![]() | ||
|
![]() |
#1 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
I will first test a few waterblocks using my AMD64 system and IHS.
Then the testing will be repeated on AMD64 sans IHS. Blocks will be chosen to be as diverse as possible in design: Swiftech Storm vs Apogee leaps to mind from the forum debates. Any comments/questions/suggestions before it starts? Expect the test results to be a modest upgrade from my old testing at this time.
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
#2 |
Pro/Staff
Join Date: Oct 2001
Location: Klamath Falls, OR
Posts: 1,439
|
![]()
Could we try some tests before and after IHS removal? Also, if you have a really low end block to compare with the known good blocks, that would be nice.
|
![]() |
![]() |
![]() |
#3 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
![]()
Yep, an old block, any old block.
This will not address the contention that the IHS's TIM joint is variable (but is not the objective here). pH, you understand that by removing the IHS, you'll never be able to get it back the exact same way? One way shot. |
![]() |
![]() |
![]() |
#4 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
Yes Brian this is the point
Testing on a few wbs with IHS present followed by testing the same blocks with a bare die And yes Ben I know what I'm doing. Consider that I am basically making the choice to ditch the IHS as a variable in my testing. The conclusion that we're coming to in staff discussions is basically that Procooling AMD users are gonna pop the tops if they're bothering with H2O for performance purposes. If that's a bad assumption to make then I'll get another CPU. I look at the findings obtainable from this kind of test to be worth the $100 outlay of cash to be honest. If nothing else it'll provide some real data to go along with all the arguing on the forums right now ![]()
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
#5 |
The Pro/Life Support System
Join Date: Dec 1999
Location: Denver, CO
Posts: 4,041
|
![]()
See IMHO I think the tim join variability should be assumed. Even if it isnt completely variable for every CPU, its still an unknown. maybe some machine had a bad batch of the TIM joint compound, maybe a big air bubble right over the core when the tim compound was applied, maybe the IHS is warped, etc... Since we cant predict the QC going on for TIM joints I would go with the thought that you cant trust it.
For that fact I think its smart to do one test with the IHS and the additional TIM, and then remove it to do all tests in the future without the IHS assuming that if you want to see the temps in the reviews, you need to remove your IHS. Its not that big of an assumption to make that people who are serious would remove that.
__________________
Joe - I only take this hat off for one thing... ProCooling archive curator and dusty skeleton. |
![]() |
![]() |
![]() |
#6 | |
Cooling Savant
Join Date: Sep 2002
Location: ohio
Posts: 140
|
![]()
Agreed - good way to proceed. Definately test an old block or one you have tested before.
Quote:
BUT, could be first step in such. If pH does this with all subsequent (say AMD) cpu's he uses on this bench (IF there are subsequent cpus) then something could be said about the variability or lack thereof of the IHS joint, no? need large sample size basically, not suggesting to pH to do this, but his data might be worth it down the road, that's all. |
|
![]() |
![]() |
![]() |
#7 |
c00ling p00n
Join Date: Jun 2002
Location: L.A.
Posts: 758
|
![]()
So Derek does this mean you will have 2 CPU's? One sans IHS and one with? I think that would be best. IF you need a donation to buy an additional CPU(s), say the word.
![]()
__________________
*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* E6700 @ 3.65Ghz / P5W DH Deluxe / 2GB 667 TeamGroup / 1900XTX PC Power & Cooling Turbo 510 Deluxe Mountain Mods U2-UFO Cube Storm G5 --> MP-01 --> PA 120.3 --> 2x DDC Ultras in Series --> Custom Clear Res "Artificial intelligence is no match for natural stupidity." 1,223,460+ Ghz Folding@Home aNonForums *:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* |
![]() |
![]() |
![]() |
#8 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
![]()
Good point Al, but how is pH going to measure the top side temp, and will the next tester be able to replicate it?
Last edited by bigben2k; 12-07-2005 at 06:00 PM. |
![]() |
![]() |
![]() |
#9 | |
c00ling p00n
Join Date: Jun 2002
Location: L.A.
Posts: 758
|
![]() Quote:
__________________
*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* E6700 @ 3.65Ghz / P5W DH Deluxe / 2GB 667 TeamGroup / 1900XTX PC Power & Cooling Turbo 510 Deluxe Mountain Mods U2-UFO Cube Storm G5 --> MP-01 --> PA 120.3 --> 2x DDC Ultras in Series --> Custom Clear Res "Artificial intelligence is no match for natural stupidity." 1,223,460+ Ghz Folding@Home aNonForums *:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* |
|
![]() |
![]() |
![]() |
#10 | |
Cooling Savant
Join Date: Sep 2002
Location: ohio
Posts: 140
|
![]() Quote:
May not be relevant, as IF the IHS joint is variable, then some cpus would see a large temp drop (core temp) and others not so. In theory this would be enough to say that IHS joint is variable? but anyway, best (IF objective is to prove IHS variability) is probably to groove the IHS on each cpu before testing, see Link not too familiar with Derek's new setup, how is he monitoring temps? --Jay |
|
![]() |
![]() |
![]() |
#11 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
![]()
As Al suggested, there's an opportunity to measure IHS tim variation, which means that one would have to measure the temp on both sides of the IHS.
