Overclocking Crossfire 5870's help please

[vulcan78]

Hi everyone on techinferno, Unclewebb told me this is THE PLACE for Alienware enthusiasts.

I have an M17x R2 with an i7 920xm, 6 GB RAM, and an X-fire set up. I have had some success with throttlestop, my present settings are TDP/TDC 90/75 with a 26 multiplier, no BIOS OC, default voltage. I want to flash my cards with a value that wont compete with the CPU for juice as I believe some values will do this.

Is 800/1100 a good value? What about voltage?

I assume the instructions in the following links which were originally for 4870 cards are applicable to the 5870's? Are these the best, recent instructions to use?

Alienware M17x w/ATi 4870 info thread!

Thanks in advance.

[StamatisX]

Welcome to T|I vulcan,

you would like to check two tutorials, one for performing a CPU and GPU retention mod in order to improve your temperatures

Alienware M17x Teardown & Cooling Modification | Tech|Inferno

and a second tutorial on how to overclock your GPUs

Miscellaneous Tutorials | Tech|Inferno

(there is also an update of this guide at the end)

800/1100 is a very moderate overclock and most likely won't require any voltage increase (I can run mine up to 850/1160 without voltage but I have performed a retention mod and replaced thermal pads with MX-4 so I have very low temps).

I can run mine at 950/1220 when gaming without any issues (my temperatures never exceed 89C)

your CPU with the current settings probably won't be able to sustain a 26x multiplier, so you may want to consider at least a 3% BIOS OC with 75mA and 1.6 for the RAM and a multi of 25x (running wprime 1.55 with TS open so you can monitor your temps will give you a good idea of the situation)

Let us know if you have any further questions

[vulcan78]

Thanks for the reply.

Well I did it! I had great success with a 800/1100 OC flash.

Monitoring in Furmark the temps were 77 and 83 after 10 minutes. I am concerned as to why there is such a big temperature difference. Is this typical?

I tried an 825/1200 flash but had the AMD drivers fail in a stability run in Furmark.

What voltage should I be running at 825/1200?

Ive done the retention mod as well, using Shin Etsu X23 and Tuniq MX-3.

How much longer can I use x26 TDP/TDC 90/75? I havent had any issues yet, I attempted to monitor with wprime but I have version 2.04, I suppose I need ver. 1.55?

In the bench on throttlestop the FID did not back off from 26. The 1024M run time was 371. or so, I know its not the greatest.

Would I get better performance with 3% OC, 75 mV and x25 over 0 OC, standard voltage and x26 multiplier?

What voltage increase should I use for the 825/1200 settings on the GPU's?

Is there a better GPU OC setting you can recommend? I may just go back to the 800/1100 as it was stable with that setting, Im not comfortable with playing around with the voltage until I get an idea of what safe amounts others are using.

Edited June 10, 2011 by vulcan78

[StamatisX]

Normally you should see the primary GPU run hotter than the secondary, that you see a hotter secondary GPU could probably mean that your retention mod wasn't that good or you need to reapply thermal paste (or both).

wprime 1.55 or 2.04 doesn't really matter, just keep an eye on your multipliers and your temperatures with TrottleStop on while you run the test, I for instance use 92/88 and still get my CPU to the limits, I cannot maintain 3.5Ghz, at the end of the test I am at 2.997 GHz.

By doing an OC via BIOS you will increase your systems overall performance not just the CPU's, that's why it is prefered over increasing just the multipliers.

For 825/1200 you probably won't need any voltage increase, just run 3DMark11 and if you complete the test without crashing then you are ok. You can try 840/1150 and see how it goes (again it depends on the ambient temps and how good overclockers your cards are).

The maximum voltage you can get out of your cards is 1.15V so worst case scenario they will shut off if they overheat, my advise is that for clocks below 900/1200 don't exceed 1.1V cause you will just add more heat (increase it only if you are unstable)

[vulcan78]

So I finally ran wprime 1.55, apparently its 1024M is more rigorous than its newer version 2.04. Originally using a 26 multiplier and TDC/TDP 90/75 via throttlestop as thats what was possible using wprime 2.04, I got BSOD using wprime 1.55 with 8 threads going.

I reduced the multiplier to 25 and it passed wprime 1.55's 1024M test in 285.63 seconds with a max CPU temperature of 97 C.

I havent increased the voltage or overclocking in BIOS, also the DDR3 voltage is at 1.5.

I tried a combination of 3% BIOS OC, 75 mV, and a 25 multiplier TDP/TDC 90/75 and got BSOD in wprime 1.55. But not only that, the reported speed in CPU'z and throttle stop was fluctuating wildly, dropping as low as 900 Hz, and the FID was scaling back significantly as well. With just throttle stop and no BIOS OC or additional voltage, the reported speed is much more stable, fluctuating between 3.1 and 3.3 Mhz and the FID only scales back to low 24's (25x).

With the GPU's, I got BSOD running 825/1200 in Furmark under a stability test. I went with an 825/1150 which passed Furmark with flying colors but had Crysis 2 freeze up on me, Im not sure if its throttlestop or the overclocked cards or whether they are now both competing for a limited amount of incoming wattage that is culprit. I reflashed the cards to 800/1100 and will run a stress test with Crysis 2, fingers crossed.

I think with the cards moderately overclocked and throttlestop also drawing more current, the problem may be a power supply issue, not in that the power supply is faulty, its just that the increased demand isnt something it can accommodate.

Im not shooting for a bench OC, I am trying to attain a solid, everyday use OC as high as possible but keeping the CPU temperatures in the mid 90's and the GPU temperatures in the low 80's.

It seems that its only with wprime 1.55 that I get such high temps. While playing Crysis 2 the CPU was in the low 80's, with the GPU's around 72 and 74.

Edited June 10, 2011 by vulcan78

[StamatisX]

ok it seems you wont need the extra voltage since you don't BSOD with your current CPU settings.

For the GPU find first your stable memory frequency i.e.1150 and then see what frequencies your card can handle with the current temps and voltage, if you have headroom with temps but BSOD increase voltage.

If you want a more intense cpu stability test use prime 95.