Nvidia HQ vs Q vs P vs HP Benchmarks for ya...

[dug777]

Thought these benchmarks might interest some of ya, i started with the Source Stress test since it's very quick and easy to run

These are run on a tbred-b@2.1Ghz, 1gb RAM, ASUS A7V8X mobo, and a AGP 6600GT@583/1140mhz...

CSS Stress Test

Firstly at 1280x960, 2xAA, 8xAF, everything else maxed in the game video options.

HQ Q P HP

64.73 67.38 67.22 68.09

1280x960 with 4xAA/16AF, everything else maxed in the game video options.

HQ Q P HP

51.18 58.12 58.52 59.74

Now at 1600x1200, 2xAA, 8xAF, everything else maxed in the game video options.

HQ Q P HP

21.35 25.39 26.47 26.89


As you can see, there's a relatively significant drop going from Q to HQ, but an extremely small difference between any of the other settings...

I certainly feel the difference in IQ between HQ and the other settings is significant, and personally i always run with HQ...i was interested to see at 1600x1200 while fps only dropped ~5.5fps from HP to HQ, without seeing the figures i would have said it was a much greater drop, as it was extremely jerky with HQ, and still relatively smooth, albeit slow, with the other settings...

I'll do a few more at intermediate settings, but i suspect below 1280x960 with Source my CPU is going to start limiting things...i'll try a 1280x960 with 4xAA/16AF now & have those up shortly...

EDIT:

i also tried Q and HQ in 3dmark05, and it's a total joke compared to real life, making almost no impact on the scores at all...went from 3038 to 3091

EDIT1: Added 4xAA/16xAF, quite a dramatic drop there...

EDIT2:

Bleh, well with 91.33, the same oc and all the optimizations off (which i assume equates to HQ under the old CP?) the fps at 1280x960, 4xAA/16AF drops to 49.66 (an average of three runs, all within 49.XX)...no idea what the other settings equate to since the driver is a beta & i can only use advanced view...where no matter what i select quality wise it doesn't seem to change the optimizations.

I'll try another with all the optimizations on now...

EDIT: double bleh, with all the optimizations turned on, the fps only jumped to 55.36, and the texture crawl/shimmering was vomit inducingly bad

So yeah, i think i'll be going back to my 84.43s Slower HQ, no thanks

EDIT3:

Well, tried the 91.31s, and while i only did HQ and Q, it confirmed that the 90 series ones bring a performance drop in HL2 for my 6600GT.

1280x960 with 4xAA/16AF, everything else maxed in the game video options.

HQ Q

49.29 55.27

 

[The Sly Syl]

Can you get some screenshots of each image type? I'm curious as to what actually changes. (6800GT (massively OC'd) here as well)

 

[dug777]

Originally posted by: The Sly Syl
Can you get some screenshots of each image type? I'm curious as to what actually changes. (6800GT (massively OC'd) here as well)

sure i'll do that later tonight mate

I was going to, but i couldn't remember the screenshot key for source games while i was doing it...certainly texture crawl/shimmering was far more noticeable once you stepped down from High Quality, but i didn't notice any real change between Q/P/HP while i was watching the benchmark...

 

[dug777]

I guess the moral of the story is that you shouldn't kid yourself anything lower than Q will bring any worthwhile performance gains I benched FEAR as well, will have those results up at some stage too

 

[schneiderguy]

what driver version are you using? just wondering

 

[Polish3d]

Interesting. Hopefully Nvidia will fix this HQ issue, their generally poorer image quality regardless, and their huge framerate swings (ie, lower mins) with G80 which I might buy. Otherwise I'll just wait. The experience going from 7800GTX to x1900XTX was kind of like going from 4400+ to E6600

 

[dug777]

Originally posted by: schneiderguy
what driver version are you using? just wondering

The last ones that were entirely stable at my overclock, so whql 84.43.

