PDA

View Full Version : He lives!


pHaestus
09-17-2004, 11:34 AM
I'm back guys!

Chiller is at my house now (it ONLY cools and doesn't heat so it's probably ok for wb testing but no good for radiators)

2 new 100 ohm 3 wire pRTDs arrived this morning

A replacement Epox 8K3A+ mobo arrived yesterday

It's the weekend so I SHOULD be able to get this stuff all sorted out and be up and running again. I am thinking seriously about picking up a Barton Mobile CPU for doing the water cooling kit testing; just seems to make more sense.

Anyway things are again progressing after a hellish work schedule for the last two weeks. I am teaching now so I can slack more in evenings.

bigben2k
09-17-2004, 04:59 PM
Got spill proofing resolved? :D

Etacovda
09-17-2004, 05:59 PM
lol, would siliconing an inside/out static bag to the mobo with a hole for the cpu work :D

jaydee
09-17-2004, 07:21 PM
Might be a good idea to setup a seperate loop just for leak tests before you install the blocks.

What else is on the road map? Other blocks? Just kits? I know I got nothing to send. I have had about 6 days at home in the last 3 months so not much progress on anything. All future free time will be spent on the die sims. I got another damn week in the hell hole in CA unfortunatly. Not sure whats going on when I get back. I hope to be in town for a week or 2. Bought a newer truck a month ago and havn't even put 90 miles on the damn thing yet. :( It has been sitting at the airport parking lot.

Joe
09-18-2004, 08:40 AM
I still say to stop leaks, cover the poly of any plexy covered block with a coating of Goop or flowable silicone (sensor safe kind). So if crack happens its held back buy the flexible coating over the top.

HammerSandwich
09-18-2004, 09:42 AM
Chiller is at my house now (it ONLY cools and doesn't heat so it's probably ok for wb testing but no good for radiators)
Why can't you test rads with water temp < ambient? Seems like that should work if you set the same delta.

BillA
09-18-2004, 11:03 AM
pH has yet to taste the bile resulting from rad testing;
just really difficult on these tiny things we put in computers,
AND really really difficult with ONLY 0.01°C resolution !

do the math, its a bitch
a suggestion pH, test rads only as part of a kit

Groth
09-18-2004, 12:53 PM
Tailor the heat load to your temperature resolution.

BillA
09-18-2004, 12:59 PM
here's the rub, the environmental chamber can only dissipate ~210W
with a deltaT of 10°C and a M fan, a MCR120 is about there, at high flow rates I have insufficient resolution - at low flow rates the flow rate measurement itself is a problem
(I need a moving avg over a 2 tc period)

Groth
09-18-2004, 01:16 PM
Yeah, you need a more powerful chamber, but I've alway thought it a little backwards the way you did rad tests. Rather than used a set delta-T(inletwater-inletair) and trying to determine power from difficult delta-T(inlet-outlet), I'd apply a set heatload and measure the resultant delta-T(inletwater-inletair). Delta-T(inlet-outlet) is good for a sanity check, but I don't like it for deriving results.

BillA
09-18-2004, 01:22 PM
and to apply a set heat load to a rad ?
I'm thinking wind tunnel and conditioned air
(the air flow rate is much more difficult than the coolant)

nice but outa my budget

Groth
09-18-2004, 01:58 PM
Compact closed loop, with just heater, submersible DC pump, rad, temp and pressure sensors. Measure the power in, account for secondary losses as function of temperature. Liquid version of all the die-sim concepts.

A large environment chamber would be neat, but I don't think we can raise the funds to get one for pH.

BillA
09-18-2004, 02:27 PM
no, no
if you're after Q in any fashon, you need to know the (mass) flow rate on one side or the other - ideally both

Groth
09-18-2004, 02:52 PM
Which type of Q? Heat flow, volumetric flow, mass flow...

Is C/W vs. delta-P less useful than C/W vs. flow?

BillA
09-18-2004, 04:45 PM
I meant Q as in work

since deltaP is determined vs. flow, C/W likewise can be expressed vs. either
I use both, for different things

Groth
09-18-2004, 05:23 PM
Work, as in a force times distance? How are you tying that into mass flow rates? Sure you aren't thinking power or heat-flow?

And yes, deltaP is determined by flow (and temperature dependent viscosity), hence the pressure sensors as a surrogate for a flow meter. A surrogate that is physically smaller, has lower secondary heat losses, and lower 'thermal inertia'. So why the previous emphasis on (mass) flow rate?

BillA
09-18-2004, 05:37 PM
work for a heat exchanger, an energy balance calc would require the temps and mass flow rates on both sides

the flow rate in all kits is known by means of pressure sensors, but this only works once the device has been 'calibrated' so that its pressure drop is known

your suggestion is appropriate, if only regarding the thermal 'mass' of the system, it has become so large that I have a ±0.02°C cycle that has to be addressed in all measurements
- it is more convenient to get the pd at the same time, perhaps I can no longer afford to do so, the unending quest for more accuracy and repeatability is a bugger

Groth
09-18-2004, 07:09 PM
Sorry, I though it was obvious that the rad would also need to visit a loop with a flowmeter to establish its PQ curve. Yeah, I can see the convenience of doing all the measurements at once, but eventually complexity and compromises cause problems.

As to the heat exchanger calcs, the whole deltaT, specific heat, mass flow rate thing, the idea is to avoid them, to calculate a radiator's C/W with a directly measured heat input. Sure you still gather flow/deltaT data and use it for verification, but a thermometer will never match the resolution of a multimeter.

Again, the die-sim analogy: Is it easier/more accuate to measure the electrical power input or to measure the temperature change? I'm just thinking what's good for the waterblock is good for the gander.

Your ±0.02°C cycling: your chiller and chamber controller weren't made to work in conjunction with another active temperature controller, mutual feedback is making for oscillation. Too much proportional control, not enough derivitive (the third part of the PID trinity, integral, has little effect on oscillation).

BillA
09-18-2004, 07:49 PM
are you suggesting a wb as the source ?
completely insulated, blah blah
how is that better than the chiller ? (res of dmm vs. thermometer ?)
but I will still have thermometers ??

the chiller cycle is due to the length of the system/flow rate; it is too long/slow
for sure the chamber is influencing it but its amplitude is half and the tc much longer - as near as I can tell w/o realtime monitoring (LabVIEW again)

I will welcome a overhaul of the system, needs fresh thoughts

Groth
09-18-2004, 09:40 PM
No, not suggesting a waterblock be used, just an analogy.

I'd begin with a DC pump submerged in as small a reservoir as possible. Submersible so that all the electrical power ends up as heat in the water one way or another, DC so that the pressure/flow can be varied remotely without bulky valves. Next comes the immersion heater, again with a variable power supply. Using the shortest possble tubes, you get to the sensors - a temperature probe and a pressure sensor for the for the inlet and for the outlet. The whole thing gets heavily insulated.

Start with no rad, connect the two sensor pods to each other, and apply power with the pump and heater. At this point all the heat loss is secondary, and there should be a nice linear relationship between equilibrium water temps and applied power. Voila, secondary losses as a function of temp.

Insert rad, set pump power to give desired delta-P across the rad, set heater power to get desired total heat load, set fans to as desired, wait for equilibrium. C/W = (Tinletwater - Tinair)/(PumpPower+HeaterPower-Secondary), double check using delta-T-water and a flowrate derived from delta-P-water, triple check with delta-T-air and your air flow measurement of choice.

