Hello,
I'm from Holland, can't write german very good so I try in English; hopefully you'll understand.
I have modded my Powercolor HD4870 with the Accelero S1 V2 and it's working like a charm!
I first had installed it without the famous red backplate but with the RAM sinks on and a sink on the VRM's. Arctic 5 thermal paste for the gpu.
With one 2000rpm 92mm fan aimed at the card the screen stayed blank, no input signal whatsoever. I thought I toasted the 200 Euro card because 1 or 2 times I saw a few characters on the screen of the POST and then it goes black again.
Whatever, I'm going back to 60C for GPU and 82C for VRM (fan at 30%) and try again sometime later; maybe I did something wrong then... I put the stock cooler back and finger-crossed turned on the PC and the card came alive and kicking again.... PFFFFFFFFFFFFFFFFFFFFF !!!!!!!!!! sweat sweat sweat...
On my search on the Net I stumbled on this thread and read it all. Main conclusion was I had to maintain the red backplate and put 4 of the RAM sinks on the chips like in the red backplate picture in the startpost. I used Arctic Silver 5 and added two 92mm fans (2x 120mm fans are on its way as we speak). I use a fancontroller, a Skythe Kaze Master, and configured the fans around 1500rpm. One fan takes the 12v and pulse from the other fan and then a 3 pin cable goes to the fancontroller; gives me 1 pulse and 12v for 2 fans, so it's installed as 1500 x2 = 1500rpm but air displacement at 3000rpm!
The sound is neglible, my HDD's are a bit louder.
Well, I installed the card in my PC, turned it on and it gave me the POST and the BIOS! Booting Windows XP, everything still fine. It auto-started Everest and gave me these temperatures at idle after 10 minutes of waiting for the card to heat up:
(FYI: ambient: 27c / 2 fans @ 1500rpm / OC'd @ 815-1120 stable / Diamond TOP bios update, TT Armor bigtower closed)
All temps in this test are the MAXIMUM temps I monitored during a particular demo, thus the highest temps the card showed me in Everest on a secondary screen in extended viewing mode. Temps varied alot during sequences in ranges of 6c for the GPU and 18c for the VRM (Voltage Regulator Modules) in the demo's and games (thus not in the dreaded Furmark) but never got over the MAX temps showed here.
@ Idle:
With stock cooler - GPU: 60c / VRM: 82c
Accelero S1 rev2 - GPU:
41c / VRM:
59c
A difference of 20c !! Very impressive indeed!
I loaded furmark at default settings, the so-called unrealistic GraKa killer... in which I deeply agree!
60 secs: GPU 60c / VRM 114c (!!) and rising...
120 secs: GPU 65c / VRM 126c (aahhhh!!) and rising... (125c spec max??? huhhhh?? WTF!!!??? ........ AHHHHHH!!!!!)
so I immediately cut off the testing before I toasted the card. From idle to 60 secs in Furmark the VRM temp rise quickly but after that 114c it was going quite slower but never going stable at any point and that stays frightening.
My opinion.... NOT IN ANY WAY REALISTIC because in games you have very active sequences and you have very inactive sequences so the VRM won't hit the 100c ever, which is well within specs.
Up to 3Dmark06... loading at default settings.
SM 2.0 GFX:
GT1: GPU 51c / VRM 80c
GT2: GPU 51c / VRM 84c
Very stable temps but the VRM is again going to rise.
HDR/SM 3.0 GFX:
HDR1: GPU 54c / VRM 86c
HDR2: GPU 51c / VRM 81c
Looks like the second demo (deep freeze) didn't put much stress on the card, unlike the first demo (canyon flight).
Without cooling off the card I proceeded to the game DEMO within 3Dmark06.
GPU 55c / VRM 92c
In which the first demo, return to proxycon, was the most stressfull for the card and in the third demo, canyon flight, were the direct water-scenes quite stressfull.
After returning to the desktop the temp dove quite fast, within 2 minutes, to 42c for the GPU and 59c for the VRM at which it holds steady.
Conclusion:
the GPU NEVER hit 56c. The VRM NEVER hit 93c (besides in Furmark) but it rose very slowly during the testing as you can see so stressing the card long enough will eventually make him hit the 100c for VRM's and higher. After a 1 hour cool-off of the card I got 52c GPU and 88c VRM. So the VRM chips are much too hot for the original back-plate to handle, maybe another, added, cooling alternative or thermal grease will lower the temps or slow it down a bit when rising but I guess the Accelero makes in my opinion a BIG difference in overall temp and didn't add 1 degree in my case temp (ambient) of 27c.
I advise to never cool the 4870 passive, also the 4850. Maybe the 38xx series will do but these high 48xx temps will ruin your card; the airflow in your case alone will simply be not enough, unless you place a deep freezer in front of your intake (so the lollypops will be in armslength also!)
Also I advise you to get rid of that Furmark test because it's unrealistic and it does some gruesome things with your card which will mostly never happen in real (gaming) situations. It's a killer, indeed.
For the record, I did a auto-tune in Ati-Overdrive and the temps were at max.: GPU 51c / VRM 80c.
(very unreliable programming, it always gives you clocksetting in which any game will freeze almost immediately; it gave me 875-1190.... can you imagine? Every game locked up..... even FURMARK (duhhhh!)
For the next few weeks I will monitor the temps during gameplay but I think the cooling will stand.
I will later shoot some pictures of my card and post them
here if you're interested.
I hope I was of some help for the still doubtfull ppl out there.
Porto
ps: FYI I played COD4 yesterday and wrote down the max temps... settings: 1920*1200, everything at MAX, 4x AA) ..... GPU 49c / VRM 73c