I'll try the beta 91.33s with the same benchmarks, hopefully my oc will be ok with them, but the 91.45s didn't play ball

Downloading now, so i'll update the scores in an hour or so i guess after i've installed & run them...

 

[dug777]

Bleh, well with 91.33, the same oc and all the optimizations off (which i assume equates to HQ under the old CP?) the fps at 1280x960, 4xAA/16AF drops to 49.66 (an average of three runs, all within 49.XX)...no idea what the other settings equate to since the driver is a beta & i can only use advanced view...where no matter what i select quality wise it doesn't seem to change the optimizations.

I'll try another with all the optimizations on now...

EDIT: double bleh, with all the optimizations turned on, the fps only jumped to 55.36, and the texture crawl/shimmering was vomit inducingly bad

So yeah, i think i'll be going back to my 84.43s Slower HQ, no thanks

 

[tuteja1986]

Most Website benchmark Nvidia's GPU on Image Settings: Quality which is pretty stupid :*( as they should bechmark on HQ but wait nvidia takes a massive hit so it wouldn't be good of nvidia .

Anyways XBit lab is one of the biggest offender and they now really cater for the noobs. Only the good i can think that the website do now days is alot of benchmarking of alot games but useless for high end gpu with setting they use.

 

[Elfear]

Thanks for taking the time to do that dug777.

On a side note, Behardware.com seems to test both cards at HQ settings and they have some nice reviews up with plenty of games.

 

[dug777]

Originally posted by: Elfear
Thanks for taking the time to do that dug777.

On a side note, Behardware.com seems to test both cards at HQ settings and they have some nice reviews up with plenty of games.

No worries

I'll get off my backside and use the whql 91.31s today i think, since i'm concerned my fps dropped with the beta 91.33s...

I guess the point of this is, a) the last three slider settings are essentially the same performance-wise and b) if you test at q rather than hq you will see a significant difference in fps, and imho a significant difference in IQ...

 

[dug777]

Well, tried the 91.31s, and while i only did HQ and Q, it confirmed that the 90 series ones bring a performance drop in HL2 for my 6600GT.

1280x960 with 4xAA/16AF, everything else maxed in the game video options.

HQ Q

49.29 55.27

 

[ronnn]

Originally posted by: tuteja1986
Most Website benchmark Nvidia's GPU on Image Settings: Quality which is pretty stupid :*( as they should bechmark on HQ but wait nvidia takes a massive hit so it wouldn't be good of nvidia .

Anyways XBit lab is one of the biggest offender and they now really cater for the noobs. Only the good i can think that the website do now days is alot of benchmarking of alot games but useless for high end gpu with setting they use.

So it seems the penalty for hq may be higher with the more budget cards? Yikes this whole reviewing cards just keeps getting more complicated and nothing seems a clear winner anymore.

 

[dug777]

Originally posted by: ronnn

Originally posted by: tuteja1986
Most Website benchmark Nvidia's GPU on Image Settings: Quality which is pretty stupid :*( as they should bechmark on HQ but wait nvidia takes a massive hit so it wouldn't be good of nvidia .

Anyways XBit lab is one of the biggest offender and they now really cater for the noobs. Only the good i can think that the website do now days is alot of benchmarking of alot games but useless for high end gpu with setting they use.

So it seems the penalty for hq may be higher with the more budget cards? Yikes this whole reviewing cards just keeps getting more complicated and nothing seems a clear winner anymore.

Not quite sure where you got that from, since only one budget card was involved in this test mate

Given the sheer unpleasantness of the texture crawl/shimmer Q and below exhibit in action, i can't see why anyone would ever bother using anything other than HQ...of course that's just my personal opinion

 

[tuteja1986]

try the latest review by xbit lab that used Quality instead of High Quality and then all the Nvidiot ran wild with such happy face of false victory.

 

[dug777]

Originally posted by: tuteja1986
try the latest review by xbit lab that used Quality instead of High Quality and then all the Nvidiot ran wild with such happy face of false victory.

BWAHAHAHAHAHA :laugh: