Cold Gpu Preparation

[Voidn]

The PCB will get cold / transfer the cold. And should keep other parts cold enough. I think Ed said not to use fans because it will melt the frost. One pro/con of eraser is that it transfers the cold. So if you really need to keep part of the card cold you could run earser towards it. though newer cards had problems with eraser making the mem too cold.

Also there are always memsinks. I debated weather or not they were necessary when the pcb is so cold.

 

[ChanceCoats123]

Yeah, I am not worried at all about the mem since most of the stock gpu coolers don't even make contact with the ram. I'm just thinking about the PWM section. I can always see how it goes.

 

[nzaneb]

A fan will melt the growing frost. Especially if it's a powerful card. Use paper towel to protect the areas you don't want to be hit with the fan air. Might as well wrap the pot up too. The frost on it will melt even if your fan is pulling air past it. On LN2 it would be a whole different ball game. Dice just isn't cold enough, so to speak.

 

[ChanceCoats123]

I figured so much. I'll probably have it wrapped up, and I'll keep and eye on the VRM's. if they start getting a bit too hot, I will section them off and get a fan going.

 

[Bobnova]

I would put heatsinks on the mosfets, mine were hitting 150f at 1.138v on the 480 on dice. The ram will be frozen solid, no issues there.

 

[ChanceCoats123]

I don't believe I have anything that would suffice.

However, I think a 5850 vs. a 480, the 5850 probably draws a bit less power. Since I don't have anything to put on them (and this card will have the air cooler put back on later), I will watch their temps, and if they're getting too hot, I'll throw a fan on them. Best I can do.

 

[ChanceCoats123]

Well this isn't happening tonight, I guess. The place I normally buy DIce is all out, and the other sellers within driving distance all close at 4 (45 minutes ago). This sucks. So now, I'm off to ship Hardass this 2600k.

Edit: PO was closed so I decided to call one more place... the night is back on track! DICE, check!

 

[Angry]

Sweetness, we want DICE'ed GPU p0rn.

.....Did you sell your 2600k?

 

[ChanceCoats123]

It was never mine.

 

[ChanceCoats123]

Anyway, pictures and results. It wasn't the most productive, and I'm not quite sure why, but I learned quite a bit.

http://hwbot.org/community/submission/2160079_chancecoats123_3dmark03_radeon_hd_5850_99940_marks

http://hwbot.org/community/submission/2160080_chancecoats123_3dmark05_radeon_hd_5850_41657_marks

http://hwbot.org/community/submission/2160081_chancecoats123_3dmark06_radeon_hd_5850_31913_marks

http://hwbot.org/community/submission/2160082_chancecoats123_aquamark_radeon_hd_5850_394081_marks

 

[nightelph]

Wow! Frosty!

 

[ChanceCoats123]

Yupp. Now granted, that was after the system was power-less. A little bit of DIce in the pot with no heat generated = frosty.

 

[Seebs]

Heh...

You really are squeezing that card for all will give you! Nice going there kid; I'm sure you'll figure out how to make the 5850 give you MOOOAR!

 

[hokiealumnus]

Excellent runs chance. I've been at it for nine hours and couldn't touch that many boints. Great work!

 

[Brolloks]

Great results Chance, excellent work

 

[ChanceCoats123]

Thanks, guys. I would've liked to see the other benchmarks cooperate. I think after the third time Vantage crashed for no reason, I decided it a night. And for some reason, with 01, all the tests were lacking the FPS. My previous best N score was like 975core/1275mem and it got me 1010fps, last night, I was pushing 1150core/1300mem and barely breaking 1000fps.

 

[Angry]

Did you try backing off the clocks a bit?
The memory when clocked to much can actually degrade the performance. Had that issue with my 5830.

 

[Bobnova]

GDDR5 has built in ECC, it'll pretty much never artifact to tell you that it's loosing it, the score just drops. It's brutal that way.

 

[ChanceCoats123]

Weird. Because I set literally all the same settings as a run I had previously done, and the score still had the roughly 5% difference. I literally had to bump core and mem to offset the mysterious loss in FPS.

 

[Bobnova]

Some memory doesn't appreciate being cold much, don't know if that can show itself as slight ECC performance loss or not though.