The Official Graphics Card and PC gaming Thread

Guess he's good to go then

@Est - just saw your userbar
I'm a gorillaz fan too ;D

 

The second HD4870X2 arrived first.
System is now built, albeit with very messy cables - there are just too many to spend any less than hours tidying them and if I end up using the Zalman instead it'll all be for nowt.
Overall, the BeQuiet is slightly noisier than the HX520, obviously I can't say under full four GPU load as the HX never did that, but in general idle noise is higher.
Idle draw at the wall is 310W or so, highest AC draw I've seen was 770W.
Overall, 3dmark06 doesn't seem to scale in any test but Canyon Flight, which was only enough to counteract Vista's reduction of the CPU score, so my 3dmark currently hasn't changed. I tried raising the GPU speeds but as yet haven't completed a stable run, though I suspect that is CPU OC related not GPU.
I will shortly be running the tests again.
God does this system make a racket when everything's going full pelt. Fortunately very little of said racket is the PSU fan. The X2 fans (or at least one of them) reach 65% under full 4 core loading, presumably as the tight spacing of the card raises the temperatures higher. 55% is the highest I ever saw before.

 

Looks like it's game over for 3dmark06. Superb scaling occurring in Crysis Warhead (1920x1200 All Enthusiast, 4xAA playable), yet my 3dmark06 increased by 400 points. By contrast, my Vantage Performance score rose from 12500 to 19500. Quite chuffed about that.
Idle system draw at 3.7-3.8Ghz: 305W
Single card system draw: 480W
Dual card system draw: 530-580W
Quad card system draw: 550-780W (only experienced 780 in 3dmark vantage fill rate tests, elsewhere not surpassed 725, usually less than 700)
BeQuiet's performance is admirable. Idle noise is higher than I'd like, but even at 720W load, noise level from it seems the same, and can only be 25dB ish, quieter at least than the Toughpower's idle.

Mirrors Edge still being a pain in the butt, but a fresh install of Vista onto my spare Raptor has helped matters a great deal.

Pardon the terrible cable management, when I know which PSU I'll be keeping I'll tidy it up. Right now it's hard to even keep the cables out of the fans and from preventing the case side closing, that's how cramped it is.



http://img3.imageshack.us/img3/7210/s5000645sw1.jpg

 

My god Sam that is nasty. Maybe the wiring has something to do with the higher fan speeds? lol

I think with either PSU, you'll be quite happy. Maybe you could invest in some better card cooling? Is there anything out there for the 4870X2? Either way, you are treading on seldom walked ground.

Can we get some performance comparisons? I'm mostly interested in your Crysis and Warhead scores.

 

Is there a benchmark tool for Warhead like the original?

As far as custom coolers go, other than what comes with other X2s like the Palit, Asus or Gainward, it's watercooling or a homebrew combination of Thermalrights that only allows for one card.

 

Ouch. Again, I state my dislike of quad GPUs LOL!

No benchmark for Warhead AFAIK. Maybe there are some custom benchmark programs floating around? Let's at least see some results from Crysis. Both 2560 and 1920, AA and no plz

 

Crysis Warhead Fraps results (First action sequence, all settings Enthusiast)

1920x1200 25-40
1920x1200 4xAA 22-36
2560x1600 15-26, odd dips to 12
2560x1600 4xAA 12-23, one dip to 10

It's worth noting that the first time I tried 2560 with 4xAA, Crysis maxed out my 4GB of RAM and the system started paging. It took me about 5 minutes to alt-tab, open task manager and close the process after I realised this, that's how badly the system ran. The problem has yet to recur, but that was quite amazing. Idle memory usage is around 700MB, putting Crysis' demand above 3.4GB, well above, as it only reached 80% loading.

Now going to try normal Crysis benchmark tool.

 

Crysis Benchmark tool results
(First run only, don't have the time to run multiples, so results will be slightly lower due to loading lag)

1920x1200 30.03
1920x1200 4xAA 28.29
2560x1600 23.60
2560x1600 4xAA 19.77
All settings Very High.

