6002-A worked well for me w/ low head pump - 2860 MHz stable

[DDogg]

I wanted to post and thank all of you experienced folks for your posts that have helped me so much.

Also, since most of you seem to use expensive equipment and pumps, I thought it might be useful to post some results using the 6002-A with a low volume, low head 120 gph LG $19.00 submersible fountain pump from Lowe's.

Some more information here. I hope it is ok to post a link. I just didn't want to type all that again. http://forums.extremeoverclocking.co...&postcount=275

Now I do realize that 54-55c is too hot, but I thought it was meaningful that an affordable mass produced block and a pitifully little pump like this could allow stability at this wattage when using a VCore bios setting 2.125 volts. SiSoft shows 205 watts. I'm told that number may not be accurate, but it might at least serve as a comparison number.

When the bios voltage is reduced to 2.05, which still shows [184] watts in SiSoft, I can comfortable run 2821 MHz in 25-26c ambient and hold =< 50c under load. For under a hundred buck investment plus some sweat equity, I'm a real happy guy.

The block did seem a little finicky to me about seating. It just did not seem to want to tolerate any sidewards pressure from the hoses, even when it seemed only slight. I guess that makes perfect sense, but it was aggravating to get a solid seat. In fairness, I was using that darn stiff 3/4 OD re-enforced tubing with some 90's to bring it down and out the back of my case to my external radiator box. When I changed to the light flexible 5/8 OD cheap stuff I had no more problems.

I did break the rules a little and used two matched spacers, ~ 1/16 - 3/32 in., under the springs to allow a little more pressure on the CPU. It seemed to make a fair difference, but I have no measured quantifiable data to back that up. I have learned from reading here how you guys are about that subject Also, hopefully I will not get chewed out about the shims from a lurking old curmudgeon Swiftech guy.

Thanks again to all for the informative posts and this excellent resource, and thanks also to Swiftech for making a high performance block like the 6002-A affordable.


Current daily -

ABit NF7-S Rev 2 | Mobile XP 2600 @2821 MHz (13*217) VCore @2.06 volt (bios) | Chipset Vdd @ 1.70 volt |(2) 256 meg PC3500 HYP_X (11-2-2-2 VDim 2.9 v Dual Channel @ 434 MHz) | DIY Water cooled - Swiftech 6002-A block, Fedco 2-342 rad < O'Reilley Auto part # 399090 < $20.00

[Myth]

I have a setup just like u, just using a Swifttec mcp and get 2850 at around the same temps as u.. just rocks!

[BillA]

sounds awfully good under the circumstances
truly hoses are a greater 'problem' than most realize

we are limited in our recommendations by the CPU mfgrs design guides,
more pressure is the conventional manner of addressing high loads from the tubing

[Mars]

Good job. Starting and finishing a project with success always puts a guy in a good mood. You'll most likely tamper with it again....and again....and again. It's inevitable.

[DDogg]

One of the things that really astonished me was how a bad seat could be masked by using too much AS5.

At the time I removed the block I could run solidly at 2750-2775. One would not think of an improper seat with those numbers. At least I certainly would not. I had noticed that an increase in VCore to 2.1 would nearly always cause a bsod 'machine check error' and assumed this was very natural at that high a VCore.

When I removed the block, I noticed only a half clear imprint of the CPU die on the block. The other half of the imprint was AS gray. Remember, I could run at 2775/VCore 2.05 perfectly overnight Prime95 stable like this!

I had not followed the instructions and had applied AS5 to both surfaces instead of a semi-transparent layer to the CPU only. Frankly I thought some of you were acting like old women with their favorite recipe I thought, heck, it will either work or not, "what's the big deal?"

After replacing the hose to remove all pressure to the WB, scrubbing both surfaces with a coffee filter and alcohol, and using an ultra thin coat (I'm talking near transparent) to the CPU only I immediately noticed a subtle difference in behavior, plus the not so subtle difference in 2860 stable. When pushing VCore aggressively, Prime would drop out politely, instead of bsod which jibes with some of the pinprick hot-spot theories I remember reading aout.

