SPBHM wrote:

is it possible to run doom 3 without SSE?
I only have 1.31 and it wont work on my Pentium II

Think the lowest anyone has ever run it on was a P3.

Leaving everything else alone I increased ram on my test system from 2gb to 4 gb: [2x] 1gb PQI PC6400/4-3-4-14 vs [2x] 2gb PQI PC6400/5-5-512. Given how D3 devours ram I expected that doubling the ram would boot the Timedemo at least a bit, but frames remained virtually unchanged: 171 to 176fps depending on Timedemo's mood.

System is same as the one I've been fiddling with the last few months:
mATX Biostar P4M890M-TE, C2D E6700/2.67ghz, 7900GTX/512mb, XPpro SP3, D3 v1

Is this a sign that the cpu or gpu are bottlenecking performance? Or could looser ram timings completely offset the benefit of the extra 2gb ram?

tincup wrote:

Leaving everything else alone I increased ram on my test system from 2gb to 4 gb: [2x] 1gb PQI PC6400/4-3-4-14 vs [2x] 2gb PQI P […]
Show full quote

Leaving everything else alone I increased ram on my test system from 2gb to 4 gb: [2x] 1gb PQI PC6400/4-3-4-14 vs [2x] 2gb PQI PC6400/5-5-512. Given how D3 devours ram I expected that doubling the ram would boot the Timedemo at least a bit, but frames remained virtually unchanged: 171 to 176fps depending on Timedemo's mood.

System is same as the one I've been fiddling with the last few months:
mATX Biostar P4M890M-TE, C2D E6700/2.67ghz, 7900GTX/512mb, XPpro SP3, D3 v1

Is this a sign that the cpu or gpu are bottlenecking performance? Or could looser ram timings completely offset the benefit of the extra 2gb ram?

In the list your result with that board, CPU and video card is 147.9 FPS, with 2GB much slower memory.

You have a better result in the same ballpark as your new score with the very similar ECS P4M890T-M V2 with the same CPU and video card, 168.6 FPS.

Perhaps you (or I) are mixing up the boards as I would say going from 147.9 FPS to 176 with only more and faster memory is very good 😀.

I will update the result in the list after you confirmed which of the two boards you used for the new better score.

It would also be nice with a more exact value than 171 to 176, just post the best score if you have ran the benchmark more than one time with the exact same setup. Normally you get the best score the first time you run the timedemo after restarting Windows but with some systems its the second time you run it that gives the best score.

Running with 2GB or 4GB should not affect the performance in Doom 3 at all. I have not seen any significant difference between 1GB and 2GB either when running a slimmed XP-SP3 install.

I should have been clearer - and more up to date. After getting a much better score on the ECS board [147 vs 168; same 2gb ram/cpu/gpu], what I did was roll back to the BioStar again but this time use the VIA driver set from the ECS. With the ECS chipset drivers the Biostar then performed as well as the ECS. VIA drivers made a big difference here.

Next I upgraded the ram to higher quality sticks; PQI 4-4-4-8, which ended up running @ 4-3-4-12/267. This increased the score to 176 or so. I may not have entered that score in the data base.

Finally I doubled the ram as described in my previous post; 4gb PQI 5-5-5-12/267. I expected a bump but the numbers were almost identical at 4gb as they had been at 2gb. Ram timings were a bit different between the 2gb and 4gb setup, but other than that it was the same box. Made me curious.

Regardless - I'm very happy with the rig and it is presently in the 'permanent legacy rig' collection - dual XP/W98se....

tincup wrote:

I should have been clearer - and more up to date. After getting a much better score on the ECS board [147 vs 168; same 2gb ram/c […]
Show full quote

I should have been clearer - and more up to date. After getting a much better score on the ECS board [147 vs 168; same 2gb ram/cpu/gpu], what I did was roll back to the BioStar again but this time use the VIA driver set from the ECS. With the different chipset drivers the Biostar then performed as well as the ECS. VIA drivers made a big difference here.

Next I upgraded the ram to higher quality sticks; PQI 4-4-4-8, which ended up running @ 4-3-4-12/267. This increased the score to 176 or so. I may not have entered that score in the data base.

Finally I doubled the ram as described in my previous post; 4gb PQI 5-5-5-12/267. I expected a bump but the numbers were almost identical at 4gb as they had been at 2gb. Ram timings were a bit different between the 2gb and 4gb setup, but other than that it was the same box. Made me curious.

