Doom 3 timedemo shootout with period correct hardware.

Discussion about old graphics cards, monitors and video related things.

Re: Doom 3 timedemo shootout with period correct hardware.

Postby Skyscraper » 2019-4-06 @ 17:31

This turned out to be a bit boring! :D

No need for Geforce 7900 GTX SLI or changing CPU to the X2 6400+

I still have the A64 X2 6000+ Asus M2N-SLI AM2 system I used for benchmarking for the list many years ago (with the slower v1.3 Doom 3 version) intact as I have used it for some other projects. I only reinstalled Doom 3, v1.0 this time, added an X-Fi XtremeMusic sound card and changed out the memory kit for a better one. This system runs Windows XP Pro SP3 with most of the bloat still active as it's a system I used for some Internet gaming such as WoW "vanilla" a few years ago.





Athlon 64 X2 6000+ (3.0), Asus M2N-SLI Deluxe nForce 570, 2Gb DDR2 750 3-3-3-8, Geforce 7900GTX, 92.91, X-fi XtremeMusic, XP SP3

Doom 3 v1.0 1024 Ultra: 190.1 FPS

A64_6000plus_M2NSLIDeluxe_2GB750mhz3339_7900gtx_Doom3_timedemo.JPG








Athlon 64 X2 6000+ @3.5 14x250, Asus M2N-SLI Deluxe nForce570, 2x1GB DDR2 1000 4-4-4-12, 7900GTX 720/1800, 92.91, X-Fi, XP SP3

Doom 3 v1.0 1024 Ultra: 221.4 FPS

A64_6000plus_3500_250_M2NSLIDeluxe_2GB1000mhz44412_7900gtx_720_1800_Doom3_timedemo.JPG




The voltage in CPU-Z is shown as 0.2V lower than it really is, both @stock and overclocked.

DDR2 750 MHz is what you actually get when you choose DDR2 800 MHz using the 15x multiplier with Athlon 64 AM2 CPUs running at stock 200 MHz reference clock. This is because the wanted memory multiplier is 2x as in (2x200)x2 (for DDR) = 800 but as the memory multiplier is tied to the CPU multiplier the nearest we can get is 15/8 = 1.875x and the memory speed comes from ((15/8)x200)x2 = 750. This can also be presented as CPU/8 (x2 for DDR) or (3000/8)x2 = 750.

At even multipliers the 800 MHz memory setting works as intended as all even numbers can be divided by 2 without the result beeing a fraction. For example 16/2 = 8 and to get 800 Mhz memory speed ((16/8)x200)x2 = 800 or (3200/8)x2 = 800.

If you think of the memory clock as the CPU speed divided by the most suitable integer or the reference clock multiplied by the CPU multiplier divided by the most suitable integer dosn't really matter as it's exactly the same thing.
Main PC: Dual Xeon X5690@4.6ghz, Evga - SR-2, 48gb memory, Intel X25-M g2 SSD and a Nvidia GTX 980 ti.
Retro PC #3: K6-2 450@500mhz, PC-Chips m577, 256mb sdram, AWE64 and a Voodoo Banshee.
Skyscraper
l33t
 
Posts: 3400
Joined: 2013-9-18 @ 15:36
Location: Sweden

Re: Doom 3 timedemo shootout with period correct hardware.

Postby Standard Def Steve » 2019-4-07 @ 01:51

Now that I'm running this beast of a PIII machine at 1628MHz, I thought I'd rebench my 9800 Pro. :)
Don't bother adding this result to the list, as I used the High preset. The 9800Pro only has 128MB of VRAM, and indeed there was a bit of stuttering on Ultra.

PIII-S @ 1628MHz (155 FSB), 2GB DDR @ 310MHz 2-2-2-5, QDI Advance 12T (Apollo Pro 266T), Radeon 9800 Pro (Cat 4.12), X-Fi Platinum, XP MCE 2005 SP3
1024x768 High:
PIII-1628, 9800Pro-D3High.png


Looking at the Toplist, this machine with a mere 9800 Pro actually managed to beat my old TUV4X/6800GT score! That machine only got 41.5 fps, which speaks volumes about how much slower the 694T chipset is!
Standard Def Steve
Oldbie
 
Posts: 974
Joined: 2012-9-15 @ 08:04

Re: Doom 3 timedemo shootout with period correct hardware.

Postby SPBHM » 2019-4-12 @ 16:52

Standard Def Steve wrote:Now that I'm running this beast of a PIII machine at 1628MHz, I thought I'd rebench my 9800 Pro. :)
Don't bother adding this result to the list, as I used the High preset. The 9800Pro only has 128MB of VRAM, and indeed there was a bit of stuttering on Ultra.

PIII-S @ 1628MHz (155 FSB), 2GB DDR @ 310MHz 2-2-2-5, QDI Advance 12T (Apollo Pro 266T), Radeon 9800 Pro (Cat 4.12), X-Fi Platinum, XP MCE 2005 SP3
1024x768 High:
PIII-1628, 9800Pro-D3High.png


Looking at the Toplist, this machine with a mere 9800 Pro actually managed to beat my old TUV4X/6800GT score! That machine only got 41.5 fps, which speaks volumes about how much slower the 694T chipset is!


I don't notice any stuttering with the FX5900SE 128mb (I'm using the pre caching command 1), 1GB of ram and 128mb AGP aperture size on ultra,

but when I tried the same PC a while back with 768MB of ram and a 9500PRO 128MB it had bad stutters, but increasing the aperture size to 256MB seemed to fix it, well at least on the second run.

in any case, it definitely feels "smooth" with the 5900se 128mb on ultra (in terms of not having freezes, but the average framerate is not that great anyway)
SPBHM
Oldbie
 
Posts: 563
Joined: 2012-10-26 @ 15:59

Re: Doom 3 timedemo shootout with period correct hardware.

Postby Standard Def Steve » 2019-8-08 @ 05:06

Here's your first PowerPC Mac result.

PowerMac G4 MDD, dual G4 1.25GHz (166MHz FSB, 256K on-die L2 cache, 2MB external L3 cache per CPU), 2GB PC2700 DDR, GeForce 7800GS AGP, OS X 10.4.11
1024x768 Ultra:
G4 1.25 Doom 3 7800GS.png


Yeah, that's pretty bad. I don't know if it's a sloppy port or if the G4 just plain sucks at gaming. Of course, there could be something else at play here. Perhaps the Motorola chipset has poor memory or AGP performance. Kinda hard to measure these things on a Mac. Still, my 1.63GHz PIII-S managed over twice the frame rate with a 6800GT. Remember this?
PIII-1628-D3-MCE.png


EDIT: I just ran Quake 3 on the Mac, and it actually outperformed the PIII-S! I'm just gonna conclude that Doom 3 for OS X is a lazy, unoptimized port. :dead:
EDIT 2: But Unreal Tournament 2004 is another game that runs faster on the PIII-S box, so who knows? Could it be that both Doom3 and UT2K4 are crappy ports, or does the G4 simply fall flat in certain games/workloads?
EDIT 3: I was running the timedemo again today and noticed that the OS X version of Doom 3 is not multi-threaded. If you run the timedemo on a dual-core Windows system, you get near 100% CPU usage. However, on the Mac I was only getting 50% CPU usage, so that 2nd G4 was just sitting there doing nothing. If it was multithreaded, this system probably would've been able to hit 40-45 fps. Which is still slower than the single overclocked PIII-S, but at least it would be somewhat playable.
Standard Def Steve
Oldbie
 
Posts: 974
Joined: 2012-9-15 @ 08:04

Re: Doom 3 timedemo shootout with period correct hardware.

Postby dottoss » 2019-8-12 @ 17:21

Standard Def Steve wrote:EDIT 3: I was running the timedemo again today and noticed that the OS X version of Doom 3 is not multi-threaded. If you run the timedemo on a dual-core Windows system, you get near 100% CPU usage. However, on the Mac I was only getting 50% CPU usage, so that 2nd G4 was just sitting there doing nothing. If it was multithreaded, this system probably would've been able to hit 40-45 fps. Which is still slower than the single overclocked PIII-S, but at least it would be somewhat playable.


Actually I think that Doom 3 never was multithreaded, at least not in any advanced way. Quake 4 though is SMP out of the box, which is a tad newer engine of Doom 3. Doom 3 BFG is SMP though.

EDIT: Source - https://fabiensanglard.net/doom3_bfg/threading.php
User avatar
dottoss
Member
 
Posts: 104
Joined: 2014-7-08 @ 11:30

Previous

Return to Video

Who is online

Users browsing this forum: lost77 and 2 guests