The weak point is the secondary losses, but since its linear with respect to temperature it shouldn't be a big deal. At high flow rates and/or low heat loads, directly measured power will easily kick the ass of flow&temp based estimates. It also gives the ability to switch coolants without worrying about precisely knowing the specific heat.

Your chamber has no derivitive control, so it'll always have some temperature oscillation (both 'high' frequency as the heater cycles, and low based on the time constant of the contents). The chiller doesn't sound like it has any either. If you keep adding to the chiller loop and get the tc closer to the chamber's, you could get some fun harmonics.

pHaestus
10-11-2004, 01:58 PM
Hi guys!
Figured out something potentially very useful yesterday. My Epox 8K3A+ isn't water damaged; it has leaky caps! The 4 nearest the power regulation area are bulging and have a little dried brown juice on the top. I noticed this because the replacement 8K3A+ someone sent me has THE SAME PROBLEM. I haven't yet booted the new board, but I am pretty sure I can salvage the old 8K3A with some new caps. KnightElite has a socketA mobo graveyard at his house I think; gonna try to get 8 decent caps from his Abit boards.

I did the Greystar mod on my 1700+ and bent the pins down so there is no chance they connect with mobo. I also put some heatshrink tubing on top of the pins to serve as an insulating layer between wires and socket. I STILL see a temperature drop when the CPU goes into the socket. There is no change from in socket with ZIF closed or open though now (this was my problem a few months ago). What could this be? I will play more today. I am plugging MAX into the mobo's PSU for now and jumpering the atx supply to do this testing. I guess as long as when I retest a few wbs that the results are the same then it doesnt really matter but still it's annoying.

I thought I'd give a line item summary of my progress:

WB testing: Cathar sent me a Storm G4 and 2 AlphaCool wbs. These will be tested on the same system as the other wbs I have run. I should have that setup running (and with a fresh coolant purge/filtering too) by tonight and will have reference wb tests repeated and be ready for Storm testing by Thursday.

Kit testing: If I can get that spare 8K3A working reliably, I am going to build a separate system JUST for kit testing. Skulemate donated a Barton mobile CPU for this project (thanks!) and all I need is a stick of RAM I think (maybe a hdd).

GPU wb testing: MaxxRacer has really been instigating this one. I am still thinking about the best way to do this. My instinct is to use the GF3 Ti200 I have which has an internal diode and then measure dT across the wb to estimate W. Will think upon this for a bit prior to testing though.

The "Les and pH care about this" testing: Details shortly; gotta go get some plumbing parts for my kitchen sink :/

KnightElite
10-11-2004, 10:35 PM
As to my socket A motherboard graveyard, most of it is not at my house at the moment, but is at my parent's house in Regina. The only dead board that I have here is an Asus A7N8X-VM400, the other three abit boards, and the one ECS board are all in Regina. Don't you have another old 8k3a kicking around somewhere though? Isn't the one you're using now my old one, and you had one in there before that?

pHaestus
10-11-2004, 11:30 PM
I did have another 8K3A but I must have thrown it away. Such is life.

Update on progress:

Drained and *sol'd the test loop; leaving it running overnight to hopefully clean out any grime. I'll refill with deionized water/glycol tomorrow evening.

Set up new 8K3A+; it seems to work fine so far. I also just for kicks upgraded the monitoring PC to SP2 and cleaned up my test room quite a lot.

yoshana
10-12-2004, 12:49 AM
Sorry to hear the that replacement board may have the same issue.. :mad:

I hadn't noticed that before I sent it.. then again I did pretty much yank it and send it.. Well hope is does come in useful at any rate...

pHaestus
10-12-2004, 08:00 AM
No worries it runs fine atm.

I would guess that the caps tend to leak on this board because (a) they're cheap and (b) this board goes up to 2.1V or so on VCore. Probably those caps get a bit more of a workout than the average mobo :)

Cathar
10-12-2004, 09:33 AM
WB testing: Cathar sent me a Storm G4 and 2 AlphaCool wbs. These will be tested on the same system as the other wbs I have run. I should have that setup running (and with a fresh coolant purge/filtering too) by tonight and will have reference wb tests repeated and be ready for Storm testing by Thursday.


??? You're testing the Storm on Thursday after I repeatedly asked that the block you have not be tested because it was a very early block subject to a number of quality control issues for which there is an open recall on all non-stepped base-plate blocks, and that I was sending you a final block to test which is indicative of post-recall (stepped) blocks. Am posting the replacement Storm/G4 via express courier tomorrow. You should have it early next week.

Chew_Toy
10-12-2004, 09:50 AM
Why not just go buy replacement caps? They are not expensive and would be a known good quantity, since over the years cap failure has been a known problem for many different MB manufacturers.

BillA
10-12-2004, 10:41 AM
now Cathar, once its out of your hands . . . . .
lol, you know the (wretched) song

pHaestus
10-12-2004, 11:42 AM
jeez

By Thursday the system should be READY. I can get by without extensive filtering for most wbs; the jets on the storm (and the #5 nozzle insert from DD) require a filter though. Was just saying that I am configuring everything in such a way as to avoid any issues in the future. I don't even HAVE your package yet so I can't exactly guarantee a test on a certain date.

relax

Cathar
10-12-2004, 03:10 PM
jeez

By Thursday the system should be READY. I can get by without extensive filtering for most wbs; the jets on the storm (and the #5 nozzle insert from DD) require a filter though. Was just saying that I am configuring everything in such a way as to avoid any issues in the future. I don't even HAVE your package yet so I can't exactly guarantee a test on a certain date.

relax

You have to admit, the way you had worded it sounded very much like you were going to test the block you had on Thursday. Am presently unsure if you were trying to poke a little fun at me.

BillA
10-12-2004, 03:46 PM
want to feel 'unsure' ?
did you send a wb to JoeC ?

Cathar
10-12-2004, 03:50 PM
want to feel 'unsure' ?
did you send a wb to JoeC ?

Not yet. Am still struggling very much with the clear disparity between commentary and measured results, and this has stayed my hand every time.

What to do with a reviewer with an accurate testbed, but who says the exact opposite of what he just measured with said testbed?

BillA
10-12-2004, 04:21 PM
btsoom (beats the shit out of me)

cougem
10-12-2004, 05:44 PM
Why acronym then explain?

jaydee
10-12-2004, 06:02 PM
Why acronym then explain?
For dumb shit's like myself that had no idea what "btsoom" meant. :)

bigben2k
10-12-2004, 06:24 PM
For dumb shit's like myself that had no idea what "btsoom" meant. :)
btsoom (beats the shit out of me)
Seems self explanatory... :p

BillA
10-12-2004, 07:33 PM
it was an answer to Cathar's question, is an Aussie expected to know it ?

btsoom, like fubar, is a WWII expression - but seems not so well known
after the fact they are obvious also to the blind

cougem
next time I post one you can decipher it

Cathar
10-12-2004, 11:54 PM
I don't even HAVE your package yet

Well a new package with the stepped & revised Storm/G4 is now on its way to you. Should have it by Monday if the couriers are true to their word.

Chew_Toy
10-13-2004, 12:01 AM
it was an answer to Cathar's question, is an Aussie expected to know it ?

btsoom, like fubar, is a WWII expression - but seems not so well known
after the fact they are obvious also to the blind

cougem
next time I post one you can decipher it

I guess it is kind of like when I explained to my kids what fubar and snafu meant.