Regardless - I'm very happy with the rig and it is presently in the 'permanent legacy rig' collection - dual XP/W98se....

OK that makes things more clear.

I updated the list with the Biostar systems better score.

Cool.

Anything goes (i guess)

Athlon 64 3200+ sk754 (NewCastle)
1GB of ram DDR1 (2x512MB)
500GB IDE hard drive
ATI Radeon HD 3850 AGP
Windows XP SP3
Doom3 v1.0 Ultra Settings at 1024x768 and sound enabled:

Not an awesome performer by any means when compared with some machines here and considering the beefy agp video card (the best ever?), so probably it's the socket 754 NewCastle CPU or most likely the Asus motherboard with a SIS chipset that is holding it back.
I get 64,8 fps at 1280x1024.

rick6 wrote:

Anything goes (i guess) […]
Show full quote

Anything goes (i guess)

Athlon 64 3200+ sk754 (NewCastle)
1GB of ram DDR1 (2x512MB)
500GB IDE hard drive
ATI Radeon HD 3850 AGP
Windows XP SP3
Doom3 v1.0 Ultra Settings at 1024x768 and sound enabled:

Not an awesome performer by any means when compared with some machines here and considering the beefy agp video card (the best ever?), so probably it's the socket 754 NewCastle CPU or most likely the Asus motherboard with a SIS chipset that is holding it back.
I get 64,8 fps at 1280x1024.

If you are using the 1.3 version of Doom 3 the score is normal. If I remember right the SIS A64 chipsets wasnt much worse than the other K8 chipsets. SIS 755 had gimped HyperTransport bus though but I do not think it matters much when it comes to Doom 3.

rick6 wrote:

Anything goes (i guess) […]
Show full quote

Anything goes (i guess)

Athlon 64 3200+ sk754 (NewCastle)
1GB of ram DDR1 (2x512MB)
500GB IDE hard drive
ATI Radeon HD 3850 AGP
Windows XP SP3
Doom3 v1.0 Ultra Settings at 1024x768 and sound enabled:

Not an awesome performer by any means when compared with some machines here and considering the beefy agp video card (the best ever?), so probably it's the socket 754 NewCastle CPU or most likely the Asus motherboard with a SIS chipset that is holding it back.
I get 64,8 fps at 1280x1024.

Pretty sure it's the SiS chip. I have a similar VIA system with a 6800GT and a 3200+ which scores quite a bit lower than a Pentium 3.2 with an Intel chipset. Totally sucks.

It's the SIS chip for sure. With the patch v1.3.1 i get about 56.5 fps.

By the way, i assembled today one Celeron rig, a Intel G1840 coupled with a ECO motherboard from MSI (some economic office board green type) and one stick of 4GB of DDR3. So this Celeron with it's intel GPU, Windows 7 and with the patch v1.3.1 scores 56.8 fps at Ultra settings! The exact same as the Athlon 64 3200+ with the ATI HD 3850 agp 😀
Just for reference i installed a ATI HD 4670 PCI-E on this Celeron which is a good match for the ATI HD 3850 AGP and it scored 156,8 fps, exactly 100 fps more than with only the GPU inside the CPU.
This is one fast celly! Most likely i'm handicaping the performance by only using one stick of ram instead of two ( no dual channel).

Yeah, SiS chipsets aren't as fast as nVidia and Intel. Also not helping is the fact that AMD's video drivers just aren't as lean as nVidia's. On my PIII-S system, my 6800GT almost always comes out ahead of my x1950Pro AGP. At 640x480, the difference in driver efficiency really shows. Even the newest driver available for GeForce 6 & 7--307.83--is surprisingly easy on old processors.

rick6 wrote:

Just for reference i installed a ATI HD 4670 PCI-E on this Celeron which is a good match for the ATI HD 3850 AGP and it scored 156,8 fps, exactly 100 fps more than with only the GPU inside the CPU.
This is one fast celly!

No surprise there, it's a Haswell. 😀

Intel chipsets for K8 would have been nice 😀

I will some time in the distant future bench a Fujitsu-Siemens D1607 VIA K8T800 Socket-754 motherboard with an Athlon 64 3200+. The board is a first generation K8 board from a Scaleo tower, it will be interesting to see how it fares against the SIS chipset board, I even own a HD 3850 AGP 😀

Standard Def Steve wrote:

No surprise there, it's a Haswell. 😀

True, true.
Still, doesn't this makes it the fastest "Celeron" here running Doom3? I know it's unfair though 😁

By the way, here is a Compaq Deskpro EN from the year 2000 very upgraded with a GeForce 6200 512mb of VRAM (PCI version, no AGP on this baby), 512MB of ram (max the chipset allows), a 1Ghz Coppermine Pentium III, a 160GB SATA hard drive through a SATA\PCI controller card and a USB2.0/PCI controller card.

