Spring RTS Engine

A newish benchmark was just made which tests the compute performance of your DX11 card. Requires Vista/7 and a DX11 card.

Very helpful in determining real-world overclocking effects, especially since I know at least a few (if not all) Radeon 5xxx series cards throttle the memory clock if you set it too high.

What's nice is that it exclusively tests your GPU. Your CPU and other components don't factor into things at all, but your compute performance translates fairly directly into shader performance, which means game performance... more or less.

http://www.friendsea.com/ComputeMark/

Just remember it's a benchmark, not a stress-test though.

Edit: Scores seem highly repeatable too. My 5850 @ 800/1200 gave 13578. I downclocked it to 800/1100 and got 13204. Then brought it back up to 800/1200 and got 13572.

Edit 2: Interesting, interesting, it's true what they say about the memory throttling. When I take the memory speed above 1250, my score plummets to below-stock speeds.

MS needs to scale back on the hardware specific iterations of DX. Then again, the manufacturers and them have a wonderfully profitable relationship that screws the end user, so not much chance of that changing.

Does OGL support hardware tesselation on non-DX11 hardware? Go build a virtual machine on your calculator or something...

~~~~~

I've got my 5850 running stable now at 880/1250 with 1.15v. Stock is 725/1000 with 1.08v! 5870 stock is 850/1200, not sure of the voltage... probably 1.1 or something. 5850 only has 1440 shaders though versus 1600 on the 5870.

Furmark would probably still kill it of course, but Heaven (which seems to be the most sensitive "normal" test of late) ran fine for about 3 hours.

I just did a before and after comparison... 5850 at stock speed (725/1000) gets 12154 points in 180 seconds. With the same fan settings as I use normally, the temperature peaks at 82 degrees and the fan spins up to 33%, still practically silent (23% is where it becomes audible above a whisper IMO).

With my new overclock, it gets 14582 points in 180 seconds! However, the temperature reaches 86 degrees and the fan spins up to 45% (again, the same custom fan settings as above). Still not a 100+ degree Fermi vacuum-cleaner-jet-engine, so I've got nothing to complain about.

Now I just need a hexacore and I'll get ALL the ladies.

...So.... sorry, normally I wouldn't be so enthusiastic about something like this, I'm just impressed by how accurate it is.

With things like 3DMark or whatever, you can do the same test 3 times and get significantly different results, with this one you can do it 10 times and be within 10 points from beginning to end.

Or you can bump your memory speed by as little as 10 mhz and you will see a definite repeatable effect on every test after that, etc.

I don't know how well it represents actual gaming performance, and the amount of VRAM you have will obviously have a major effect on things if you have high resolutions or high anti-aliasing mixed in, but I still think it's impressive.

OpenGL can actually tesselate via geometry shaders on any OGL 3 compliant card, which IIRC is gf8 series and above, and it'll do it in XP too

wbt radeon

Overclocking can be really fun but is often useless.
In my case i could get maybe 5 frames more out of my gfx card by overclocking.
I dont think its worth any risk.

Also my new phenom 965, I overclocked it but its not of so much use+ my psu is a little weak. (humming sound at 100% cpu usage)

So I set it back to default and undervolted it, I like my pc as silent as possible.
Now it is more silent than the water in a heater when its turned on, with air cooling.

I agree with you half-way... it's true that overclocking doesn't usually have a linear performance improvement, but it's effectively "free" and there's very little risk unless you're reckless. In my case I've got a free performance upgrade from a 5850 to a 5870, which costs about $100 more.

On the other hand I do like my computer environment to be noiseless, so what I've focused on more lately is shopping with efficiency higher in my priority list.

As a result, basically the only part of my PC that makes any noise at all is the GPU, and even then it's still silent unless I'm playing a game.

I set up a custom fan profile with MSI Afterburner which keeps the fan running at 25% until it reaches 70 degrees, at which time it jumps to 30% and increases linearly from that point up to 46% at 84 degrees.

Beyond that the speed rises quickly for the sake of preventing a disaster in case something actually pushes it that high... but aside from purpose-built stress testers like Furmark, nothing does.

edit: I forgot to mention, I actually tried my settings above (880/1250) in Furmark... it got pretty warm but it didn't crash!