Swiftech Apoggee review by Robotech systemcooling

[Orkan]

Originally Posted by Joe

Which Storm block did you test against?

swiftech

... and the quality of the storm is 10x that of their new golden child, Apogee. Didn't find any metal shavings in the storm. Just some delrin shavings in one of the screw holes that clamp the block together. ... outside the water loop.

[RoboTech]

Originally Posted by ricecrispi

Thanks fo the input robotech. ...would you dare pop the lid on IHS and see what happens?

I "could" but: (1) I don't have a spare CPU I'm willing to sacrifice if I screw up and (2) its not really necessary as you can buy them with the IHS already removed (actually never put on) => mobile CPU's.

As BillA pointed out in another thread, the ONLY real purpose for having the IHS is mechanical protection. The copper base of a HSF or waterblock makes a very good heat spreader without introducing another set of thermal boundaries into the mix the way an IHS does.

Agreed, one of the best TRUE tests of waterblock performance is how much of a stable overclock you can get. Who cares what the IHS surface temp is?...

[nikhsub1]

Originally Posted by RoboTech

As BillA pointed out in another thread, the ONLY real purpose for having the IHS is mechanical protection. The copper base of a HSF or waterblock makes a very good heat spreader without introducing another set of thermal boundaries into the mix the way an IHS does.

Actually, if you mean this post http://www.systemcooling.com/forums/...ndpost&p=33175 that was me... forgot to type me name! Guess I should just register lol.

[Etacovda]

joe - by 'the cpu tests' i mean the cpu tests that Lee did - tests like pH's are very useful.

As i said earlier, overclocking tests are extremely 'hit and miss', even with the same cpu - but hell, its more useful than the cpu numbers which dont tell us ANYTHING now... was more meaning it as something else to test and discuss, than a be-all and end-all performance conclusion - id say it'll be a while before some other testing method will be agreed upon, so its certainly something to consider.

[Cathar]

Originally Posted by RoboTech

Agreed, one of the best TRUE tests of waterblock performance is how much of a stable overclock you can get. Who cares what the IHS surface temp is?...

http://www.xtremesystems.org/forums/...ad.php?t=82566

Mind you, even though that test favors the Storm, I still don't strictly agree with using IHS's, even for overclocking ability tests. Can change to a different CPU from the same batch, and the positions can reverse, all due to CPU-die->IHS->wb variations.

The use of IHS's presents a serious issue in terms of the goals of the average overclocker. The goals of CPU manufacturers for using an IHS are incompatible with the goals of overclockers, or even "small overclock with silence" seekers.

In terms of assessing waterblock performance with an IHS involved, to date I have yet to find a single research paper that willingly introduces multiple thermal interfaces when attempting to determine h(eff) and h(conv) of a cooling apparatus. The use of IHS's are also incompatible with cooling apparatus performance assessment.

The only thing IHS's are good for are for mechanical protection of a potentially fragile silicon die (they're not that fragile mind you), and possibly to assist in spreading heat in scenarios of badly non-flat heatsinks, or improperly designed heatsinks that use insufficient base-plate thickness to spread the heat to the primary convectional area. Arguably somewhat helpful for the older style and cheap all aluminium heatsinks due to slightly improved heat spreading ability. For copper base-plated heatsinks, they offer zero benefit (except in limited scenarios mentioned above).

IHS's incorrectly shift the focus away from correct heatsink manufacture and quality control, and instead replace correct design with a "numb and dumb" attitude.

[bigben2k]

Thank you Lee (Robotech).

Just a reminder to those (apparently most) who forgot:
the IHS temp was taken on the side of the IHS.

What this data tells me is that there are no reliable temperature measurements on this particular CPU test, and leaves more questions:

-Is it because the CPU temp was measured by the mobo? Would it be more accurate if a Derek (pHaestus) style temp measurement was taken?

-Wouldn't it be simpler for Lee (Robotech) to implement the Intel style temp probe on this CPU? (i.e. the IHS groove with TC). Would it churn out more meaningful data?

I don't see the need to pop off the IHS just yet, I'd recomend taking it one step at a time; if you pop off the IHS, it's a one way trip. What you need most right now, is a more reliable and accurate CPU temp measurement.