Added: I'm not sure that Derek is willing to groove the IHS. As long as one can replicate the top side measurement, it's all good. |
![]() |
![]() |
![]() |
#12 | |
c00ling p00n
Join Date: Jun 2002
Location: L.A.
Posts: 758
|
![]() Quote:
__________________
*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* E6700 @ 3.65Ghz / P5W DH Deluxe / 2GB 667 TeamGroup / 1900XTX PC Power & Cooling Turbo 510 Deluxe Mountain Mods U2-UFO Cube Storm G5 --> MP-01 --> PA 120.3 --> 2x DDC Ultras in Series --> Custom Clear Res "Artificial intelligence is no match for natural stupidity." 1,223,460+ Ghz Folding@Home aNonForums *:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:* |
|
![]() |
![]() |
![]() |
#13 | |
Cooling Savant
Join Date: Sep 2002
Location: ohio
Posts: 140
|
![]() Quote:
Ok, so from Derek's test bench status thread, he's got wires on the cpu diode pins and a custom reader, hopefully no interaction from motherboard? So anyway, he would have to do this to every cpu he tests, and in theory would get diff. temps each time. However the waterblock would be the same, so then he pops the IHS off the cpus and tests again and all temp readings should be the same now (assuming same conditions, same under-load wattage dump, etc...) within error margin. Since the baseline are all the same, but with IHS on the temps are NOT all the same, conclusion is there is variability in IHS joint. *again, assumes cpus and load conditions are identical - many will argue against but is there data to say otherwise? --Jay |
|
![]() |
![]() |
![]() |
#14 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
![]()
Right, his test setup is completely isolated from the mobo (hence why he gets so much admiration from those who can appreciate the difficulty).
Once the IHS is off, there's going to be a CPU temp drop, for each block tested. C/W estimates for each block should remain the same (if calculated that way); any variation would come from the resized contact area, which is going to be smaller without the IHS; a small difference, almost insignificant. IHS tim joint variation assumes that the internal tim joint varies, between mounts (among other things), so it really needs to be tested on several CPUs. I suspect that the internal tim joint may undergo some curing (one of many possible explanations, if variations indeed do exist), which would put Derek's CPU way past that stage (since it's already been used several times). I'm lead to believe this from my previous inquiries with a few manufacturers, as well as user experience, which all indicate that TIM joints (with few exceptions) need to be cured (i.e. put through a hot temperature, to settle). I can dream up many other ways for the internal TIM joint to vary. |
![]() |
![]() |
![]() |
#15 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
No electrical interference from motherboard as far as I can tell; diode reader reports same temperature whether mobo is powered or not.
I calibrate the diode in a water bath over 0-45C always vs. the same temperature probe. So while the offset of the diode is a little different from CPU to CPU the dT CPU-water is always the same provided I get the same voltage delivered on the same mobo (assummmmppppppttttions!) I was honestly planning to do all my testing after the first few "with IHS" blocks using a bare die. That's open for debate I s'pose
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
#16 | |||
Cooling Savant
Join Date: Sep 2002
Location: ohio
Posts: 140
|
![]() Quote:
"the dT CPU-water is always the same provided I get the same voltage delivered on the same mobo (assummmmppppppttttions!)" Is always the same for multiple mountings on one cpu? Or always the same across cpus? yes many assumptions on this part. Quote:
Quote:
--Jay |
|||
![]() |
![]() |
![]() |
#17 |
Cooling Savant
Join Date: Aug 2004
Location: uk
Posts: 109
|
![]()
blocks to be tested .... can you test the "MP-05 XTREME" at this time? can u get 1? and a ww as a baseline? maze 3 as a baseline?
