5850 overclocking gains: 725/1000 compared to 950/1200 in Serious Sam HD

[dug777]

For those of you interested in 5850 results while overclocking in a game that you don't see reviewed or tested very often (or indeed at all!) and at settings you might not see reviewed/tested, I whipped up a quick comparison in Serious Sam HD: The First Encounter

Settings are 1920x1200, everything maxed in the various control panels (some customisation to crank it even more within the GPU settings in game), but 4xAA set in the game, and SSAA is set in CCC.

Specs of computer are : Q6600@2.7Ghz/Asus P5Q PRO/4GB DDR2/XFX 5850 BE.

Blue line is stock 5850 clocks - Avg of 57.033
Purple is stock XFX Black Edition clocks (765/1125) - Avg fps of 61.367
Red is 950/1200 - Avg fps of 68.178
Green is also 950/1200 but dropped the CPU to 2.4Ghz to check for CPU bottlenecks - Avg FPS of 67.633

http://pics.bbzzdd.com/users/dug777/5850SSHD.jpg

The FPS collapse at the end is the demo recording I set up stopping and going back to the menu screen.

The GPU overclock makes a huge difference to how smooth it all is in real world play, but I can't really explain the bit where they all converge at ~75-83 seconds as it doesn't seem to budge regardless of CPU speed or GPU/RAM clocks, maybe it's a system memory bandwidth issue (my RAM is nothing fancy, probably running at reasonably loose settings in dual channel mode at some speed I forget)?

Any thoughts, comments or queries? I could run again at 3.3ghz if anyone was interested

For amusement value, here is 725/100 (stock reference 5850) against 1000/1225:

http://pics.bbzzdd.com/users/dug777/5850SSHD4.jpg

Avg fps of 70.156 compared to stock avg of 57.033

 

[Kenmitch]

How about doing runs with 950 core and memory at 1000, 1100, 1200 to see how much of a diff it makes on FPS. Think it may be helpful to those on the edge about flashing the bios to the hacked asus 5870 bios as it sometimes will let you get to 1300 or close. i know on 3dmark Vantage it makes somewhat of a diff.

 

[dug777]

How about doing runs with 950 core and memory at 1000, 1100, 1200 to see how much of a diff it makes on FPS. Think it may be helpful to those on the edge about flashing the bios to the hacked asus 5870 bios as it sometimes will let you get to 1300 or close. i know on 3dmark Vantage it makes somewhat of a diff.

Your wish is my command

http://pics.bbzzdd.com/users/dug777/5850SSHD2.jpg

Apologies, it is getting a bit messy there

EDIT:

Here are all the 950 runs by themselves:

http://pics.bbzzdd.com/users/dug777/5850SSHD3.jpg

 

[dug777]

For amusement value, here is 725/100 (stock reference 5850) against 1000/1225:

http://pics.bbzzdd.com/users/dug777/5850SSHD4.jpg

 

[ArchAngel777]

For amusement value, here is 725/100 (stock reference 5850) against 1000/1225:

http://pics.bbzzdd.com/users/dug777/5850SSHD4.jpg

Thanks for taking the time to post this. The gains are substantial. It has me considering this card...

 

[dug777]

Thanks for taking the time to post this. The gains are substantial. It has me considering this card...

No worries mate. At least you know it's raw power and the real deal behind a hokey custom benchmark like this, I can't imagine anything less likely that ATi or Nvidia would optimise for

 

[solofly]

dug,
Why are you running your Q6600 only at 2.7?

 

[konakona]

damn, looks like core (or shaders, probably) matters more than memory. I am thermally limited on the VRMs, gonna try to put on some copper shims once I get some time to work on the card again.

 

[FragKrag]

Those gains are so damn impressive.