Delirious
10-13-2004, 10:20 PM
jeez

By Thursday the system should be READY. I can get by without extensive filtering for most wbs; the jets on the storm (and the #5 nozzle insert from DD) require a filter though. Was just saying that I am configuring everything in such a way as to avoid any issues in the future. I don't even HAVE your package yet so I can't exactly guarantee a test on a certain date.

relax

When you do the review what water pumps are u planning on doin the testing with? The reason i ask is i have a 1250 and am wondering if i should get something better to get max performance out of the storm on the pump end of things.

threeputt
10-14-2004, 08:58 AM
pH's testing method (http://www.procooling.com/articles/html/procooling_testing_methods_-_p.php)

hydrothruster?

Delirious
10-14-2004, 09:15 AM
Thx for the link, so he uses just one pump correct? I thought he was goin to test the storm with a few popular pumps to see which achieves the best performance?

pHaestus
10-14-2004, 09:20 AM
I vary the flow rates using valves. Never said I was going to test with multiple pumps...you are getting dangerously close to "work" instead of a hobby then

threeputt
10-14-2004, 09:33 AM
I thought this WAS "work" for you... What happened to all the "Big Money" you made???;)

BillA
10-14-2004, 09:36 AM
yea, pushing that 5/8" garden hose ?

Delirious
10-14-2004, 09:54 AM
I vary the flow rates using valves. Never said I was going to test with multiple pumps...you are getting dangerously close to "work" instead of a hobby then

ah ok, I see now. Never said that u said that you were goin to do that, just though i heard someone else say that u were.

Very nice setup, after reading about the setup i now see you put alot of work in to the reviews you do. Easily the best on the web.

pHaestus
10-16-2004, 01:50 AM
Where would I buy good replacement caps? Digikey? What spec/part number should I be looking for?

I am dedicated to actually getting my own real life PC water cooled and finished tomorrow. I need to add another hdd or two as well; they fill up so fast nowadays.

I bought a new digicam so if Cathar's Alphacool wbs ever make it I can do a proper review now

BillA
10-16-2004, 09:04 AM
I thought a digicam was all that was needed to 'do' a review ?

snowwie
10-16-2004, 12:54 PM
that, and a thesaurus

pHaestus
10-16-2004, 03:06 PM
well I clearly recall that talcum's MCR120 review was labeled "worst review ever put on the web" by someone not impressed with his camera. So I suppose you either need a digicam or a friend at a webstore/mfgr willing to take the pictures for you. I remember Bob Dyl used to do hs/fan "roundups" with just a bunch of pictures from a sponsor. Saves on shipping I guess!

murray13
10-16-2004, 05:12 PM
pH check your PM and I'll get you your caps

pHaestus
10-18-2004, 09:04 PM
Running the test system again with a new motherboard. It seems to take exactly 0.25V less VCore to yield the same W (calculated from dT water across block) compared to the old board. The MBM voltages at these settings also look very similar. I hate testing with a CPU...

Will hopefully sort this out tonight/tomorrow. Temperatures at first glance seem a lil high, but the paste application wasn't so good and usually my first couple mounts after a time off from testing are a little off anyway.

pHaestus
10-18-2004, 10:08 PM
I think it's gonna be "OK". Looks like 60 minute equilibration with same heat load as previous tests yields pretty close to the same number as earlier testing (a dT of 9.7 at 1.50GPM vs my average dT of 9.5 over 10 runs)+/-0.4 before). I'll run one or two more reps tonight and then should definately be able to run a flow test tomorrow and hopefully start with G4 soon thereafter

pHaestus
10-19-2004, 10:20 PM
Some days it's like I'm cursed. Video card in my labview/monitoring PC died and reservoir on test loop somehow got a leak. I think it's sorted out now though; running a flow test of cascade to have something to compare the G4 against. All nighter MIGHT be bad idea because I have a prof from calgary coming to use an instrument in my lab tomorrow and Thursday. I won't let you guys down though. I'll stay up as late as I can working on this

Cathar
10-19-2004, 10:35 PM
Are you going to retest any other blocks as part of the calibration thing? I remember you saying that you had changed your procedure slightly since the first test of the Cascade block which would have knocked out some of the worse mounts that dragged its average higher, and was wondering if there were other more recently tested blocks you were calibrating against?

pHaestus
10-19-2004, 10:40 PM
Yes. What I will do is retest the Cascade, run the G4, and show the test results vs. cascade, and then retest a few other blocks. It may be that I have to do a bit of an adjustment now that I've changed mobos to have the G4 results generally comparable. If so I am hoping full flow tests of 3 wbs will do the trick.

Bottom line is I am trying to do what I must to get a good G4 review up (and the wb winging down to JoeC) and then I'll worry with any minor adjustments I need to make to put the G4 on my interactive comparison page.

Cathar
10-19-2004, 10:59 PM
Thank you for all the effort you're putting in this Phaestus.

I'm just biting my nails, that's all.

pHaestus
10-19-2004, 11:10 PM
Actually the numbers look very sensible on the old vs. new Cascade testing now. I'll post a comparison graph of old and new data before bed (another 90 minutes or so). I'll then install the G4 and just check for a minute before sleep.

I ran the cascade first because I figured it'd be the block everyone wanted to compare to anyway...

pHaestus
10-20-2004, 12:38 AM
Probably bad form to post such preliminary results but what the hell it's late and everyone's eager to see...

http://phaestus.procooling.com/forcathar.jpg

That's a whopping 1 point from one really quick test mount on the G4. That means paste hasn't fully settled and I haven't gotten as familiar with mounting the wb as I will tomorrow. So don't go crazy (good or ill) yet.

Cathar
10-20-2004, 01:37 AM
Probably bad form to post such preliminary results but what the hell it's late and everyone's eager to see...

<snip image>

That's a whopping 1 point from one really quick test mount on the G4. That means paste hasn't fully settled and I haven't gotten as familiar with mounting the wb as I will tomorrow. So don't go crazy (good or ill) yet.

Hmmm, would've hoped that the 8C mark would've gotten broken at the least on your testbed - but that's cool - still early days...

Am somewhat more interested in what you end up measuring for the lower end flow rates too as that's something that I really concentrated on, somewhat at the expense of top-end cooling performance. My goal for this block was more about realistically achievable high-end performance even with quite weak pumps, rather than the fairly steep sloping Cascade/White Water performance curves. Very interested to see what the 0.25-1.0gpm points will be.

pHaestus
10-20-2004, 09:09 AM
Yea like I said don't sweat it yet. It's so much thicker than the Cascade that I suspect it'll take a little fiddling to get good mounts with the old hardware. I didn't measure, but are the new bolts a little longer than the old ones?

//edit: Re the low flow performance: Note the 0.25GPM point. I'm slowly following the requests of the readers

Cathar
10-20-2004, 03:09 PM
Yea like I said don't sweat it yet. It's so much thicker than the Cascade that I suspect it'll take a little fiddling to get good mounts with the old hardware. I didn't measure, but are the new bolts a little longer than the old ones?

//edit: Re the low flow performance: Note the 0.25GPM point. I'm slowly following the requests of the readers

25mm thick (Storm) vs 20.5mm thick (Cascade). I seem to have worked out the mounting though by very firmly holding the block against the CPU while feather tightening the springs, before clamping down.

The new bolts are 75mm long, as opposed to the old bolts which are 65mm long. The extra height along with the mounting bracket would mean that the 65mm bolts are just fractionally too short for P4 systems. The 65mm long bolts are still fine for Socket A though.