__________________
"<pH> I'll stab you in the genitals with a rusty shank if you touch my computer stuff" "we are only 'mean' to the persistently ignorant, lazy, and anyone who questions us" BillA |
![]() |
![]() |
![]() |
#18 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
Allbigger:
Are you asking if CPU diodes are PRECALIBRATED from the factory? Well they have a tolerance and different CPUs generally are within ~1C of another. But the solder joint introduces resistance in the diode reader lines and this varies every time you solder a new CPU up. The calibration vs. a known thermometer in a water bath is designed to deal with all the possible variables in the diode reader/wires/CPU diode system so that it's in cal. The dT CPU-water is the same for the same wb at the same flow rate and at the same vcore/FSB/multi on multiple chips. Since I did the calibration of the whole soldered up system described before. And YES if CPU temps go down then C/W would decrease. How much though? Is this a constant change (an offset) for all wbs, or do some cool the IHS or die area in a way that makes them particularly affected by the change? THIS is a good question I think. flatline: I have an MP-05 Xtreme downstairs and a WW. I have a question though. If I am choosing only 3 or 4 wbs for the IHS vs. bare die test, then what would be the justification for the ones you have named?
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
#19 |
The Pro/Life Support System
Join Date: Dec 1999
Location: Denver, CO
Posts: 4,041
|
![]()
well I think you need to get your mits on an Apogee, I mean with all the BS on the forums here, I would love to see what the reality is for one of those blocks.
__________________
Joe - I only take this hat off for one thing... ProCooling archive curator and dusty skeleton. |
![]() |
![]() |
![]() |
#20 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
hmm bigfootcomputers doesn't have stock yet :/
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
#21 | |
CoolingWorks Tech Guy Formerly "Unregistered"
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
|
![]() Quote:
(since obviously the 'as bought' camp is distinct from the pop-a-top group) "I would love to see what the reality is for one of those blocks" on a CPU no less ? for a wb that 'lost' the die sim tests ? pH you really do want to make an initil internal TIM joint characterization that will be directly comparable with subsequent CPU testing results (IS there the initial variability, to what extent ?) - if you test with an IHS over time you will be able to monitor the internal TIM joint in the same way note that a dT will not yield the same info (unless only one variable is assumed) |
|
![]() |
![]() |
![]() |
#22 |
Cooling Savant
Join Date: Aug 2004
Location: uk
Posts: 109
|
![]()
as i see it the MP-05 Xtreme will be the only block with jets and a proper mounting system now the storm is discontinued im thinking what ppl will actualy want to buy and are able to. ww as i bouth have 1 and there performance is seen on most wb testings graphs
__________________
"<pH> I'll stab you in the genitals with a rusty shank if you touch my computer stuff" "we are only 'mean' to the persistently ignorant, lazy, and anyone who questions us" BillA |
![]() |
![]() |
![]() |
#23 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
Bill:
I have to consider the hassles of getting a grooved CPU and 40 gauge TC affixed to the top of the IHS and then desoldering pins on the underside and adding wires. Then I have to consider that every time I change hs/wb I have a substantial chance of breaking either the CPU, the diode wires, or the TC wires. It doesn't seem worth it to me. Then if the TIM joint itself is degrading over time then that will be another problem that I have to deal with. I suggest that if you want to test wb performance on IHS-bearing CPUs with TCs then you should use Intel. If you want to take CPU die temps on any modern CPUs I would suggest you use a bare AMD. I could be way off-base though? //edit: I am aware that differences in pre- and post-op dT includes at least two variables: heat spreading and TIM resistance. But in both cases we're at least getting the useful number right? CPU core temperature? //edit2: Do we have someone here willing to do the grooving of the IHS? I could always buy a second CPU, send to them for grooving, get a 40 gauge tc, and then neuter the processor I guess. Any takers?
__________________
Getting paid like a biker with the best crank... -MF DOOM Last edited by pHaestus; 12-08-2005 at 01:14 PM. |
![]() |
![]() |
![]() |
#24 |
CoolingWorks Tech Guy Formerly "Unregistered"
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
|
![]()
yes, IF one were to do CPU testing with an IHS I believe an Intel 775 package would be best for durability
note that I am not recommending anything by Intel to this crowd, lol no good reason to waste $ to groove an AMD IHS if the joint is to be considered variable before the fact yes; some kind of core temp can be had from AMD CPUs, for this purpose that makes them better no, the analysis is complex if seeking correlation; #s per se are easy many assumptions are being made but if the goal is limited to describing performance as tested all is good, if the presumption is to extrapolate all bare CPU wb performance parameters from a single test, . . . . . "neuter the processor" ?? |
![]() |
![]() |
![]() |
#25 |
Big PlayerMaking Big Money
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
|
![]()
Well I would think at least having 4-5 good dT vs. Q curves for + and - IHS using different blocks would be a start. We can for sure look at whether the IHS is just an offset or if the relationship is more complicated. Right?
Another point in favor of the tests is that if I choose wbs I have previously tested on SocketA then we can also look at die size effects (well maybe but this is a whole 'nother can of worms).
__________________
Getting paid like a biker with the best crank... -MF DOOM |
![]() |
![]() |
![]() |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
|
|