Makes GTX 470 vs HD 5850 so hard ;_;

 

[Kenmitch]

Your wish is my command

http://pics.bbzzdd.com/users/dug777/5850SSHD2.jpg

Apologies, it is getting a bit messy there

EDIT:

Here are all the 950 runs by themselves:

http://pics.bbzzdd.com/users/dug777/5850SSHD3.jpg

Well looks like your memory is scaling fine up to 1200 without engaging error correction.

Looks about what I'd expect. You would have about the same gains if you could get to the magic 1300. Any free speed s worth the effort!

I'm sure this will be helpful to others. Just watch the heat guys/gals

 

[Kenmitch]

damn, looks like core (or shaders, probably) matters more than memory. I am thermally limited on the VRMs, gonna try to put on some copper shims once I get some time to work on the card again.

Those damn vregs get hot! Reason I bought a water block for mine. Just to keep them in the 80's I had to pump the fan speed to unbearable noise levels at specs in sig when testing with furmark for stability. Under water with specs in sig temps dropped alot 40 min plus furmark stability test core was 38 vregs low 50's

 

[dug777]

dug,
Why are you running your Q6600 only at 2.7?

Lazy and greedy for the power savings, both lame excuses

I just use the Asus EPU-6 software for the oc mainly, as it drops me down below 1v for idle at ~1600mhz, but the chip is good for 3.4 or so no worries.

Pretty happy with the performance benefits in situations like this in any even

 

[Kenmitch]

Lazy and greedy for the power savings, both lame excuses

I just use the Asus EPU-6 software for the oc mainly, as it drops me down below 1v for idle at ~1600mhz, but the chip is good for 3.4 or so no worries.

Pretty happy with the performance benefits in situations like this in any even

I kinda like to take full control when overclocking my stuff. I think doing it in bios is the best way to do it. But 3rd party apps do get the job done for the Lazy as you say

When overclocking your cpu you don't have to whack all the energy saving features like in the guides. I like to leave the C1E and Speedstep enabled when overclocking. Doesn't seem to effect any chip I've overclocked yet. Mine in sig idles at 1800mhz @1.15v with it enabled which is plenty of power for surfing the web and doing the light duty stuff.

 

[Tempered81]

thanks for the benches dug

 

[MagickMan]

I kinda like to take full control when overclocking my stuff. I think doing it in bios is the best way to do it. But 3rd party apps do get the job done for the Lazy as you say

When overclocking your cpu you don't have to whack all the energy saving features like in the guides. I like to leave the C1E and Speedstep enabled when overclocking. Doesn't seem to effect any chip I've overclocked yet. Mine in sig idles at 1800mhz @1.15v with it enabled which is plenty of power for surfing the web and doing the light duty stuff.

I'm not "lazy", I just like having a lower power profile as my card default. I did notice a difference in wattage, even at idle. So I just let Afterburner handle it and I adjust my OC according to how much power I need; 925/1300 for WoW, 975/1300 for modern games, and 1050/1300 for benchmarking (all set on hotkeys). Works really well.

I also leave C1E on, otherwise I don't consider it a valid OC.

 

[Kenmitch]

I'm not "lazy", I just like having a lower power profile as my card default. I did notice a difference in wattage, even at idle. So I just let Afterburner handle it and I adjust my OC according to how much power I need; 925/1300 for WoW, 975/1300 for modern games, and 1050/1300 for benchmarking (all set on hotkeys). Works really well.

I also leave C1E on, otherwise I don't consider it a valid OC.

We was talking about CPU's

But now that you butt'd in....How's your temps on your card at those overclocks? Kinda like to see what they are at about 15min into Furmark stress test if you can swing it @1050/1300

After I figured out my setback on my overclocking(Frick'n cpu vtt needed to .025v it up) I'm currently testing mine out at 1025/1300 1.287v which is 100% stable....Breaks 20k in 3dmark Vantage now....GPU anyways my CPU lags behind even at 4.2ghz go figure!

Guess I'll have to move my CCC sliders to 1075 core 1350 memory and do 1.30v and see how high she'll fly and get it over with....But I kinda like to milk the overclock along as once I get to the top the fun is over!