Was very concerned yesterday - I mounted the block and temps started to skyrocket, and I'm thinking WTF? Quickly pulled it off and I have no idea how but something very small and hard and whitish was stuck in the Arctic Ceramique that created a divot in the block's base and was leveraging the block away from the CPU. Quite frankly I was surprised that the CPU core didn't crack.

pHaestus
10-20-2004, 10:23 PM
Remount suggests G4 is indeed a little better than the point last night suggested. Hopefully a full flow test to follow.

Cathar when you say you designed for low flow performance, did you mean >1LPM? At 0.25GPM the block loses all its performance edge vs Cascade if tonight's run is representative.

Cathar
10-20-2004, 11:25 PM
Cathar when you say you designed for low flow performance, did you mean >1LPM? At 0.25GPM the block loses all its performance edge vs Cascade if tonight's run is representative.

I kinda meant Eheim 1046 sort of pumping, which appears to be the quintessential "low-flow" pump. With the Eheim 1046 in a full 1/2" setup a flow rate of around 2.5LPM (0.65gpm) is predicted for the Storm/G4, as opposed to about 3LPM (0.8gpm) for the Cascade.

I guess by low-flow I really meant around 1.5-3LPM. Below 1.5LPM is really "very low flow" and is not normally achievable unless one artificially chokes the crap out of the system with needless restrictions. I only properly tested it down to around 1.5LPM (0.4gpm) myself with performance falling away fairly rapidly below that point, but I was unconcerned with flow rates that low.

I suspect that some of it is explained by the fairly small die-size of the T'Bred B. The Storm's cooling patch is around 400mm² in size, and the T'Bred B is just 88mm². If you're feeding the block 1LPM, only about 0.3LPM of that is actually being used to cool the CPU. Could probably shut off 2/3rds of the jets and still achieve the same cooling performance with 1/3 of the jets at 0.3LPM as with all of the jets at 1LPM. I had to keep in mind that there are up to 250mm² CPU dies out there which then are covered with IHS's that spread the heat out over an even larger area.

Also possibly explained by the G4 only having about half the internal surface area (within the primary cooling patch) of the Cascade available for convection. As the primary impingement effects start to lose efficacy the Cascade has more surface area to fall back on to prop its performance up. Still we are talking about, IMO, super-low flow rates that one has to do a lot wrong to get that low.

pHaestus
10-20-2004, 11:41 PM
Let me post this first and then I'll try to see how it relates to your post:

http://phaestus.procooling.com/temp/G4prelim.jpg

//edit: That was one complete flow test run as soon as I felt I had a "good" mount. Final performance may move up or down a bit when I run 4-5 reps at 1.5GPM tomorrow.

Cathar
10-20-2004, 11:44 PM
Let me post this first and then I'll try to see how it relates to your post:


Cheers. I was seeing slightly better low-flow performance than either of the Cascade's, but again I suspect that may be due to the differing die sizes being used.

pHaestus
10-20-2004, 11:54 PM
I'm going to guess that the viscosity of my coolant starts to become a bigger issue at really low flow rates. I still use 25% ethylene glycol/75% RO water (by volume). I would have intuitively thought that the little center pin in the Storm design would offer a boost in turbulence under such low flow conditions compared to the cascade. I ran the 0.25 GPM point at the very beginning of testing and again at the end and it was very reproducible (within 0.05C dT). Your explanation of die size being important makes sense.

I have a Morgan core Duron and a mobile Barton here too. I am itching to run tests with 3 difft CPUs at (a) same MHz/VCore and (b) same dT measured across wb) and look at effect of die size and cache size on performance. I am personally wondering which is bigger effect on the diode temp: Size of die (aka heat density) or size of cache (aka diode placement relative to the hottest CPU bits). Any bets? Would also be good to compare performance effects on a wb like the cascade vs a block like the MCW6000.

Cathar
10-21-2004, 12:11 AM
I am personally wondering which is bigger effect on the diode temp: Size of die (aka heat density) or size of cache (aka diode placement relative to the hottest CPU bits). Any bets?

Since there would undoubtedly be smallish variations in the cooling effect in different locations of the cooling patch, how about the relative location of the die's thermal probe to the cooling effect of a jet?

I had a T'Bred B here for a short while (someone brought it around) and the on-die diode read somewhat higher for it than the Barton-dies, even with what should have been a lower heat load per unit of surface area. Quick test and totally uncalibrated in any fashion though, so it could just be a bogus data point.

After a lot of testing I find myself asking myself more and more often - just what is it that I'm actually measuring here?

Re: die-size, the simulations show that the smaller die sizes with high heat densities favor a slightly thicker base-plate, while the larger dies with lower heat densities favor a thinner base-plate. So hard to strike an acceptable balance.

Now that I have the G5 middle plates ready I'll try to get one off to you ASAP. With a jet/cup density 70% higher than the G4, but same basic cooling effect, it may provide the necessary extra information to further speculate on.

pHaestus
10-21-2004, 12:16 AM
Is it possible to track that first package? It's been quite a while now... FYI, I would expect this G4 review to get posted about 1-2am on Friday my time (so around 26 hrs from now). I figure that'll give readers plenty of time prior to the Procooling server upgrade Monday :)

Cathar
10-21-2004, 12:28 AM
Is it possible to track that first package? It's been quite a while now... FYI, I would expect this G4 review to get posted about 1-2am on Friday my time (so around 26 hrs from now). I figure that'll give readers plenty of time prior to the Procooling server upgrade Monday :)

Not possible to track the package unless I put in an insurance claim, at which point it becomes somewhat moot because it's not like I can just get more Alphacool blocks and re-send them. More likely that they are still stuck in Canadian customs somewhere though. Have only had a single international parcel get lost in over two years now, and that one was uninsured so I have no idea what really happened to it.

Thank you again for all the work Ph.

pHaestus
10-21-2004, 12:44 AM
No problem. Thanks for the chance to test the G4 out. Just as an aside how does pricing of the G4 compare to the Cascade SS? I'd assume the G4 is quite a bit cheaper?

Cathar
10-21-2004, 01:00 AM
No problem. Thanks for the chance to test the G4 out. Just as an aside how does pricing of the G4 compare to the Cascade SS? I'd assume the G4 is quite a bit cheaper?

G4 = $116 AUD (Australian dollars)
SS was $255 AUD

G5 will be around the $240 AUD mark.

$1 AUD =~ $0.735 USD =~ $0.915 CAD at present.

BillA
10-21-2004, 08:31 AM
nice work Cathar, interesting
and Thanks also to the tester

nicozeg
10-21-2004, 09:27 AM
This is becoming complex!

To the usual testing @ full flow rate range, plus several mounts.... You now have to add different liquid solutions and die sizes! pheeew! :eek:

How you dare to sleeep! :evilaugh: :p

pHaestus
10-21-2004, 11:02 AM
Well.. Looking at different solution conditions on G4 performance will depend on how quickly Cathar wants me to send this wb to JoeC I guess. It is an interesting question though and one I'd guess Les could model reasonably well (to keep me honest).

BillA
10-21-2004, 11:12 AM
well, since the viscosity changes also with temp, you could run at 4 or 5 different cooling temps and develop the curve . . . . .

