Overclocked GTX 570 6200ish graphics score 3Dmark 11

[Jay85]

Been messing with the overclock settings again today using 3Dmark11 and heaven benchmark. I managed to score some decent numbers however slightly worries about the voltages I'm using to get them stable.


Overclock settings:
Core: 852
Shader: 1700
Memory: 4002 (2000)
Vcore: 1.065v

Heres the graphics score results:

Stock settings on performance level
3Dmark11: 6200ish
3Dmark Vantage: 21702
Heaven (4xAA, 1600 res as per ocuk thread): 1204


Now I'm running a lot of gear off a corsair HX650w psu and I'm thinking because of this that I'm having to use more volts to get my gpu stable, would I be right in saying this or is the card just a higher volter in general? Also I can run 3Dmark11 on much lower volts but will crash on heaven bench unless I push them much higher.

Has anyone encountered this before? Also if you could post up your results Aswel and most importantly the vcore used and benchmark tests used to test for stability.

Thanks!

 

[harmony]

If you push your PSU to the edge, the current might be less stable. However I don't see 650W anywhere near insufficient.

It is relatively easy to pass 3DMark11 with low voltage. This is not because the load of 3DMark11 is low, but because each test scene is short and the GPU takes a break after every test scene to let the temperature drop, which plays a significant role to stabilize the card.

Meanwhile, Unigine Heaven puts the card on load continuously thus it causes more *continuous* stress and requires more volts to stabilize.

Since nVidia started to cheat in Furmark since a specific driver version (maybe 270.x), my favourite stress test methods for nVidia cards are now Unigine Heaven and Metro 2033 benchmark. (I measure highest power consumption in Unigine Heaven with normal tessellation, 4AA and 16AF, and Metro 2033 with everything max except PhysX.)

 

[Jay85]

If you push your PSU to the edge, the current might be less stable. However I don't see 650W anywhere near insufficient.

It is relatively easy to pass 3DMark11 with low voltage. This is not because the load of 3DMark11 is low, but because each test scene is short and the GPU takes a break after every test scene to let the temperature drop, which plays a significant role to stabilize the card.

Meanwhile, Unigine Heaven puts the card on load continuously thus it causes more *continuous* stress and requires more volts to stabilize.

Since nVidia started to cheat in Furmark since a specific driver version (maybe 270.x), my favourite stress test methods for nVidia cards are now Unigine Heaven and Metro 2033 benchmark. (I measure highest power consumption in Unigine Heaven with normal tessellation, 4AA and 16AF, and Metro 2033 with everything max except PhysX.)

How exactly do you calculate max power consumption during stress testing? Is there some specialist equipment that I would need? Ive had a blue weird blue screens Aswel when I bump up the clock on my CPU from 3.6 to 3.8 while the card is overclocked but seems to run okay without the overclock on the card which pushes me to believe that im lacking power somewhere. I'm also unable to tweak the voltage on motherboard for some reason so I can't raise the vcore on my CPU.

Could a faulty (touch wood) psu cause an unstable supply of voltage to my components? I want to be absolute sure before I make any purchases, last thing I want is to spend £130 on a psu and find nothing has improved. I don't have access to the metro 2033 game unless the benchmark suite is available to download elsewhere (not pirated).

For heaven I ran 3 loops on my said settings and it runs smooth but with a few screen tears (due to vsync off) but otherwise runs at high fps with good
Minimums (lowest being 31.9fps). How many loops do you do for a good stability run?

Thanks.

 

[harmony]

How exactly do you calculate max power consumption during stress testing? Is there some specialist equipment that I would need?

I just used a £15 power meter bought on the internet to measure between the wall and the PSU.

Ive had a blue weird blue screens Aswel when I bump up the clock on my CPU from 3.6 to 3.8 while the card is overclocked but seems to run okay without the overclock on the card which pushes me to believe that im lacking power somewhere. I'm also unable to tweak the voltage on motherboard for some reason so I can't raise the vcore on my CPU. Could a faulty (touch wood) psu cause an unstable supply of voltage to my components? I want to be absolute sure before I make any purchases, last thing I want is to spend £130 on a psu and find nothing has improved.

You mean you can't get it stable if you overclock your CPU and GPU at the same time? I had that PSU before, and it powered up a pair of 5870s without any problem, though I didn't overclock them.

I don't have access to the metro 2033 game unless the benchmark suite is available to download elsewhere (not pirated).

If you have a steam version of Metro 2033, the latest version already includes a file called metro2033benchmark.exe in the game folder, which is not pirated. However I don't think it's available separately without installing the game.

