CPU Scaling Benchmarks (X2 + 4850)

[kllrnohj]

I finally upgraded from my aging 7950GT to a sexy new 4850. However, my CPU is a bit on the slow side (3ghz X2), especially when compared to the rigs that the major benchmark sites are using. Since there hasn't been a CPU scaling article recently, I decided to make my own.

System specs:
AMD Athlon64 X2 5000+ BE
Asus M2N32-SLI Deluxe wireless
Sapphire 4850 512mb @ stock
4x1GB Corsair XMS2
Vista Ultimate 64-bit

Using Cat 8.7s

First, the GPU-Z of the card:

Here are the CPU-Z screens for the 3 CPU frequencies I used

First up, is the GRID Demo. My testing is similar to that of [H]'s in that I simply did a 3 lap race and recorded the FPS using FRAPS. I only did the race ONCE per setting. Graphics settings were maxed out in the game, 8xAA (MSAA in game - no edge detect), 16xAF (forced through CCC), and the resolution is 1280x1024. I did the Drift race in the demo for each test.

In GRID, there does seem to be a CPU bottleneck going on, even with 8xAA. Every 200mhz bump seems to correspond to a ~3fps boost in average FPS. So a 15.3% rise in clockspeed (2.6->3) brought a 12.5% rise in average FPS(50.43->56.75). However, it should be noted that even at 2.6ghz, the X2 still managed to be fast enough to keep a minimum FPS of 37, which is very playable. So even if you have an older CPU, it won't be too much of a bottleneck in GRID, unless you find 37 fps minimum to be unacceptable.

Here are the CPU scaling results for Guitar Hero 3 (all settings maxed, crowd at "packed", 8xAA/16xAF (through CCC). Song was "Raining Blood" by Slayer on Expert in the Kajiju Megadome venue. I only did 2.6ghz and 3ghz on this test after seeing how well 2.6ghz did.

Again, a faster CPU is a benefit, but both were playable (note: the 2.6ghz test had a small hiccup and a very slight bit of stutter for 1-2 seconds, the 3ghz one had no such stutter - not sure if it was related to the CPU, game, or just vista deciding to choke for a second)

// MORE GAMES TO COME //

So far all I've done is GRID, and next will likely be Supreme Commander - anyone have any other suggestions for games they would like to see tested? If I have it, I'll test it

 

[gorilla show]

Great, I built my girlfriend an X2 system with a 4000+. I overclocked it to 2.7GHz and it's nice to know that she can get a 4850 and expect some nice fps on a 19" monitor.

Thanks.

 

[wolf2009]

nice, i was thinking of doing the same thing .

i think your choice of race was wrong, drift races are hard to replicate since you cant drift the same every time. You should choose a more identical benchmark like world in conflict, Company of heroes, or even in grid a circuit race .

 

[BioSergeant]

Good benchmark. Those are some high AA setting though. I'm usually happy with no AA as long as the monitor is at it's native resolution. In that case, I don't need as much of a powerful CPU as some would suggest getting.

I would like to see Company of Heroes. But instead of using the standard video stress test. Use actual gameplay to make the benches.

 

[kllrnohj]

nice, i was thinking of doing the same thing .

i think your choice of race was wrong, drift races are hard to replicate since you cant drift the same every time. You should choose a more identical benchmark like world in conflict, Company of heroes, or even in grid a circuit race .

I thought of that, but since my wheel is still packed up (just moved), it is much easier to get more repeatable runs with drifting on an xbox 360 controller. That, and I'm not used to GRID yet, so my circuit runs tend to be me running everyone off the road - which isn't very repeatable either (but its so much fun ) The scores really don't line up with the average FPS results, and the overall race time was within a couple of seconds on each one.

Good benchmark. Those are some high AA setting though. I'm usually happy with no AA as long as the monitor is at it's native resolution. In that case, I don't need as much of a powerful CPU as some would suggest getting.

I would like to see Company of Heroes. But instead of using the standard video stress test. Use actual gameplay to make the benches.

AA doesn't use any CPU power - it is completely on the GPU. So since it looks like I'm CPU bottlenecked, the AA is likely free. That said, I found that GRID needed AA. On my 7950GT I dropped a couple settings down just to get AA.

EDIT: Grid 3ghz w/o AA: min/max/avg: 34, 68, 47.279, so the AA is, in fact, free

So I got recommendations for CoH and WIC - I don't have either of those games, but both have demos, so I'll give them a shot.