NamelessPlayer wrote:

I'm becoming increasingly convinced that SOMETHING is seriously holding back my P4 build's performance compared to you guys. I just don't know what it is right now.

Maybe it would help if I had another S478 board with 800 MHz FSB support to play with, preferably one that actually gives me some overclocking options. I mean, Extreme Edition CPUs have unlocked multipliers and I can't even take advantage of that with my current industrial board.

In the meantime, I'm seriously thinking about replacing the 6800 Ultra with the FX 5950 Ultra just to see how much the performance actually drops as a result.

Yeah, I was reading the whole topic again yesterday and your results are pretty weird. You should easily have the advantage on me for example!

It seems Doom 3 does support dual cores.

Here are some results with my AMD64 FX60 and the Geforce 7900GTX

1024*768 Ultra 127.7 FPS
AMD64 FX60 2.6Ghz, A8N32-SLi Nforce4, 2GB DDR475 CL2.5-3-3-3-5, Geforce 7900GTX 92.91, onboard audio, XP-SP3

The card still scales with a faster CPU though.

1024*768 Ultra 143.4 FPS
AMD64 FX60 2.6@3.0Ghz, A8N32-SLi Nforce4, 2GB DDR500 CL2.5-3-3-3-5, Geforce 7900GTX 92.91, onboard audio, XP-SP3

[Edit]

I did also test the 7800gtx 256mb in SLI with the overclocked FX60 and I got more or less the same score as with a single 7900gtx.
But at least it diddnt lose performance when adding a second card as with the single core CPU.
The scaling at 1024*768 in Doom 3 is perhaps 5% and in 3dmark 2001 10%, these cards needs a faster CPU at this low resolution.

With the 7800GTX SLI in Doom 3 I get exactly the same FPS at 1024*768 Ultra and 1280*1024 Ultra 142.5 FPS +-0.5 FPS

I am still using the low res LCD...

[/Edit]

F2bnp wrote:

NamelessPlayer wrote:

I'm becoming increasingly convinced that SOMETHING is seriously holding back my P4 build's performance compared to you guys. I just don't know what it is right now.

Maybe it would help if I had another S478 board with 800 MHz FSB support to play with, preferably one that actually gives me some overclocking options. I mean, Extreme Edition CPUs have unlocked multipliers and I can't even take advantage of that with my current industrial board.

In the meantime, I'm seriously thinking about replacing the 6800 Ultra with the FX 5950 Ultra just to see how much the performance actually drops as a result.

Yeah, I was reading the whole topic again yesterday and your results are pretty weird. You should easily have the advantage on me for example!

I agree, something isn't right. The P4 EE Gallatin should have left my P4 Northwood so far behind it wouldn't even be funny, but it doesn't.

@NamelessPlayer: you said you're using a Socket 478 industrial board. Does it have ISA slots? If it does, then maybe the PCI to ISA Bridge used to make them work is affecting the results somehow. The only way to find out is to get a hold of another S478 board, like you said.

JayCeeBee64 wrote:

F2bnp wrote:

NamelessPlayer wrote:

I'm becoming increasingly convinced that SOMETHING is seriously holding back my P4 build's performance compared to you guys. I just don't know what it is right now.

Maybe it would help if I had another S478 board with 800 MHz FSB support to play with, preferably one that actually gives me some overclocking options. I mean, Extreme Edition CPUs have unlocked multipliers and I can't even take advantage of that with my current industrial board.

In the meantime, I'm seriously thinking about replacing the 6800 Ultra with the FX 5950 Ultra just to see how much the performance actually drops as a result.

Yeah, I was reading the whole topic again yesterday and your results are pretty weird. You should easily have the advantage on me for example!

I agree, something isn't right. The P4 EE Gallatin should have left my P4 Northwood so far behind it wouldn't even be funny, but it doesn't.

@NamelessPlayer: you said you're using a Socket 478 industrial board. Does it have ISA slots? If it does, then maybe the PCI to ISA Bridge used to make them work is affecting the results somehow. The only way to find out is to get a hold of another S478 board, like you said.

It's a BCM BC875PLG, purchased specifically because it had one ISA slot with DMA support, currently occupied by an AWE64 Gold. (Would be an AWE32 instead if it fit in the case.)

I don't actually use it for anything but the AWE64's gameport when booted into XP, though. Audio is handled strictly by the X-Fi Prelude under that environment. It's more about being able to have one computer that can handle native DOS, 98SE and XP all at once.

Unfortunately, the BIOS options are limited. The manual showed multiplier and FSB options, but they're clearly missing and I don't know of any way to enable them; peeking in with an AwardBIOS tool to see if they're just hidden and waiting to be exposed yields nothing of the sort. There are no voltage options, either. All I can do is adjust the RAM timings, which I'll probably do since the SPD seems pretty relaxed for this set.

I happen to own a Gallatin as well.
I will put a system together with a i875P or i865PE motherboard too see if I can get some decent numbers.

I did also notice that Steves Prescott on a Nforce4 platform performed much better clock for clock than my Prescott on i925X.
I am not sure I own a s775 Nforce4 board but I do have a first generation socket 775 Radeon express board I want to try out.
It is a DDR1 board so I doubt it will perform better than a s775 Nforce4 DDR2 board.

Was reading this today:
http://www.pcgamer.com/revisiting-crysiss-ama … s-lost-its-way/

Think I know what the next thread should be about....

Hmmm, let me try a few things on this Gallatin rig...

-tightens RAM timings to 2.5-3-3-6
-re-enables Hyper-Threading

Well, those things didn't help all that much.

-DOWNGRADES drivers from 91.31 to 77.77

72.4 FPS on 800x600 High.
72.0 FPS on 1024x768 High.
64.2 FPS on 1600x1200 High.

You've gotta be kidding me. All I did was switch to the driver version F2bnp was using to get rid of one more variable, and now I'm already getting performance more in line with his system.

Looks like some graphics driver experimentation may very well be warranted here; whatever the case, 91.31's out for this system. Especially since it kept nagging me that SLI couldn't be enabled while somehow overlooking that it's an AGP card.

I'll post a full set of updated results shortly in light of this...

...no, wait, I SPOKE TOO SOON. Rebooted into XP and all of a sudden, the performance is back down to where it used to be. What gives? I don't know what's going on with this thing, but I'm reasonably sure it's at the software level now...

Wait, I think I figured it out. The framerate goes back up when I jack up my monitor's refresh rate (which was 60 Hz most of the time for some reason, except for when I was deliberately testing out my i-glasses HMD that requires a 640x480 60 Hz signal to sync alongside the usual lower DOS-era resolutions), meaning there's some manner of Vsync involved even after I specifically force Vsync off in the NVIDIA control panel.

Oh, and don't bother using the 3D stereo driver in Doom 3 unless you like slideshow framerates. Seriously, performance tanks a lot more than expected with stereo mode on.

Athlon XP 2400+ (2GHz, 266MHz FSB), 2GB DDR333 CL2.5, nForce2, 6800GT AGP (182.06), onboard audio, XP Pro SP3

640x480 Medium: 49.0 fps
1024x768 Ultra: 48.6 fps
1280x1024 Ultra: 48.2 fps

Only 7 frames faster than the overclocked Tualatin. I was expecting a little more out of this one. Still, it goes to show how much of a leap forward AMD made going from K7 to K8.

Now, I know I have a 2800+ Barton somewhere.

I got my socket 478 system running.
This is 100% CPU limited, the difference between 6800GT clocks and 6800Ultra clocks is 0.8 FPS. I could not get Intel "Performance Acceleration Technology" running. It worked last time I used this board with the same CPU and memory but this time it results in instability.

1024*768 Ultra: 89.2 FPS
P4 EE 3.4, Asus P4C800 Deluxe i875P, 2x512MB DDR400 CL2-2-2-5, Geforce 6800Ultra AGP 92.91, onboard audio, XP-SP3.

Yours has a 200 MHz clock speed advantage (again, just because my board doesn't let me overclock), but I somehow don't think that's enough to result in a whopping 20 FPS advantage at 1024x768, especially on Ultra as opposed to High.

What's going on here?

NamelessPlayer wrote:

Yours has a 200 MHz clock speed advantage (again, just because my board doesn't let me overclock), but I somehow don't think that's enough to result in a whopping 20 FPS advantage at 1024x768, especially on Ultra as opposed to High.

What's going on here?

The memory timings alone could be ~5 FPS and 200 MHz core speed another ~5 FPS.
The rest of the difference is harder to explain as you use a simular driver as the one I use.

But look at Steves results, he beats us all when it comes to peformance with certain hardware so Im sure there are more factors to consider such as, how updated your Doom 3 install is, how bloated your XP install is and what motherboard chipset you use.

I was a bit curious to know how much impact the Gallatins L3 cache has on the performance in Doom 3 so I installed a vanilla Northwood 3.4 and ran the timedemo with the exact same settings.
The answer is: the L3 boosts the performance with ~10%.

1024*768 Ultra: 79.8 FPS
P4 3.4 Northwood, Asus P4C800 Deluxe i875P, 2x512MB DDR400 CL2-2-2-5, Geforce 6800Ultra AGP 92.91, onboard audio, XP-SP3.

Skyscraper, or anyone else.

Is there a command line parameter or batch file that allows me to:

- Set a resolution
- Run the benchmark

So that I have several shortcuts for each resolution to allow quick benching. Working on something a little bit more extensive to do with the Doom 3 benchmark 😀

I have no idea but I do also want to know.

Im bumping this thread with an "anything goes" entry.

I think Windows XP 32bit is a better OS choice for a Netburst system with 2GB memory, well at least for benching.

128.1 FPS 1024*768 Ultra
Pentium D 950 @ 4533 MHz, Gigabyte X48 DS4, 2GB 1066 MHz DDR2 CL4-4-4-15, Geforce GTX285 335.23, onboard audio, Win-7 64.

If you are still working on this you should be able to make a myresolution.cfg file in the base folder with the following:

1r_mode #
2vid_restart
3timedemo demo1 1

with valid values for # being 3 = 640x480, 4 = 800x600, 5 = 1024x768, 6 = 1152x864, 7 = 1280x1024, 8 = 1600x1200, and -1 for use with the r_customwidth and r_customheight commands.
You can then make a shortcut with +exec myresolution.cfg added to the end of the target.

TeddyTheBear wrote:

If you are still working on this you should be able to make a myresolution.cfg file in the base folder with the following: […]
Show full quote

If you are still working on this you should be able to make a myresolution.cfg file in the base folder with the following:

Copy code to clipboard

1r_mode #
2vid_restart
3timedemo demo1 1

with valid values for # being 3 = 640x480, 4 = 800x600, 5 = 1024x768, 6 = 1152x864, 7 = 1280x1024, 8 = 1600x1200, and -1 for use with the r_customwidth and r_customheight commands.
You can then make a shortcut with +exec myresolution.cfg added to the end of the target.

Thanks for the information.

I changed some chipset settings and memory sub timings and suddenly Doom 3 gained 10 FPS with the Pentium-D system. 138 FPS 😀

I'm working on a test of retro CPUs - there are many benchmarks, Doom3 included. 😀

Since this is a CPU benchmark, I'm using the fastest VGA possible to avoid limitation by the graphics subsystem. So it doesn't exactly fit the time frame. Still - GF 6800 Ultra (CPU unlimited) scores about 120fps... so just pretend there is 6800 Ultra used, if you want. 😎

Doom3 Ultra details, 1024x768 noAA, noAF, timedemo demo1
GeForce 7800 GS AGP (G71 16pp) @ 500/1400 MHz, drivers 275.33 WHQL, Win XP SP3
Asus P4P800 DLX + 1GB DDR400 CL2.5-3-3-5 (PAT enabled), dual-channel
Abit NF7 + 1GB DDR266 CL2-2-2-5 / DDR333 CL2-2-2-5, dual-channel
Asus AM1I-A + 2GB DDR3 1600 CL9-9-9-24

Sempron 2650 included in the graph is just a little bonus - to find out if the current lowend is really so slow and useless by todays standards. Well, apparently it is.
All processors are running at default speed, no overclocking.

Pentium 4 3.0 (HT) - Northwood, FSB800
Athlon XP 2600+ - Tbred B, FSB333
Athlon XP 2800+ - Barton, FSB333

havli wrote:

I'm working on a test of retro CPUs - there are many benchmarks, Doom3 included. :) […]
Show full quote

I'm working on a test of retro CPUs - there are many benchmarks, Doom3 included. 😀

Since this is a CPU benchmark, I'm using the fastest VGA possible to avoid limitation by the graphics subsystem. So it doesn't exactly fit the time frame. Still - GF 6800 Ultra (CPU unlimited) scores about 120fps... so just pretend there is 6800 Ultra used, if you want. 😎

Doom3 Ultra details, 1024x768 noAA, noAF, timedemo demo1
GeForce 7800 GS AGP (G71 16pp) @ 500/1400 MHz, drivers 275.33 WHQL, Win XP SP3
Asus P4P800 DLX + 1GB DDR400 CL2.5-3-3-5 (PAT enabled), dual-channel
Abit NF7 + 1GB DDR266 CL2-2-2-5 / DDR333 CL2-2-2-5, dual-channel
Asus AM1I-A + 2GB DDR3 1600 CL9-9-9-24

Sempron 2650 included in the graph is just a little bonus - to find out if the current lowend is really so slow and useless by todays standards. Well, apparently it is.
All processors are running at default speed, no overclocking.

Pentium 4 3.0 (HT) - Northwood, FSB800
Athlon XP 2600+ - Tbred B, FSB333
Athlon XP 2800+ - Barton, FSB333

I updated the first post. I did put you in the almost period correct list, its less crowded there anyhow.
I did some guessing when it comes to what core revisons your Duron 1400, XP 2000+ and XP 2200+ could be. Shout if I guessed wrong.

I played around some more with the Pentium D system.
I think it will be hard to get much more performance out of this system using air cooling.

1024*768 Ultra.