Still, I'm amazed that I could run stable at 2775 with bios VCore 2.05 with a bad seat and having only half the CPU top firmly touching the block. The ability of AS5 to mask a bad seat speaks favorably about it, yet it's a sneaky thing to figure out. Now I'm going to always use the thinnest possible layer on the cpu only on the theory that a bad seat will show up near instanly, instead of being masked.

I think the attachment methodology of the center fed 6002-A coupled with hose side loading, and in fact many of the spring loaded mounts used on WB may lend themselves to subtle and sneaky seating problems, but I'm sure no expert. The only thing I did like about my old Koolance was the way it put direct pressure on the top of the WB directly alligned over the CPU die. Of course it was side fed with cross flow so there was nothing to get in the way of that style mount.

@BillA - I wondered about doing a mod to the mount so that four springs were used on the middle of the blue crossbrace equally spaced like an inch from the center of the pressure point so that the pressure is controlled within specs yet is applied more centrally over the die face. Any thoughts on that? Probably not too practical I would guess.

[DDogg]

I wanted to resurrect this old post to add some new information. It seems I still did not get the ultimate seat as I thought I did above. I recently reseated again, making sure my spacers did not bind on the screws, and my temps dropped again. I have also added another of the pumps in series. Now with an inlet temp of 26C my onboard sensor shows 52C which for comparison purposes is about 4C less than before. This allows a stable 24 hour 2860 Mhz @ bios VCore of 2.150 which is an actual of around 2.08 v as measure by 8rdavcore.

I guess the morale of the story is the 6002-A is awesome, but a user should pay particular attention to the quality of the seat (on any WB) and perhaps use the same method as the experienced testers by trying 5 or 6 test seats to get an idea of the best one.

Between my first improper seat described at the top of this thread and the present one is something like 10C, IIRC, although all seats yielded a stable 24 hour prime test. In short it is hard to tell if you've got it right without multiple tests to compare.

[Ruiner]

Cooling aside, that is one awesome chip.

[DDogg]

Yeah, I'm very pleased with it. I got a little stupid yesterday, while my inlet temp was around 23, and ran it up to 2899 @ 2.15 actual VCore. Not too stupid, just long enough to run Super-Pi 1M a couple of times. No problems, but things were warming up pretty fast. Only a 35 second 1M pi though. I thought it would be a little better at that speed.

Certainly should tide me over for 6 months to a year waiting on the AMD64s and mobos to mature and drop in price.

[Cranky]

tbh my 6002 is my 2nd swifty block and i couldnt be happier with it. i saw a 2-3 c temp drop on my p4 2.53 vs. my 5002..its gonna be nice to get that block on my new a64 rig soon.

[trans am]

Maybe you could help me. I am trying to tec a amd64 and I have a mcw6002-64. Do you use a cold plate or just put the cold side of pelt directly on cpu heaspreader? Someone said 6002 doesn't work for pelt, but you disproved them. Please help.
http://forums.procooling.com/vbb/showthread.php?t=10950

[DDogg]

trans am, I don't know if you were addressing me, but I believe you might be confused a bit. I don't use a TEC with my 6002. Even so, and to answer your question as best I can, a cold plate would be mandatory.

The fact that you asked that question, and I mean no offense, shows a lot of confusion on the subject. I would advise you to do quite a bit more study on the subject. TECs are not a good solution for most people. Quite a bit of specialized knowledge is required to use them without destroying your CPU and motherboard.

Here is a forum where you might find answers better aligned with your particular needs. Good luck with your project.

The picture below is from the SystemCooling review of the MCW5002-64T.

[trans am]

Originally Posted by DDogg

trans am, I don't know if you were addressing me, but I believe you might be confused a bit. I don't use a TEC with my 6002. Even so, and to answer your question as best I can, a cold plate would be mandatory.

The fact that you asked that question, and I mean no offense, shows a lot of confusion on the subject. I would advise you to do quite a bit more study on the subject. TECs are not a good solution for most people. Quite a bit of specialized knowledge is required to use them without destroying your CPU and motherboard.

Here is a forum where you might find answers better aligned with your particular needs. Good luck with your project.

ok, sorry for being in the wrong section. I am already a member of eocf. I just heard this is the forum for TEC. I think maybe 5002-64 is my best bet then.

[BillA]

http://www.swiftnets.com/products/mcw5002-64T.asp