For heaven I ran 3 loops on my said settings and it runs smooth but with a few screen tears (due to vsync off) but otherwise runs at high fps with good
Minimums (lowest being 31.9fps). How many loops do you do for a good stability run?

Well it really depends. Because I'm a heavy gamer when I play games (can keep playing for over 10 hours when I have time) so usually I leave Unigine Heaven running there for a whole night while I sleep and I verify if it's still up and running the next day I get up. (The demo would loop indefinitely and if there is anything unstable from the card, Unigine Heaven would likely to stop upon error such like device closed or whatever error message it is.) Also, the folding@home GPU client is a good alternative - it doesn't stress the card with that much power consumption, but it is very sensitive to errors. I also run this for 10+ hours to verify a set of stable oc variables before I flash the BIOS of my graphics cards for a permanent oc.

 

[Jay85]

I just used a £15 power meter bought on the internet to measure between the wall and the PSU.

You mean you can't get it stable if you overclock your CPU and GPU at the same time? I had that PSU before, and it powered up a pair of 5870s without any problem, though I didn't overclock them.

If you have a steam version of Metro 2033, the latest version already includes a file called metro2033benchmark.exe in the game folder, which is not pirated. However I don't think it's available separately without installing the game.

Well it really depends. Because I'm a heavy gamer when I play games (can keep playing for over 10 hours when I have time) so usually I leave Unigine Heaven running there for a whole night while I sleep and I verify if it's still up and running the next day I get up. (The demo would loop indefinitely and if there is anything unstable from the card, Unigine Heaven would likely to stop upon error such like device closed or whatever error message it is.) Also, the folding@home GPU client is a good alternative - it doesn't stress the card with that much power consumption, but it is very sensitive to errors. I also run this for 10+ hours to verify a set of stable oc variables before I flash the BIOS of my graphics cards for a permanent oc.

In that case I think I'll invest in one and see what my max power draw is at the wall. It would be interesting for me to know what kinda power I'm pulling when I'm benching.

Yes that is correct however it will stabilise when I downclock my CPU from 3.8 to 3.6 while my gpu is still overclocked. I get a bscode: 50 i think which indicates a lack of power to the CPU. But like I've mentioned, I am unable to tweak the vcore in the bios setup for whatever reason and googling comes up blank.

Is the metro 2033 benchmark really that good? Is it worth paying £10 for the game just for the benchmark suite? I'm really tempted of the software is indeed reliable for benching. I've used folding before but never fully understood it's purpose so i stopped using it. Did make my gpu nice and toasty though (ATi 5870).

I think my biggest concern is the vcore I have used to get the overclock to run stable. I've seen others using reference and nom reference cards running the same card but at much lower volts (unsure on what benchmarks they have used hence this thread too). Theres been a few running 930core at 1.1v but at that vcore level I'm only stable on 870core which was disappointing.

 

[harmony]

Your BIOS cannot tweak vcore of CPU? I only encountered such situation on Intel motherboards...

It is not worth £10 for the metro 2033 benchmark alone. The benchmark is just an extra to the game itself

 

[555BUK]

Your voltage looks fine for a 570. My 480 takes 1.125v to run the same clocks and I get almost exactly the same score (linky). Your PSU should also be fine. In the past I ran a 480 alongside an i7 920 @ 4GHz on a Corsair 520W without any issues.

Don't forget it's hot now. The extra 10degrees or so could make the difference between stable and unstable.

 

[Jay85]

Your BIOS cannot tweak vcore of CPU? I only encountered such situation on Intel motherboards...

It is not worth £10 for the metro 2033 benchmark alone. The benchmark is just an extra to the game itself

In the past when i've used Asus boards and Gigabyte boards it has been fine however since upgrading to this board, i can't seem to change from the auto selection to manual tweak. I can change the multipliers and also use the OC genie switch on the board but the Vcore voltage tweaks are unchangeable unless i'm missing something.

I would get the game however i'm slightly ashamed that these games freak me out abit... not a fan of jumpy games!

Your voltage looks fine for a 570. My 480 takes 1.125v to run the same clocks and I get almost exactly the same score (linky). Your PSU should also be fine. In the past I ran a 480 alongside an i7 920 @ 4GHz on a Corsair 520W without any issues.

Don't forget it's hot now. The extra 10degrees or so could make the difference between stable and unstable.

That makes me feel slightly better as i know the 570 is essentially a suited up 480 in disguise but runs cooler and draws less power. I'm running a lower memory clock than you aswel as i dont think i could stabilise it at your level.

Very true about the heat, the extra 10 degrees could make all the difference. I've noticed my cpu temps idle much higher than normal too in this heat from 33 upto 39 when its really hot outside.