Which benchmark best matches AH?

[bockko]

Hello all, can anyone tell me which benchmark best correlates to AH?  I see the benchmarks on tomshardware, but I am not exactly sure which ones to use in picking hardware for the game. I assume its a dx9 benchmark, but from there I am not familiar enough with the frame rate tests to know which one is the most cpu intensive.

Thanks!

[humble]

From what I understand AH is very CPU intensive...I'm not sure that any benchmark really fits it well. I've got a amd 64 3200+ & X800 pro. I OC'd both the card (550/1050) and MB (220FSB) moderately and it's benchmarking ~68000 in aquamark and ~11800 in 3D mark 2003. In AH it'll run at 75 up high but fall to mid 30's in hvy furball or right on the deck. Ave in tower is 28/29.

No appreciable differance in stock 475/900 200FSB vs "OC" even though there is a pretty big jump in benchmark 58000/68000 and 10600/11800....

[eagl]

3dmark 2001SE seems to best match AH performance in my opinion.

[Skuzzy]

Actually, Aquamark is probably a better one eagl.  It leans towards the CPU, but uses DX9 shaders in many of the scenes.

None of them will be a true reflection of how the game will run, but would be okay for a relative comparison.

[bockko]

thanks gents -- if keeping a pc "current" doesnt make the pc owner the donkey chasing the carrot, I don't know what is...i'd complain, except it IS fun!

[eagl]

True.  3dmark2001 won't do any dx9 stuff at all.  From what I saw in performance between my GF4-4200 and GF 6800GT though, having hardware dx9 support on the gpu doesn't seem to really do much for framerate in AH2.

Maybe I didn't look hard enough for the AH2 differences, but until I went up to the A64 the main thing I got going from a non DX9 card to a DX9 card was the ability to run FSAA without a framerate penalty.  Since AH looks fine and runs ok on even an older GF2, it seems like the dx9 use in AH2 just isn't that stressful on the gpu and doesn't rely much on hardware dx9 support.

I'd like to see benchies of some of the GF 5xxx series cards compared to the GF4-MX cards.  Some of the slower GF 5xxx cards like the 5200 are really just GF4-MX cards with very basic DX9 hardware support tacked on.  Comparing AH framerates between some of those cards might show whether hardware dx9 support really does anything at all for AH2.  From my experience with the GF4-4200 though, I'd bet the difference would be minimal.  My GF4 would blow away any lower end GF5 card in AH framerates even though it has zero dx9 hardware features.  DX9 support on the GF4 is supposedly 100% driver/software based.

It'd be funny if HT accidentally left on a flag that forces AH2 into "software" dx9 support on all cards  I doubt that's what's going on, but AH2 runs so well on older cards that it's tough to see what hardware dx9 gets you.

[Skuzzy]

The game will fool you.  It is more dynamic about what it does and how it does it.  Older cards cannot run the DX9 shaders in the game so they revert to an older form of the shaders.
Hence the display differences in the animated water.

Obviously I cannot get into the nuts and bolts on how it all works without giving away the farm.  All I can say is the game will abuse the poop out of video cards if given the chance.
You cannot go by the frame rates.  It is not a good indicator of what is being used and where and when.

Oh, we tested with software mode, and it is terrible.  The frame rates tank on any card.  Just asking too much of the CPU with all the model changes to pile the video on it too.

[eagl]

Thx for the explanation skuzzy.  I guess not knowing exactly what to look for makes it tough to see all the differences between the alternate render paths.

Will HT ever offer the option to select which render path to use?  For example, could I select the path for older non-dx9 cards even when using my 6800gt?  My priorities in graphics quality have always been framerate first, visibility distance second, resolution third, distant detail fourth, and overall image "quality" last.  If disabling specific enhancements would let me run the sliders all the way to the left giving me full visibility range while still keeping high framerates (over 40 is my goal), then I'd probably end up flying that way.  The only way to do that right now is to lower resolution, and that looks ugly and sometimes makes it tough to read gauges or see details like enemy bank angles and stuff when farther away.

I suppose I could load up the old nvidia 30.82 drivers and see what happens, but my guess is that the game wouldn't even run.

[Grits]

61.76 is the oldest nVidia driver that supports the 6800 series. 53.03 was the oldest driver I got to work with my 5200 and 5900.

[Skuzzy]

Shader performance will normally run best when running in the video card's native mode.  However, there are exceptions.

The FX5xxx family of cards cannot run shaders well in thier native mode.

I do not recall eagl, but what video card do you have?  I had thought you were pretty CPU limited, but I may be recalling bad data.  The 30.82 drivers will not work as they are not DX9 ready.  The 61.77 drivers appear to be the most stable for older NVidia cards, including the FX5xxx cards.

Forcing a path is more difficult than it sounds, as it is not a path, per se.  The things which would effect performance the most, in the game, are already available as buttons, checkboxes, and sliders.

[bockko]

May I ask a follow up question?

Which processor will run the game better -- northwood (3.2?), prescott (3.2?) or athlon 64 (say, 3.2?). I am familiar with the northwood core p4's -- like them; not familiar with either the prescotts or athlon 64s (either skt 754 or 939) -

thanks again!

[Skuzzy]

I cannot comment to the AMD part, but the Northwood, at the same clock, is faster and runs cooler (much, much cooler) than the Prescott.
Anyone with a Prescott is living on borrowed time.  The components surrounding Prescott will die a premature death, due to heat exposure.

[IronDog]

Just a blip reference the old cards verse newer cards.Replaced my Ti 4200 with a 5900XT,and lost fps.Somehow  that doesn't seem right,but I don't know that much about the intricasies of video cards.

IronDog 475th FG Satans Angels

[humble]

Originally posted by IronDog
Just a blip reference the old cards verse newer cards.Replaced my Ti 4200 with a 5900XT,and lost fps.Somehow  that doesn't seem right,but I don't know that much about the intricasies of video cards.

IronDog 475th FG Satans Angels

The Ti-4200 is a horse...at default settings I was getting ~60 FPS...when I bumped up the resolution to 1280x1024 FPS actually went up to my max of 75. Now it would fall of into the low 30's over a porked base or in a hvy furball...or down on the deck.

TheX800 pro gives me almost identical performance....but with the sliders all "maxed".....from looking at all the benchmark data I decided to hang onto the Ti-4200 till I was ready to bite the bullet on a topend card....

*** looked up the 5900 aquamark score and its pretty good....did you benchmark it in aquamark? I had a problem with my X800...when I benchmarked it and got 16000 instead of the 56000 it should I knew I had screwed the pooch somewhere on the install....after tweaking i've got it up to 68000....

[Overlag]

Originally posted by bockko
May I ask a follow up question?

Which processor will run the game better -- northwood (3.2?), prescott (3.2?) or athlon 64 (say, 3.2?). I am familiar with the northwood core p4's -- like them; not familiar with either the prescotts or athlon 64s (either skt 754 or 939) -

thanks again!

a A64 3200 will beat both of those with ease, and will cost less too.