Les
10-21-2004, 12:09 PM
Was hoping Cathar might present his model. This allows him to predict:-
"The G1, of which I had one made, simluated out to around 1C worse than a White Water, and in reality performed about 1.5C worse than the White Water. The G2 I have not had made up. In reality it should perform about on par with a White Water" (link) (http://www.employees.org/~slf/lrwb/)
I am unable to connect "h(convection)" with "h(effective)" for anything other than channeled blocks.( Using the definition of "h(effective)" used here (http://www.electronics-cooling.com/html/2004_may_cc.html)

Additionally I am bemused by "I am acutely aware that one of the principles in the operation of the G4 is more heavily affected by the kinematic viscosity of the liquid than some other designs." - from interchanges here(post 196 to 227 (http://forums.procooling.com/vbb/showthread.php?t=9144&page=10).

BillA
10-21-2004, 03:40 PM
Female hormones decide research skills of male scientists!: (http://news.newkerala.com/india-news/?action=fullnews&id=38039)
hmmm

pHaestus
10-21-2004, 03:53 PM
that's... odd.

My index finger is shorter than my ring finger by 12mm.

BillA
10-21-2004, 04:12 PM
did the ring make it grow ?
or stunt the index finger ?

pHaestus
10-21-2004, 04:26 PM
hmm sometimes I chew at the band

The basest of instinct (see animals chewing off leg to escape traps)

Thus far no luck. Truthfully I stopped wearing the band when I lost ~40lbs on Atkins. It kept falling off and I constantly fiddled with it.

Delirious
10-21-2004, 05:00 PM
hmm sometimes I chew at the band

The basest of instinct (see animals chewing off leg to escape traps)

Thus far no luck. Truthfully I stopped wearing the band when I lost ~40lbs on Atkins. It kept falling off and I constantly fiddled with it.

LMAO!

Nice work so far pHaestus, cant wait to read the review.

pHaestus
10-21-2004, 06:34 PM
Cathar when you see this please PM me with your mfgr spiel on design considerations/objectives/etc.

BillA
10-21-2004, 06:40 PM
sloppy terminology, would gospel not be more appropriate ?
j/k - so I don't get pounded

pHaestus
10-21-2004, 06:51 PM
perhaps gospel in your case; he's your god right?

BillA
10-21-2004, 07:17 PM
you are baiting me, and I am showing control
lol

pHaestus
10-21-2004, 07:27 PM
you're baiting yourself :)

Cathar
10-21-2004, 07:37 PM
My index finger is, in fact, somewhat longer than my ring finger, and as a befits my hormonal (im)balance I'm throwing a hissy fit due to the above comments, and sulking in an episode of self-loathing.

BillA
10-21-2004, 07:44 PM
never fear, rescue here

I found that keeping my fingers parallel I could bend them to make even my pinkie as long as my index
and the other way my index longer than the slave finger

we're all frigging geniuses here

killernoodle
10-21-2004, 07:46 PM
^ LOL

My ring finger is longer than my index, which explains why I just gave Cathar two black eyes, chugged a nice cold Miller, and cruised out in my shiny IROC.:D

Stupid science...

Cathar
10-21-2004, 09:12 PM
Cathar when you see this please PM me with your mfgr spiel on design considerations/objectives/etc.

May as well include it here:

Objectives:

1) Make a higher performing block than the Cascade
2) Make a block that doesn't require as high flow rates to achieve excellent cooling performance, such as would be suitable for use with an Eheim 1046 and 3/8" ID tubing for which 2-3LPM (0.55-0.80gpm) would be the typical sort of flow rate.

Commentary:

The Storm design is more of a pure-impingement design. The older Cascade design was not strictly a pure impingement design. The jet width in the Cascade design in comparison to the cup width was too large for a proper jet impingement effect to occur, with the Cascade being more of a "super turbulent mash-in-a-cup" effect. True jet impingement needs much wider cups in comparison to the jets for the impingement effect to properly develop.

I had experimented with pure-impingement designs both during and after the Cascade design's development with limited success, but was generally unable to match the Cascade design's strong performance curve over both high and low flow rates. As flow rates got lower, performance of the true jet impingement approach suffered quite heavily in comparison to the dense high-surface area Cascade approach, and at higher flow rates the true jet impingement approach could at best only match the Cascade design.

I tried a modified impingement approach which forms the basis of the Storm design. The Storm design implements jet disruptors to turbulate in incoming jets prior to striking the cup's base. The disruptors also offer extra surface area to assist in lower flow applications where the relatively broad cup-base expanse (in comparison to the Cascade) becomes a liability as thermal convection efficiency falls and the block design then has to rely more upon abundantly available surface area that is easily reached through metal conduction to sustain good low-flow performance.

The base-plate thickness for the Storm was increased slightly over the Cascade to further assist the pure impingement approach in lower flow scenarios, but again a balance had to be struck here as today's modern largish CPU dies when covered with an IHS favor a thinner base-plate approach. I strove for the best balance across what is really a fairly large variation in CPU die sizes, being anything from the bare die ~80mm² AMD CPU's, right up to the >220mm² IHS covered Intel CPU dies. I still wated to retain high-end performance for those using pumps strong enough to take full advantage of the Storm design's cooling power.




----------------

There's so much more that I would want to write about, but I figured that's probably blabbing on for long enough.

pHaestus
10-21-2004, 09:17 PM
the more you blab the less I have to write

Delirious
10-21-2004, 09:22 PM
the more you blab the less I have to write

Isnt that a good thing?

Cathar
10-21-2004, 09:22 PM
Was hoping Cathar might present his model. This allows him to predict:-
"The G1, of which I had one made, simluated out to around 1C worse than a White Water, and in reality performed about 1.5C worse than the White Water. The G2 I have not had made up. In reality it should perform about on par with a White Water" (link) (http://www.employees.org/~slf/lrwb/)
I am unable to connect "h(convection)" with "h(effective)" for anything other than channeled blocks.( Using the definition of "h(effective)" used here (http://www.electronics-cooling.com/html/2004_may_cc.html)

Additionally I am bemused by "I am acutely aware that one of the principles in the operation of the G4 is more heavily affected by the kinematic viscosity of the liquid than some other designs." - from interchanges here(post 196 to 227 (http://forums.procooling.com/vbb/showthread.php?t=9144&page=10).

I had later said that the predictive models I was using would not satisfy rigorous analysis, so I decline to post it because I know enough to know that it'll get shot down. If you want to treat my work as purely empirical, that's fine with me. If someone wants to grab the block, look at what I've done, and attempt a model for it, that's fine with me too.

Re: kinematic viscosity - jet impingement efficacy is more heavily impacted by higher kinematic viscosity than a mini/micro-channel approach. Various convection calculators seem to show that to be the case. The turbulator approach with the storm is of a similar nature. The goopier the liquid is, the more the liquid will just flow around the turbulator, rather than be disrupted by it, effectively nullifying much of its impact. That is just a theory of mine though. I can't prove it either way but it feels right to me.

Cathar
10-21-2004, 09:42 PM
the more you blab the less I have to write

Okay, I also wanted to briefly talk about the jets:

Commentary Part 2:

I paid particular attention to the jets for the Storm design, to both optimise the jet intakes in an easily machinable manner that allowed for minimised pressure loss from accelerating the incoming flow into the jet array, and also offered as close to a one-size-fits-alls level of restriction that uses the pump's pressure to accelerate the water to as high of a velocity as possible, but also without choking flow rates excessively.