Damn....It looks like my single loop cooling solution is rapidly getting to it's thermal limits! Sucks when you gotta downclock the cpu just to see how high the gpu will go!

 

[MagickMan]

1050/1300 @1.31V with my new cooling setup is 82C GPU and 98C on the VRMs after 20 minutes of Kombustor with the side of the case open. I kept thinking, "oh shit, they're gonna hit 100C" but they didn't. I had my finger on the Esc key and was a little nervous. haha

 

[toyota]

1050/1300 @1.31V with my new cooling setup is 82C GPU and 98C on the VRMs after 20 minutes of Kombustor with the side of the case open. I kept thinking, "oh shit, they're gonna hit 100C" but they didn't. I had my finger on the Esc key and was a little nervous. haha

and comments like this remind of why why I dont buy used video cards or cpus. lol

 

[Kenmitch]

1050/1300 @1.31V with my new cooling setup is 82C GPU and 98C on the VRMs after 20 minutes of Kombustor with the side of the case open. I kept thinking, "oh shit, they're gonna hit 100C" but they didn't. I had my finger on the Esc key and was a little nervous. haha

When did you get a fermi?

Ouch....Glad I went to a full water block before I decided to see what my card will do.

Those temps are probably no worse than somebodies at stock speed shoved in a crappy case with poor airflow tho

I'll boot up my sytem a little later and post a couple of shots at 1025/1300 about 15min into furmark stability test to compare. I think I'm gonna have to stop at that for now as my single loop has reached it's max cooling compacity it looks like.

 

[MagickMan]

and comments like this remind of why why I dont buy used video cards or cpus. lol

No kidding! I usually don't flog my video cards like this, but my 5850 OCs so well I can't help myself. For games my temps never reach those levels and I use more conservative clocks, but sometimes I get a wild hair and push this sucker a little further.

 

[MagickMan]

When did you get a fermi?

Ouch....Glad I went to a full water block before I decided to see what my card will do.

Those temps are probably no worse than somebodies at stock speed shoved in a crappy case with poor airflow tho

I'll boot up my sytem a little later and post a couple of shots at 1025/1300 about 15min into furmark stability test to compare. I think I'm gonna have to stop at that for now as my single loop has reached it's max cooling compacity it looks like.

Time for a second loop or a bigger rad?

I did play Dirt2 the other night for 3 hours at 1050/1300 and it was flawless, temps weren't anywhere close to that level (15C lower for GPU and 35C lower for VRMs). It was completely within my "comfort zone", but I just don't need that kind of power right now. It looked the same as 975/1300 while playing (perfectly smooth) so why burn the extra electricity?

 

[Kenmitch]

OP....Sorry to get off topic you know how it is around here

Did you get any higher on your memory speed?

and comments like this remind of why why I dont buy used video cards or cpus. lol

Just for your information I draw the line at 70*C in case you see any of my items available for sale

I'm still deciding on what I wanna do with the HAF 932 I picked up as far as cooling goes. I'm thinking I'll just get a 120.3 for uptop....Something like the XSPC RX360 Radiator (Version 2) as I like the look of it. http://www.koolertek.com/computer-parts/pc/XSPC-RX360-Radiator-Version-2-202p2217.htm

I may keep my single 120mm rad also....Haven't decided yet tho. My dual is one of those whimpy 120.2's about 1" thick.

 

[konakona]

I take it those dedicated VRM coolers are working quite well. Still, they don't seem to cover the VRM for the VRAM (the lone VRM toward the back of the 5850, or was it in the front?)

 

[soccerballtux]

No worries mate. At least you know it's raw power and the real deal behind a hokey custom benchmark like this, I can't imagine anything less likely that ATi or Nvidia would optimise for

this is a very interesting thread. Love these benches. Looks like AMD's cards scale great.

 

[soccerballtux]

I couldn't care less if VRMs or some non-complex component gets up to 120C. 125 is danger territory, everything else fine.