[ricecrispi]

Originally Posted by RoboTech

I "could" but: (1) I don't have a spare CPU I'm willing to sacrifice if I screw up and (2) its not really necessary as you can buy them with the IHS already removed (actually never put on) => mobile CPU's.

I do have a 939 sitting around. I'm on break till January so I'm doing some prelim tests. It can be a donor afterwards

[BillA]

Originally Posted by bigben2k

Thank you Lee (Robotech).

Just a reminder to those (apparently most) who forgot:
the IHS temp was taken on the side of the IHS.

What this data tells me is that there are no reliable temperature measurements on this particular CPU test, and leaves more questions:

-Is it because the CPU temp was measured by the mobo? Would it be more accurate if a Derek (pHaestus) style temp measurement was taken?

-Wouldn't it be simpler for Lee (Robotech) to implement the Intel style temp probe on this CPU? (i.e. the IHS groove with TC). Would it churn out more meaningful data?

I don't see the need to pop off the IHS just yet, I'd recomend taking it one step at a time; if you pop off the IHS, it's a one way trip. What you need most right now, is a more reliable and accurate CPU temp measurement.

yup Ben, if info is desired
- answered Lee in his forum

[Joe Camel]

thought this thread could use some eye candy

here are a few pictures to look @ and maybe stimulate some...thing.



a 3500 Venice



same 3500 Venice



3200 Winchester



FX-55 fresh out of my MachII after about 8 months of benching / gaming



same 3500 Venice and approximate thickness of IHS (.094")
(micrometer slid over the rim and around the area with TIM)



same 3500 Venice (CPU and IHS facing same way)



a FX-57 that some idiot killed while HACKING the IHS off



IMO:

phase-change VS. water:

hard-core benching (which is about the only thing cascades and LN/Dice tubes are good for) and OCing are quite different. most gamers and *STABLE* home-office PC's that are OCed arn't phase cooled. (A perfect example is nikhsub1) they are two different crowds and buy/use quite different hardware.
(im looking for a new TEC WB to mod for my GPU, so this block is of interest to me)

IHS:

as much as i HATE the introduction of the AMD IHS, i dont think the water cooling community can ignore its existence. (as 95%+ of users will never remove the IHS and/or the probable non-removable IHS of future CPU's)


"real-world" IHS VS. die temps:

could one modify an IHS by milling a slot through the sidewall, and if needed, part way into the face of the IHS so that a thermal probe could be inserted "under" the IHS and come in contact with the (side of) core?

as much we all HATE "MBM5 temps" and thermal probes on the side of a die / IHS, id say its about the best you can do to *TRY* to produce "real-world" data...

without an industry standard test-bed (as imperfect as it may be) were all chasing our tails


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

here are a few shots to remind us of what its all about...

same 3500 Venice (NAKED ) @ 3GHz

IDLE:


LOAD:



Storm G5, D4 pump @ 16V (back removed and fan cooling MOSFETS), BIXII, 4x100+ CFM fans



i keep my basement at a nice cool 65F during the summer.
(NOT a running PC but the exact set-up that produced the first screen-shots)




GOOD LUCK!!


PS

Cathar,
keep the faith...dont let the buggers wear you down:

[Marci]

Only thing of use now would be a shot of a new unused CPU, de-IHS'd , and then comparison photos vs older de-IHS'd CPUs - see if tim1 ccompression is same or whether compression effect is increased after use... (u get my line of thinking?)

[BillA]

I suspect that AMD and Intel have not been 'the same' for some time
one always needs to bear in mind that IHS 'flatness' is as mounted, not that which is seen when the package is unconstrained
Tim1 compression is as mounted, not loose; the IHS crown, edges, CPU package, socket, mobo, etc. all take part of the applied clamping force
nor are all the same (see Nicona), and BTX to come which 'float'

the mechanical loads on a 'large' IHS during the removal of a sink can be considerable, removing pinned CPUs with the sink is almost the norm
these loads, and severe thermal cycling, will de-couple a grease TIM over time
I'm quite unsure of any value derived from testing a Tim1 joint generally agreed to degrade over time

if the goal (not of this thread, eh ?) is to develop a test platform, concentrate on possible candidates

this is 'our' problem, we cannot focus and make choices - just endless debate

[nikhsub1]

FWIW, Joe Camel's Venice looks EXACTLY like the old Northwood IHS > TIM bond... to a tee. That method is horrible, don't know why AMD continues, good for those who wish to remove it though. If you cut the outer epoxy, the IHS falls right off, tis the only thing holding it on at all. Imagine that, the outer rim of the IHS epoxied to the PCB of the CPU is the only damn thing holding it on. With this method and all my Northwoods, I saw WILD reported temps and some VERY high differences between with and without IHS in terms of reported temps, some as high as 10C!