 

Impressive results for very high But with AA it looks painful to run, lol.

I can see the urge to use AA to, in effect, "max" the game. But from my own experience, AA doesn't do much for the visuals. In fact, with FSAA enabled, it disables the foliage AA already used in the game. So AA should be left to bragging rights benching and no AA should be used for real gameplay, IMO.

I'm really enjoying the ToD mod right now in most parts and not so much in a few parts. During daytime sequences, it has a profound impact on the game and really looks fantastic. But when it gets dark, it REALLY gets dark. Flashlights and night vision are actually required to navigate, lol. It actually hinders gameplay to an extent. Especially in the last parts of Relic(driving in pitch dark anyone?) and Onslaught.

The ToD mod makes it get dark in the end of Onslaught when it was originally a completely daytime level. This makes it really difficult to navigate the huge area. I actually disable the mod for doing that level. Though watching the airstrike on the ammo dump is a real treat with ToD

Awesome "band" IMO. lol

 

I may have to try that mod. You are right about AA in the original Crysis, Warhead's AA is better, but also more demanding.

 

Recent attempts have shown better Crossfire performance with Warhead. With the CCC high config and lots of tweaks, I get about 32-36FPS in normal play. And frequently down to and below 25 in the ice levels or with intense visual fireworks going on. Over all, not bad. But Crysis still runs much smoother even later on. About 36-40FPS in normal play, only dropping to the mid-high 20s or so with intense visuals. Both using nearly identical configs.

 

That's because the scaling is now as good, Warhead is just a much more demanding game than the original Crysis, in stark contrast to what they originally said.

 

Found CrossfireX off this morning, rather strangely. Not sure if last night's tests were using 2 GPUs or 4. Re-ran the Crysis Very High test at 2560x1600 no AA but with 3 runs this time, and got an average of 28.96. 22.99 with 4x AA.

 

Oh yeah. I finally did some OCing on the Phenom II this morning. 3.708GHz, 18 x 206MHz FSB w/ 1.45v. Running Orthos Prime but all looks stable after 8 hours so far. 34*C idle and 58*C load. Nice going AMD, you guys have made an excellent chip

EDIT: LOL!!!!!!!!!!!!!

AMD Phenom II 940 BE @ 3.708GHz 18 x 206MHz 1.45v
2 x 2GB Mushkin XP Ascent PC2-6400 @ DDR2 825 CL 4-4-4-12 2.1v

Core 2 Quad Q6600 @ 3.708GHz 412MHz FSB 1.45v
2 x 1GB Mushkin XP PC2-6400 @ DDR2 825 CL 4-4-3-10 2.1v

 

What voltage?

 

Read my edit. Any higher and it would have needed 1.5v or more. Max that would boot was 3.9GHz w/ 1.55v and a crazy 4GHz at 1.65v. But that didn't stay long

 

Clock for clock then, the Phenom II's probably about the same as the Q6600 then?

 

I'd say so yeah. The memory bandwidth and latency are better, but mostly the same FPS in everything, including Crysis, and the same 3D Mark score. Just under 18000 at default settings. It really is a clock-for-clock equal. And I got both for about the same price.

They even get better temps. Go figure XD AMD really brought back some competition. Check my profile for the final temp readings.

 

Well yes, at stock they did, the Q9400 is the same price I still have to pay £32 more for a Phenom II 940 than a Q6600 though and I get a free game with the latter, so when overclocked, nah. It's good to see AMD finally breaking the 3.3-3.4Ghz barrier for overclocks though, hopefully they'll get a bit further with newer chips and start becoming competitive again.

 

Well it was mostly about finding the right architecture to allow speed, value, and OCing. And they seem to have finally hit the mark with both video cards AND CPUs. So hopefully, everything new from AMD/ATi will improve on Phenom II and HD4000. We might see performance rivals to i7 and the GTX280.