Seems that Coppermine CPUs aren't very good with Doom3 aren't they? Yes i know that the PCI bus doesn't help either but still...

Time to throw another one into the "anything goes" list:

Fujitsu T902
Core i5-3320M
16 GB (2x8 GB dual-channel) DDR3-1600 CL9
Intel HD 4000
Windows 10
Doom 3 v1.3.1 Steam

800x600 Ultra: 88.2 FPS
1024x768 Ultra: 83.5 FPS
1152x864 Ultra: 79.7 FPS
(haven't bothered to test higher resolutions since I'd need an external monitor and the internal display's 1600x900)

I'm not sure how much of this is due to me repasting the CPU for the hell of it to reduce thermal throttling (one core still runs 5-6 degrees hotter on full load) and how much is due to me wiping Windows 8 to pave the way for a clean Windows 10 upgrade, but I notice GTA:SA and Doom 3 performing a lot better on this machine now.

I suppose Intel graphics are at least competent enough for decade-old gaming once you go past Ivy Bridge now, even if the latest driver updates with DX12 support only apply to Haswell onward.

NamelessPlayer wrote:

I suppose Intel graphics are at least competent enough for decade-old gaming once you go past Ivy Bridge now, even if the latest driver updates with DX12 support only apply to Haswell onward.

They might be competent enough for decade-old gaming but even so, they're still nowhere near the raw power of video cards from 2004 because they still have a cpu backing it up that is several times more powerful than what cpus were in 2004.

It would be really interesting if someone here would try a 6800 GT\Ultra PCI-E on their i3\i5\i7 cpus! I'm curious how that would even compare against core2duo\quad cpus!

rick6 wrote:

NamelessPlayer wrote:

I suppose Intel graphics are at least competent enough for decade-old gaming once you go past Ivy Bridge now, even if the latest driver updates with DX12 support only apply to Haswell onward.

They might be competent enough for decade-old gaming but even so, they're still nowhere near the raw power of video cards from 2004 because they still have a cpu backing it up that is several times more powerful than what cpus were in 2004.

It would be really interesting if someone here would try a 6800 GT\Ultra PCI-E on their i3\i5\i7 cpus! I'm curious how that would even compare against core2duo\quad cpus!

I think the fatest overclocked period correct scores are somewhat GPU limited even with K8 CPUs but the X800/X850 XT(PE) and the 6800 GT/Ultra could probably reach between 150 and 200 FPS at 1024*768 Ultra if paired with for example an i7 2600k @5+ GHz running Windows XP.

I updated the list with the new scores.

rick6 wrote:

Seems that Coppermine CPUs aren't very good with Doom3 aren't they? Yes i know that the PCI bus doesn't help either but still...

I think you're actually being limited by your video card there. My 64-bit FX5200 AGP can only do 6.5 fps, and that's with a 2.64GHz Athlon64 behind it! 12.5fps for 64-bit GF6200 PCI is quite good!

rick6 wrote:

It would be really interesting if someone here would try a 6800 GT\Ultra PCI-E on their i3\i5\i7 cpus! I'm curious how that would even compare against core2duo\quad cpus!

I doubt that a GF6800 would scale much higher than ~110fps even with a Core i CPU powering it.

Here are some 6800GT results with three different CPU architectures. The video card was not overclocked.

Core 2 Quad @ 3.33GHz (PCIe, Win7 x64): 111.4 fps
Athlon 64 @ 2.8GHz (AGP, WinXP): 108.7 fps
Pentium M @ 2.66GHz (PCIe, WinXP): 104.2 fps

The Quad is a much faster CPU than the Athlon 64, but there's only a 3 fps difference between the two. Windows 7 manages VRAM a little differently than XP, so the 6800GT might have done a little better under XP, but it's still completely GPU bottlenecked. The Pentium M is able to hit 144 fps with a GTX 260, and the Core 2 Quad can pump out 265 fps with a GTX 560.