Across a very broad range of centrifugal pumps, both actually tested and simulated, the resultant jet velocities are typically around 85-95% of the maximum jet velocity any commonly used centrifugal water pump can sustain (after other cooling loop pressure drop factors are taken into consideration). It is possible to increase jet velocity further, but the gains are very small and result in volumetric flow rates falling rapidly which is something that I specifically wanted to avoid because other waterblocks may exist within a user's system which may not act gracefully with a sharp reduction in volumetric flow rates. With any given pump, jet velocities for the Storm are typically about 1.7x that of what occurs within the Cascade when given the same pump.


In all I feel that I found the best balance that suited the widest range of pumps and system scenarios, while providing the highest level of jet velocity as it important for the jet impingement effect to work at its best, and also keeping volumetric flow rates high enough to not significantly impact the performance of other blocks in the system.

Cathar
10-21-2004, 09:48 PM
Okay, I think I'm done.

pHaestus
10-21-2004, 10:01 PM
me too (with testing). Came home at 4pm so I could finish up :)

Your review is first one w/ new camera btw cathar

BalefireX
10-21-2004, 10:45 PM
Foiled by VBB... pH's PM box is full and thus he cannot retrieve graphs..
http://monster911.com/stormg4.zip

pHaestus
10-21-2004, 11:11 PM
How can my PM box be full? I run this goddamn place!

Gonna have to ask Joe about an upgrade to 1000 PMs

greenman100
10-21-2004, 11:24 PM
How can my PM box be full? I run this goddamn place!

Gonna have to ask Joe about an upgrade to 1000 PMs


best post ever

pHaestus
10-22-2004, 01:22 AM
http://www.procooling.com/reviews/html/storm_g4_waterblock_review_-_1.php

On to different things now (finishing up dP setup, kits, and GPU wb testing I guess)

Cathar
10-22-2004, 01:49 AM
http://www.procooling.com/reviews/html/storm_g4_waterblock_review_-_1.php

On to different things now (finishing up dP setup, kits, and GPU wb testing I guess)

Many thanks Phaestus.

/me breathes a large sigh of relief.

nicozeg
10-22-2004, 01:58 AM
me too (with testing). Came home at 4pm so I could finish up :)

Your review is first one w/ new camera btw cathar

Ahhh, that explains my only concern with your review:

You need to learn how to do white balance in the new digicam! :P

Les
10-22-2004, 03:13 AM
Deleted

Reposted in http://forums.procooling.com/vbb/showthread.php?p=127844#post127844

pHaestus
11-01-2004, 01:39 PM
I am pretty sure the AlphaCool wbs and the first G4 block actually made it to my house today. Wife called to let me know that a package from Australia with a postmark dated late September had arrived. GG customs I'll update tonight.

Cathar
11-01-2004, 04:18 PM
I am pretty sure the AlphaCool wbs and the first G4 block actually made it to my house today. Wife called to let me know that a package from Australia with a postmark dated late September had arrived. GG customs I'll update tonight.

The Storm/G5 should be with you in very short time as well. On-line tracking tells me that it's not very far away if it's not already at your house. Maybe in the morning? :shrug:

pHaestus
11-01-2004, 06:16 PM
WTH size thread is this one wb? Gonna have to be creative in plumbing it I guess...

Cathar
11-01-2004, 06:32 PM
WTH size thread is this one wb? Gonna have to be creative in plumbing it I guess...

Your guess is as good as mine. I sent it how I got it. The silver-topped block has the barbs that I into it myself. It didn't originally come with barbs either. Still, with the 1/2" OD barbs I put in I don't think anyone could complain that I did the block a dis-service by fitting restricted barbs, with the barbs I fitted showing the block in about the best light possible. It uses 1/4" BSP threads, and I was pretty sure that being in North America you might have a bit of trouble finding suitable barbs for it, so I fitted the best ones I had for you.

You may also have to get a little creative with figuring out how to mount the silver-topped block and achieve a consistent and known mounting pressure. I would suggest fabricating up a top-plate for it. Increasing mounting pressure does help things significantly.

Cathar
11-03-2004, 06:31 AM
Phaestus, you may want to check with your local post office. On-line tracking tells me that delivery of the G5 was attempted today, but that no-one was there to receive it.

pHaestus
11-03-2004, 07:20 PM
Block is here. These silver blocks look so damn cool. It'll be hard to part with it :(

JWFokker
11-06-2004, 12:54 AM
Despite already owning a G4, I'm still excited to see new waterblock reviews. I need to find more hobbies.

pHaestus
11-07-2004, 02:59 AM
http://phaestus.procooling.com/g5prelim.jpg

bling bling

BalefireX
11-07-2004, 03:09 AM
Well well.. Looks slightly more restrictive, but we're into the 7s!

Cathar
11-07-2004, 04:09 AM
So pleased to know that I was not going insane when I saw the results at my end, which I must say stunned me slightly. At the same pumping power I was seeing a ~1.4-1.5C gain with the G5 over the G4 on my testbed. Historically Phaestus has always measured about bang on 40% of the differences that I measure between blocks, mostly due to the higher heatloads that I test with, and also due to different temperature measuring mechanisms.

40% of 1.5C is 0.6C, and it appears Phaestus has measured perhaps very close to a 0.6C difference here, which pretty much falls exactly in line with what I was hoping he would see once the testbed differences were factored in.

Thank you once again for the testing Phaestus.

Etacovda
11-07-2004, 04:40 AM
Sweet jesus thats impressive, amazing work Stew - I'm looking forward to your waterchiller project + this in conjunction on the 3.0ghz XP-M.

cougem
11-07-2004, 01:14 PM
Oh. Baby.

So cathar, when and how much? :)

jaydee
11-07-2004, 03:50 PM
Very cool Cathar. How much better can it get?

Cathar
11-07-2004, 05:46 PM
How much better can it get?

The block that Phaestus had is hand-lapped and probably could be lapped a little better.

The G5 totally threw out any models I had in predicting where the design was going in terms of performance. The gain that I measured between the G4 and G5 is almost twice as large as I was expecting from the model that I had formed.

It is possible to me to manufacture a G6 design variant, and this is something that I will now look into doing. It will be terribly cost prohibitive though but I enjoy chasing that last little bit of performance.

A hand-waving prediction of how much better? Given a proper machine lap (when assembled to remove any chance of assembly based warping), and I would say that a G6 in silver could probably be around 0.3C better on Phaestus's testbed, or be around 0.7C better on my testbed.

It is even possible to manufacture up a G7 design by special ordering in the tools fine enough to do the job, but I suspect that at this level that improvements will be small as the increased jet restriction of the design at this level may start to play a major role. It may also be difficult to accurately control the machining of the structures that gives the block its boosted performance at this level, especially in a soft gummy metal like silver is.

It's all becoming somewhat of an academic exercise at that level though. The machining costs get so high that even if it were possible to build a block based on the Storm design that cracked the 7C mark on Phaestus' testbed, that it simply would've been better to just go out an buy a Vapochill instead.

pHaestus
11-07-2004, 05:52 PM
It'd be interesting to compare the G4 to G5 when running chilled water wouldn't it?

greenman100
11-07-2004, 05:53 PM
It's all becoming somewhat of an academic exercise at that level though. The machining costs get so high that even if it were possible to build a block based on the Storm design that cracked the 7C mark on Phaestus' testbed, that it simply would've been better to just go out an buy a Vapochill instead.


unless someone had access to a CNC machine and you gave out the code

doubtful at best, eh

BalefireX
11-07-2004, 05:57 PM
It'd be interesting to compare the G4 to G5 when running chilled water wouldn't it?

Still have your MCWChill-452 setup laying around? :D

jaydee
11-07-2004, 06:23 PM
So when is your chiller and what not going to get up and running pH? If not to far off (3 months?) I might wait to send you my next 4 blocks. If longer I will send them as I get them done (1 done 3 to go).

Cathar
11-07-2004, 06:51 PM
Thought that this graph may be of some help to people. If you're color blind, grab the nearest female. Only a 1 in 100000 chance that she'll be color blind too. 1gpm = 3.785lpm.

http://www.employees.org/~slf/curves/pumps/g5flow.png

nikhsub1
11-07-2004, 07:30 PM
On that chart the RD-30 would be what, about 9.5lpm or so? On a side note, once the RD-30 is restricted (I only ran it with 2M tubing) it gets much quieter, I artificially restricted the outlet by pinching the tubing, whoooosh noise is gone when doing so. Will see when it is in the loop.

Am very excited about getting the new pump and G5 in the rig :D

Cathar
11-07-2004, 07:50 PM
I'll do up a quick graph for you.

Yes, all of the high pressure Iwaki's need a good amount of restriction on the outlet. If the outlet is unrestricted they will cavitate and make a horrible racket. My MD-30RZ is noisy as all heck when unrestricted, but the moment I stick a block on it it quietens right down to a simple low volume whine.

nikhsub1
11-07-2004, 07:52 PM
I'll do up a quick graph for you.

Yes, all of the high pressure Iwaki's need a good amount of restriction on the outlet. If the outlet is unrestricted they will cavitate and make a horrible racket. My MD-30RZ is noisy as all heck when unrestricted, but the moment I stick a block on it it quietens right down to a simple low volume whine.
I would bet this will be an almost identical whine... except the RD-30 will give you 30RZ 60HZ specs, can be overvolted for even more or undervolted, er well you already know all this :D

alexwai
11-07-2004, 08:02 PM
Cathar, Can you also add the line for RD-20. Thanks

Cathar
11-07-2004, 08:02 PM
http://www.employees.org/~slf/curves/pumps/g5rd30.png

Cathar
11-07-2004, 08:05 PM
Cathar, Can you also add the line for RD-20. Thanks

Will re-edit above graph - wait a moment

nikhsub1
11-07-2004, 08:08 PM
Cathar, Can you also add the line for RD-20. Thanks
Use the MD-30Z 50Hz... identical specs as the RD-20, 8M max head and 15lpm max. Should translate to the same curve.

pHaestus
11-07-2004, 08:09 PM
JayDee: I have a few chillers. One's for getting water COLD; the other one is for testing and holding temps constant.

As for wbs, I have had enough with the current testbed. I am testing the AlphaCool wb (both if I can find barbs for the lil one) and I am then "freezing" wb tests. Will perfect the dP measurements on all the blocks I've run so far, post that, and then migrate to the die simulator (assuming I get it built ever).

Cathar
11-07-2004, 08:11 PM
Use the MD-30Z 50Hz... identical specs as the RD-20, 8M max head and 15lpm max. Should translate to the same curve.

Not quite. The 30RZ has stronger mid-range pressure than the RD-20.

I've replotted the graph to include the RD-20. Reload the page to get the new graph.

As you can see, 7.5LPM predicted with the RD-20, vs 8.1 LPM with the 30RZ @ 50Hz.

alexwai
11-07-2004, 08:19 PM
Will re-edit above graph - wait a moment
Thanks for your work.

pHaestus
11-07-2004, 08:39 PM
BalefireX you have a PM :)

pHaestus
11-07-2004, 08:50 PM
Cathar
Would it be possible to use that graph in the review? Would I think be helpful to the readers.

Cathar
11-07-2004, 08:57 PM
Cathar
Would it be possible to use that graph in the review? Would I think be helpful to the readers.

Indeed, I was hoping that you would. Be my guest.

jaydee
11-07-2004, 09:00 PM
JayDee: I have a few chillers. One's for getting water COLD; the other one is for testing and holding temps constant.

As for wbs, I have had enough with the current testbed. I am testing the AlphaCool wb (both if I can find barbs for the lil one) and I am then "freezing" wb tests. Will perfect the dP measurements on all the blocks I've run so far, post that, and then migrate to the die simulator (assuming I get it built ever).
Sounds like a plan. Will give me more time to build the other 3 blocks. If the parts I get work to fix my mill I may still be able to mill up a die sim. I am afraid I might not be able to hold tight tolerances though. :shrug:

Cathar
11-07-2004, 09:09 PM
Here's the same graph with GPM units:

http://www.employees.org/~slf/curves/pumps/g4g5flowgpm.png

pHaestus
11-07-2004, 09:14 PM
Last question: PRICE?

Cathar
11-07-2004, 09:16 PM
Last question: PRICE?

Hmmm, around the $250 AUD mark. Not fully fixed, but very close to that.

Availability will be better than the Cascade SS as well.

pHaestus
11-07-2004, 09:32 PM
Cathar: PM me any corrections or clarifications (if any) that you'd like to see in the review.

Everyone else:

http://www.procooling.com/reviews/html/storm_g5_waterblock_review_-_1.php

jaydee
11-07-2004, 09:35 PM
That is just pure class! Beautiful block!

http://www.procooling.com/reviews/assets/images/g5angle.jpg

Cathar
11-07-2004, 09:37 PM
Cathar: PM me any corrections or clarifications (if any) that you'd like to see in the review.

Everyone else:

http://www.procooling.com/reviews/html/storm_g5_waterblock_review_-_1.php

Might want to include the GPM graph I provided above, in place of the LPM graph, since the units are then the same as on the main comparison graph.


Other than that, looks all good.

Many thanks again Ph.

[Edit: Oh - realised that it's just the thumbnail that's in LPM, but the link goes to the GPM graph. May want to clarify that]

pHaestus
11-08-2004, 08:13 AM
I have 1 (possibly 2) wbs I will test the next few days (the Alphacool is plumbed in now). Then I am going to basically put an end to wb testing as it currently exists at Procooling. I will first off collect dP measurements for many commercial wbs already tested and update the Pro/Testing page with an interactive dP vs. Q and dT vs. dP graphs. The test system will be converted to complete kit testing. Finally, I will find a machinist and devise a final design for the die simulator. Fun stuff. I am horribly behind in all my Procooling projects (sorry but shit happens).

HammerSandwich
11-08-2004, 02:09 PM
Phaestus, did you ever get any numbers for the Storm with straight water?

pHaestus
11-10-2004, 08:58 PM
HS: nope. I have a G4 and a G5 here now tho for good and I am going to be playing with them as time goes on.

All: I am testing the AlphaCool wb that Cathar sent right now. Not a bad wb (though restrictive) but the holddown mechanism is less than impressive. If this wb had a 4 hole top it'd be so much nicer...

Cathar
11-10-2004, 09:16 PM
The NexXxoS XP is a pretty nice block, huh Phaestus?

I had the same opinion of the hold-down mechanism. Fiddly as all crap and if I got the block on without getting thermal paste everywhere, it was a minor miracle.

Once the block is installed though, it does perform well. Well enough, I thought, to keep all us English-speakers honest.

pHaestus
11-10-2004, 10:53 PM
I am hesitant to post the preliminary results as I have NO idea how much force is applied with the allen bolt. My best guess is "more than I apply on the 4 hole wb tests" because the first mount I had the block a little bit off center and got horrible temps. I am pretty sure I cranked down a bit more than is normal the second time. But yea on the dT vs flow rate curve the NexXoS XP is a very impressive block.

I should be able to post results tomorrow

Cathar
11-10-2004, 11:49 PM
Yes, I had the exact same issues with not knowing the mounting pressure. Very easy to apply quite staggering levels of mounting pressure with the mechanism as provided. Indeed that does bring into question the whole mounting pressure vs performance mechanism.

Can't remember where I saw it, but someone recently tested a block where the stock mounting pressure as provided by the manufacturer was around 80lbs and on AMD type die. After backing the pressure off to normal levels (~20-30lbs) temps increased by around 1C or so. I personally see this sort of behavior myself with varying mounting pressures, and there are a host of papers on-line that have researched the same phenomena.

Additionally the block's pressure drop, and it's impact on the flow rates of real-world pumps, is something that cannot be ignored. I took the liberty of plotting a full system PQ graph for the XP as it helped to explain the behavior I was seeing from it.

Performance for any particular flow rate is exceptional, however the problem is actually finding pumps strong enough to give that level of flow. Trying to achieve anything much above 1gpm is no small task...

http://www.employees.org/~slf/curves/pumps/xpgpm.png

HammerSandwich
11-10-2004, 11:59 PM
Performance for any particular flow rate is exceptional, however the problem is actually finding pumps strong enough to give that level of flow. Trying to achieve even 1gpm is no small task...
After reading this, I expected worse. Basically, anything better than an Eheim does the trick.

EDIT: Cathar edited his post while I wrote the above. His new wording obviates my statement.

Cathar
11-11-2004, 12:03 AM
After reading this, I expected worse. Basically, anything better than an Eheim does the trick.


Comparison flow-rate vs pump graph of the XP vs the MCW6000 to demonstrate the flow-rate thing more clearly:

http://www.employees.org/~slf/curves/pumps/xp-mcw6-gpm.png

HammerSandwich
11-11-2004, 12:06 AM
Yeah, I'd compared it to the graph in the G5 review, where the difference is somewhat less.

Les
11-11-2004, 03:15 AM
I am hesitant to post the preliminary results as I have NO idea how much force is applied with the allen bolt. My best guess is "more than I apply on the 4 hole wb tests" because the first mount I had the block a little bit off center and got horrible temps. I am pretty sure I cranked down a bit more than is normal the second time. But yea on the dT vs flow rate curve the NexXoS XP is a very impressive block.

I should be able to post results tomorrow

Perhaps interestingly Incoherent's C/Ws for wbs (http://forums.procooling.com/vbb/showpost.php?p=129607&postcount=53) are expected to be mounting pressure independent.

pHaestus
11-11-2004, 11:04 AM
Wouldn't that require 2 temp probe holes drilled in every wb bp though? Even 1 hole drilled in a bp like a cascade or WW or Storm would be pretty difficult to manage.

Les
11-11-2004, 12:25 PM
Just indicating a possible virtue of the Flux-Block as heat-source .
Not suggesting adoption when using a cpu as a heat-source.

pHaestus
11-11-2004, 02:25 PM
I am currently testing the last block I'll run on a CPU. I hope so, anyway. I am terribly conflicted on this because I think that mounting issues and secondary effects are really important. With several hardware failures and a continuous progression of new CPU types though I don't think this sort of testing is convenient or particularly relevant any more.

pHaestus
11-11-2004, 08:52 PM
http://phaestus.procooling.com/temp/alphacool-prelim.jpg

Preliminary data. Block is very good but also very restrictive. I will finish this review either tonight late or tomorrow.

Cathar
11-11-2004, 10:13 PM
http://phaestus.procooling.com/temp/alphacool-prelim.jpg

Preliminary data. Block is very good but also very restrictive. I will finish this review either tonight late or tomorrow.

Having assessed the same block myself I wholeheartedly agree with those results.

On a pure flow vs performance basis it is extremely impressive, but its level of flow resistance pegs that impressiveness back in real world use. In real-world pumping power vs performance tests I personally found it to slot in about bang on mid-way between a Cascade and a Cascade SS pretty much right across my pumping power test range.

Definitely an impressive block though. Something for the main-stream manufacturers to keep on eye on because it does outperform most any commercial US-manufactured block across a broad range of pumping powers.

It also pretty much dismisses the whole high-flow vs low-flow myth too. Having had tested it myself I knew that was the case, but had kept my mouth largely shut on the matter until it was independently tested.

brucoman
11-11-2004, 10:22 PM
On that chart the RD-30 would be what, about 9.5lpm or so? On a side note, once the RD-30 is restricted (I only ran it with 2M tubing) it gets much quieter, I artificially restricted the outlet by pinching the tubing, whoooosh noise is gone when doing so. Will see when it is in the loop.

Am very excited about getting the new pump and G5 in the rig :D

my RD-30 is whisper quiet with the 2 Storm's/2-342 rad at 24v, drawing only 0.3 amps is odd....

pHaestus
11-11-2004, 11:04 PM
Cathar:
Was thinking I'd send it to Bill to play with when I was done with it (if he's interested). Is that ok?

Cathar
11-11-2004, 11:39 PM
Cathar:
Was thinking I'd send it to Bill to play with when I was done with it (if he's interested). Is that ok?

I personally have no problem with that at all. However it may be courtesy to fire an email off to Alphacool (Tanner the name of the guy whom I was talking to from memory). He was happy to provide to Procooling by way of me, but may have something to say about a test block finding its way into competition hands free of charge.

pHaestus
11-11-2004, 11:42 PM
Could you PM or e-mail me the contact address? I need to get a little "design/goal/marketing" blurb for this block too for the review.

Cathar
11-11-2004, 11:54 PM
Could you PM or e-mail me the contact address? I need to get a little "design/goal/marketing" blurb for this block too for the review.

Done. Taner is his name. Nice guy once we got past the initial unpleasantness resulting from that old thread.

jaydee
11-12-2004, 09:55 PM
http://phaestus.procooling.com/temp/alphacool-prelim.jpg

Preliminary data. Block is very good but also very restrictive. I will finish this review either tonight late or tomorrow.
Very good.

jaydee
11-13-2004, 03:19 PM
Hi guys!
Figured out something potentially very useful yesterday. My Epox 8K3A+ isn't water damaged; it has leaky caps! The 4 nearest the power regulation area are bulging and have a little dried brown juice on the top. I noticed this because the replacement 8K3A+ someone sent me has THE SAME PROBLEM. I haven't yet booted the new board, but I am pretty sure I can salvage the old 8K3A with some new caps. KnightElite has a socketA mobo graveyard at his house I think; gonna try to get 8 decent caps from his Abit boards.

I built a comp for my daughter a while back with one of my Epox 8K7A's. About a month ago I was getting complaints the thing would lock up and while doing so they could hear a high pitch squeal. I got the comp today and took the side cover off to find whats in the attached pic. I am now thinking the last 2 8K7A's I had that mysteriously died were from the same problem.....

This board will work fine until Paint Shop Pro is loaded and the CPU starts to get loaded up then the squeal and then the freeze up... Got to be the caps? I am going to replace them from one of the Abit or Asus boards I got laying around.


This bothers me because they were supposed to have this bad cap problem sorted out long before the 8K7A's were released. Obviously not and the 8K3A+ you got there is a good year newer... :mad: Bah...

pHaestus
11-14-2004, 02:11 PM
Lots of new data posted yesterday and today throughout forums. Expect 2-3 more reviews this week :)