Parhelia Fill rate or CPU limited??

[chhfchhf]

sir, i did a examination a moument ago
when core/ram at 200/500 ,3dmark2001pro point is 6704
when core/ram at 166/415 ,3dmark2001pro point is 59xx
when core/ram at 207/550 ,3dmark2001pro point is 6795
i find that the score is not altered distinctly when core/ram changed.

the Band width of parhelia is 512*200=10GB/s
the band width of ram is 256*500=16GB/s

i consider the GPU is not full load in the benchmark!

[Wombat]

Originally posted by Roark
ie, FSB and memory speed... this will make a huge difference with 3d Mark, as will processor speed... this benchmark in particular has been shown to be horribly processor/memory bound...

Not that we know if it's "bound" or "construed." 3DMark is a farce people, and always has been. The score is governed by arbitrary rules.
Search the forums and you'll find what happened when we took SoftFSB and used it on 3DMark. The score <B>skyrocketed</B>, with the same system booted at 100, Softed to 133 would massively outscore the same box when it was booted at 133 by the BIOS.

[Greebe]

You got that straight Rob

and so goes the history of 3Dfarq...

[DegreeC]

To WyWyWyWy:
600 points is a nice jump for a 300 mhz increase in CPU speed and is on the order of what I was expecting when I bumped my CPU speed up by 227 Mhz. Thats why my lack of any performance increase is so strange and why I think I have run into some bottleneck in my CPU/motherboard. What did you mean by "the main memory bandwidth is increased too"? Did you change your memory speed? Or did you mean that you upped the FSB from 400 to 533?

To chhfchhf:
I'm not sure I agree that your numbers indicate the Parhelia GPU is not fill-rate limited. Assuming that you tested with FSAA and Ansio off so the memory bandwidth is not a factor, your numbers do scale with the GPU speed (166mhz-59XX, 200mhz-6704, 207mhz-6795)which is an indication of a fill rate limitation.

To Roark:
I agree that upping my FSB will not contribute to a valid test of CPU scaling. What I'm really trying to acheive is more game performance and and to understand why my CPU upgrade hasn't given me what I expected. Upping the FSB(and memory speed/timing) will help me figure out where the bottleneck in my system is.

I'm now convinced that I need to ditch 3Dmark 2001 as a meaningful test and concentrate on games like UT2003, Quake3, and Commanche4. I plan on starting at lower resolutions without FSAA and ansio to limit any memory bandwidth issues.

[WyWyWyWy]

My main memory is set to 3/4 ratio to FSB.

Actually do you think, it is not fill rate limited but rather "processing power" limited? As there are lots of calculations involved with antialiasing.

Also, have you tried 4xFAA instead of 16xFSAA?

[Greebe]

Also, have you tried 4xFAA instead of 16xFSAA?

I'd like to see anyone with a P try those...

[WyWyWyWy]

In fact I often use 4xFAA when actually playing games, because I found 16xFSAA gives better quality... well at least in LiveForSpeed.

[Pace]

Why do you use 4x when 16x gives better quality?

Also, FAA/FSAA mixup there?

[Indiana]

I beleive you're confusing FAA and FSAA here....
The Parhelia can't do 4x FAA (unfortunately, would be interesting performancewise), but only 16x FAA. It can, however, do standard supersampling 4x FSAA as well, but that's way too slow for real use, I guess a multisampling implementation together with the FAA would've been better.

[DegreeC]

Yeah, the mix up is my fault for using 16X FSAA(which is impossible) instead of 16X FAA in my posts. I never considered trying 4X(or 2X FSAA) because I didn't know it was possible. Can it be enabled in Powerdesk or must it be selected in the game? WyWyWyWy are you saying that you are enabling 4X FSAA in the game and that it runs/works better than than 16X FAA?

[3dfx]

16 FAA is nice but take it from me it has to many faults imo and it doesn't alias alot of places it should so some areas on screen remain really jaggy, dont get me wrong when it does it right it does it well. Though I've compared ATi's 6x to Matrox's FAA 16x and the ATI's is of a similiar quality it also does the full monty! ..hehe ie everything! and it tends to work all the time with similiar quaility and performance hit One of the most annoying things about Matrox's FAA is that is doesn't alias translucent edges ARRGGH ..lol Water etc looks so jaggy maybe I was a little to harsh as it is a new tech and with some work it could be the best

[Roark]

Wombat, good point with wording choices... as you said, it isn't cpu/memory bound... more like the amount of memory bandwidth and speed of processor have a tremendous amount of weight on the benchmark, often times giving huge increases in areas when it shouldn't.

DegreeC, to enable 4XFSAA you do it in the Powerdesk. Instead of forcing 16X FAA you just force 4X FSAA.

Unfortunately most of the current generation games are going to be inherantly CPU bound. Unreal 2/Unreal Tournament 2003 are both tremendously CPU bound, regardless of the video card. Doom 3 will be CPU bound as well. As are most of the other games that have been released within the last 6 months/year/year and a half.

[chhfchhf]

4x FSAA is based at multysample
16x FAA is based at supersample
multysample brings lots of data to excute,so need more band width.
supersample needs enhanced fillrate.
in quality,4x FSAA will do good performance.
when use 16x FAA , it is little different with 4x FSAA
but ,when use 4x FSAA,it is slowly than 16x FAA

[DegreeC]

Roark,
How does your comment about current and future games being largely CPU bound apply to the Parhelia? Are you saying the nature of these games doesn't expose any fill-rate limitations in the Parhelia? Because of the low clock speed of this GPU I would expect that at some level of CPU power the GPU would be saturated and further increases in CPU power wouldn't lead to any performance increases.

This really is the core of my quest; to determine if there is a fill rate limitation and if so where (in general) it is. I absolutely love Surround gaming; it is a spectactular leap forward in the immersive effect of PC games. However, my current configuration is a little short of giving me the kind of performance I want. I'm willing to throw more money at this but I need to decide what kind of CPU/motherboard I need without resorting to overkill. Is and XP2800/nForce2 or P4 2.8/E7805 too much?

____________________
Athlon XP2400
MSI KT3 Ultra with on-board sound
Retail Matrox Parhelia 128 MB
512 MB CorsairXMS PC2700
Maxtor D740X 60 GB
WinXP Pro SP1

[Greebe]

Matrox has yet to see a cpu capable of fully saturating Parhelia.