FWIW, the original 640MB GeForce 8800 GTS plateaus at 290fps with a 4.5GHz Core i7. It's completely GPU limited, because adding a 2nd 8800GTS for SLI nearly doubles the frame rate: 568 fps.

Standard Def Steve wrote:

I think you're actually being limited by your video card there. My 64-bit FX5200 AGP can only do 6.5 fps, and that's with a 2.64 […]
Show full quote

rick6 wrote:

Seems that Coppermine CPUs aren't very good with Doom3 aren't they? Yes i know that the PCI bus doesn't help either but still...

I think you're actually being limited by your video card there. My 64-bit FX5200 AGP can only do 6.5 fps, and that's with a 2.64GHz Athlon64 behind it! 12.5fps for 64-bit GF6200 PCI is quite good!

rick6 wrote:

It would be really interesting if someone here would try a 6800 GT\Ultra PCI-E on their i3\i5\i7 cpus! I'm curious how that would even compare against core2duo\quad cpus!

I doubt that a GF6800 would scale much higher than ~110fps even with a Core i CPU powering it.

Here are some 6800GT results with three different CPU architectures. The video card was not overclocked.

Core 2 Quad @ 3.33GHz (PCIe, Win7 x64): 111.4 fps
Athlon 64 @ 2.8GHz (AGP, WinXP): 108.7 fps
Pentium M @ 2.66GHz (PCIe, WinXP): 104.2 fps

The Quad is a much faster CPU than the Athlon 64, but there's only a 3 fps difference between the two. Windows 7 manages VRAM a little differently than XP, so the 6800GT might have done a little better under XP, but it's still completely GPU bottlenecked. The Pentium M is able to hit 144 fps with a GTX 260, and the Core 2 Quad can pump out 265 fps with a GTX 560.

FWIW, the original 640MB GeForce 8800 GTS plateaus at 290fps with a 4.5GHz Core i7. It's completely GPU limited, because adding a 2nd 8800GTS for SLI nearly doubles the frame rate: 568 fps.

The top overclocked period correct score is 120.6 FPS using an Athlon 64 so a bit higher than that should at least be possible 😀 It wont reach much higer without serious GPU overclocking though. If someone could get their hands on a Geforce 6800 Ultra Extreme with 450 Core and 1200 memory clock he would not have a hard time reaching the top of the period correct list.

Its nice to know that all whine about poor SLI and Crossfire scaling back in 2004 - 2006 mostly was because the CPUs couldnt keep up 😀

Skyscraper wrote:

The top overclocked period correct score is 120.6 FPS using an Athlon 64 so a bit higher than that should at least be possible 😀 It wont reach much higer without serious GPU overclocking though. If someone could get their hands on a Geforce 6800 Ultra Extreme with 450 Core and 1200 memory clock he would not have a hard time reaching the top of the period correct list.

I own one of the worst 6800GTs in the world. It just doesn't overclock worth a damn. When paired with a slower CPU like my PIII-S, it lets me overclock to around 385/540, but with an Athlon 64 driving it, even 370 produces the occasional glitch. It's basically stock or nothing with a fast CPU.

I haven't tried overclocking my PCIe 6800GT yet, but honestly I don't see the point of overclocking (or even using) any DX9 PCIe GPU. If I need more graphics horsepower, I can just plug in a substantially more powerful PCIe card. 😀

Its nice to know that all whine about poor SLI and Crossfire scaling back in 2004 - 2006 mostly was because the CPUs couldnt keep up 😀

Yup. 😀
It depends on the game though. For example, in Crysis DX10, even my Opteron 185 allows for some incredible (~90%) SLI scaling with two 8800GTS cards. But in Doom3, SLI is 2.5 fps slower.

Standard Def Steve wrote:

FWIW, the original 640MB GeForce 8800 GTS plateaus at 290fps with a 4.5GHz Core i7. It's completely GPU limited, because adding a 2nd 8800GTS for SLI nearly doubles the frame rate: 568 fps.

Standard Def Steve wrote:

It depends on the game though. For example, in Crysis DX10, even my Opteron 185 allows for some incredible (~90%) SLI scaling with two 8800GTS cards. But in Doom3, SLI is 2.5 fps slower.

I must have missed something. That's quite the opposite what you just said above in the two quotes.

Also i just notice something weird on my own printscreen:

rick6 wrote:

http://i1383.photobucket.com/albums/ah283/rick62008/1_zpsirwbh1bo.jpg […]
Show full quote

3.0 volts on the cpu core? That must be a incorrect reading!