[Orkan]

What is the 'best' way to remove these amd IHS nik? I think I'm going to uncap my X2.

[nikhsub1]

Originally Posted by Orkan

What is the 'best' way to remove these amd IHS nik? I think I'm going to uncap my X2.

Carefully With the Intels, there were no capacitors on the PCB like there are on the AMD's, so you have to be carefull not to touch them with the razor blade. What I did was; Get a new straight edge razor blade. I held it with a small pair of pliers. Gently slide the blade under each corner, once that is done, work the blade around the IHS making sure not to push it in too far so you don't touch the capacitors. Once you cut round the whole IHS, it should fall right off.

[stev]

Originally Posted by RoboTech

I "could" but: (1) I don't have a spare CPU I'm willing to sacrifice if I screw up and (2) its not really necessary as you can buy them with the IHS already removed (actually never put on) => mobile CPU's.

As BillA pointed out in another thread, the ONLY real purpose for having the IHS is mechanical protection. The copper base of a HSF or waterblock makes a very good heat spreader without introducing another set of thermal boundaries into the mix the way an IHS does.

Agreed, one of the best TRUE tests of waterblock performance is how much of a stable overclock you can get. Who cares what the IHS surface temp is?...

Lee,

(3) It will void the AMD warranty on the little, but expensive CPU.

Like I had mentioned at SystemCooling, we should start a RoboTech AMD64 CPU fund raiser to take off the IHS.

The Inquire mentioned that AMD was dropping a chunk of the AMD64 939 cores. So, even the processors in inventory will remain high until they are all gone.

As for the mobil processor without an IHS skin, does that run cooler anyways? Is this the Turion processor that is only 35W?

Lee, you have provided great help this week. Thank you!

Stev

[stev]

Originally Posted by Joe Camel

...
"real-world" IHS VS. die temps:

could one modify an IHS by milling a slot through the sidewall, and if needed, part way into the face of the IHS so that a thermal probe could be inserted "under" the IHS and come in contact with the (side of) core?

JC,

My French is weak, however, you're looking for something like this ...



http://membres.lycos.fr/roscool
(From Roscol's Cooling-Masters site.)

Too bad we can't laser etch the IHS with channels and make it the base of our cooling block. But, then again, we're not IBM who holds those patents.


Stev

[BillA]

if Intel thermal design guide 'case temps' are of interest, and they have to be,
then an Intel surface groove should be used
given the non-flat IHS surface when unconstrained, a surface groove is seen as easier also

what the hole shown will give is a mean IHS temp
the special surface groove is to place the TC junction as close to the IHS top surface as possible
(it is the IHS/case temp at its face, the Tim2/bp junction is being approximated)
the groove and TC placement are VERY sensitive, not a die grinder and Kapton tape job

[RoboTech]

Update, and some personal thoughts…

I have been reviewing PC hardware for several years (mostly cooling devices, power supplies and cases). I enjoy seeing what makes things tick and write reviews as a hobby. I try to be as objective as possible and find it’s a lot more fun to write about a products strengths and assets than it is to point out weaknesses and liabilities; but in general I call ‘em like I see ‘em.

Reviewing Swiftech’s new Apogee waterblock has not been a “typical” review. Much controversy started the day the Apogee was announced, before anyone had seen/used one, let alone reviewed one. Because of this I felt prompted to do more testing than usual, testing that I wouldn’t normally do (and many regard as useless), but that I thought would provide additional perspective. (The larger 32mm die I mainly use for TEC testing and the live CPU tests were to try and replicate testing done by Swiftech.)

My day job involves product design and development so it’s natural for me to be always critiquing products in my mind and assessing strengths and weaknesses and asking, “why did you do it this way?” Frequently the designer or manufacturer will have a satisfactory answer and we go on. Unfortunately, tech site hardware reviews normally don’t work that way. What really brings this into focus for me is when an opinion I express in a hardware review negatively impacts a company and its employee’s livelihood. I find this humbling and it makes me question if this is what I really want to do in my leisure time.

For example: During my review of the Apogee, I noted some apparent thin spots in the top housing. I even went so far as to section a top to illustrate the point. My initial impression was, “I wouldn’t have designed it that way”, so I noted it as a potential concern. Making the top slightly thicker or radiusing the outside edge seemed like a better way. Swiftech came back and stated the top thickness met their design specifications and as part of their QC, they tested each assembled waterblock to 40 PSI to insure there are no leaks. (And I assume the outside edge was squared-off so the metal mounting plate would locate securely in position.)

In hind site, the top now appears to be fine. There is nothing inherently wrong with a molded versus machined part. I’m satisfied with the answers Swiftech gave, but again that is just my opinion. I have only read about one Apogee with a cracked top so unless a lot more start showing up, this appears to be a non-issue. And I never had any worries about the minor stresses from tubing side-loads breaking or cracking the plastic top.

Here are a couple more pics for additional clarification.



The red line is where I made the first cut, which was included in the review. The blue line is another section that shows one of the four “thin-spots” from a different angle.



And for those who might be wondering if Swiftech put me up to this – no they did not. I am posting this because I feel a sense of responsibility to the manufacturer as well as to our readers. It appears like my questioning the top thickness is being interpreted by many as a confirmed flaw that will result in leaks or breakage. That does not seem to be the case but the burden of responsibility falls on me for bringing it up in the first place and creating that perception. It appears my concern with the top thickness was unnecessary (I was wrong) and in my mind at least, this issue has been adequately addressed.

Thanks for listening/reading…

[nikhsub1]

Lee, I don't think many here thought the top was really much of an issue (save orkan). I agree that the top of the apogee seems to be a non issue with only one case reported, again orkan who didnt send the block back to swiftech IIRC. Anyway, why not post what you just wrote here as an addendum to the article? Anyway, I for one applaud your hard work with all this and admire your dedication and skill.

[Orkan]

Good testing is good testing.

The block is in transit back to swiftech. I find it hard to believe that I cracked the block putting the hoses on, so I am sure the block was cracked before it got to me somehow. Maybe a forklift ran it over or something... since according to gabe's "hammer" testing... that is probably the only thing that could crack it. Perhaps mine fell onto some train tracks and was ran over by a locomotive.

Be that as it may... when comparing the machined top of my maze4 and storm to the injected top of the apogee... ...
well, there is no comparison.

The mold could have been made a touch thicker in my opinion. ... in several places. But the apogee exists for one reason: To make money.

Not that I fault any company for this.

[Anonymous]

Originally Posted by Orkan

But the apogee exists for one reason: To make money.

Not quite, from what I hear. Also used to cool CPUs or throw at cars from an overpass.

[bigben2k]

Glad to see you clear that up too, Lee. I agree; it's not really an issue. A concern maybe, but I don't believe that the manufacturing would vary to a point where Swiftech would not process any more returns that usual.

Joe Camel -=> That's some good info (IHS thickness). I calculate 0.094 to 2.4 mm, which is consistent with the rim of the IHS, but would like to know the thickness right over the core area, to see how it compares to Intel's IHS (should be 1.3 to 1.4 mm). Can you check?

[Joe Camel]

that was the area inside the rim but not where the TIM was.


yellow = (about) where i passed the micrometer to get my average:





under the assumption that there isnt a "divot" directly over the die, id say its safe to assume that 2.4mm of IHS is between the die and a HSF (give or take a few TIM joints )

i cleaned the basement a few weeks ago and cant find that darn IHS ANYWHERE!!!
(ill keep looking though, i KNOW its down here)


EDIT:

soooo thats where i put it....

TIM removed, "dead center":
(lapped 3500 IHS as seen above)
.094" (2.8376mm) EIDT: 2.3876mm
(?.??" x 25.4 = ?.??mm or ?.??mm / 25.4 = ?.??")





black "glue" removed (mostly), on the rim:
(lapped 3500 IHS as seen above)
.119" (3.0226mm)

[bigben2k]

I can't imagine that there is a divot, but I'm puzzled at the difference:
Intel: 1.4mm
AMD: 2.4 mm

?!?

[Joe Camel]

edited above post.

sorry for the delay in response.