http://www.vgamuseum.info/index.php/benchmarks/311-unreal
Here's the early Unreal 3D game, there's some PCI cards in the list. I doubt you'll find any early benchmarks on half-life though 😒
@kithylin, is this the original game from 1997? I don't have the game actually. Is the benchmark 'flyby' builtin and is it available in a free demo or something?
I'm sorry.. I don't have any further information. I just found the page referenced in another thread and thought it might help you some how.
Yes it is built in (looks like this) and there is no free demo available of it.
i got a shareware demo of that unreal game. i will give it a try later today and share the results. thanks to both kithylin & d1stortion.
Keep in mind that Unreal loves Glide and at first didn't run very well with Direct3D or OpenGL. You can see that a Voodoo3 or Voodoo2 SLI beats a Geforce2 Pro or a Radeon 9600 here.
Also keep in mind that those benchmarks are run with a faster CPU than yours. I suppose a P3 850 would max out at about 80 fps even with Glide.
If you really want to run Unreal with your current hardware you might want to try even a more modern version of Unreal or a Glide wrapper.
wrote:Keep in mind that Unreal loves Glide and at first didn't run very well with Direct3D or OpenGL. You can see that a Voodoo3 or Voodoo2 SLI beats a Geforce2 Pro or a Radeon 9600 here.
Also keep in mind that those benchmarks are run with a faster CPU than yours. I suppose a P3 850 would max out at about 80 fps even with Glide.
If you really want to run Unreal with your current hardware you might want to try even a more modern version of Unreal or a Glide wrapper.
@idspispopd, yeah i tasted that bitterness when i tried to play deus ex, an unreal engine game, with single digit fps in some scenes. honestly i'm not hoping or expecting much by the way. the shareware demo does not have the flyby timedemo. i'm now trying to patch it and will update.
Unreal can use UTGLR too. OldUnreal.com can set you up with that. Recent versions don't seem to work with 9x though. Interestingly, you can use these renderers on cards as old as Radeon 8500 and GeForce 3 and the result is very impressive. It is better than Glide I think, plus you get S3TC support.
just updated to latest patch 227i. started the game and there it goes again.
OpenGL: Lowest 17.7, HIghest 61.87 and Avg 42.06.
Direct3D9: Lowest 17.3, HIghest 62.62 and Avg 42.34.
The dip in FPS was mainly noticed at 2 places: right at the start of the flyby in front of the castle and then almost at the end of one cycle when it comes near the right side of the castle.
Software: Lowest 5.6, HIghest, 12.77 and Avg 8.34.
edit1:all settings were default(800x600, depth??) except fps limit of 60 was removed. @1280x1024 decreased the min fps by 3. alternatively disabling volumetric lighting increased the min fps by 6.
wrote:Unreal can use UTGLR too. OldUnreal.com can set you up with that. Recent versions don't seem to work with 9x though. Interestingly, you can use these renderers on cards as old as Radeon 8500 and GeForce 3 and the result is very impressive. It is better than Glide I think, plus you get S3TC support.
@swaaye, i will try that tomorrow and post the result. thanks for the info, i got the latest patch from there only. do you mean directx 9?
wrote:@swaaye, i will try that tomorrow and post the result. thanks for the info, i got the latest patch from there only. do you mean directx 9?
The Direct3D 9 renderer is UTGLR. It comes with Oldunreal's patch.
@swaaye, ok got it. as you said the new direct3d9 renderer and opengl were far better compared to the software renderer(which came with the demo) that i initially tried except for the the dip in fps. it seems to be the main issue, which if identified and resolved will result in performance as expected.
Those dips in flyby are normal, your scores are quite bad nevertheless. Here's what I get in Unreal Gold on my PIII 1100, Voodoo5, 440BX: 40.6 lowest, 170.1 highest, 102.2 average (1024x768, Glide)
And in software: 9.9 lowest, 23.9 highest, 17.9 average (same resolution)
wrote:Those dips in flyby are normal, your scores are quite bad nevertheless. Here's what I get in Unreal Gold on my PIII 1100, Voodoo5, 440BX: 40.6 lowest, 170.1 highest, 102.2 average (1024x768, Glide)
And in software: 9.9 lowest, 23.9 highest, 17.9 average (same resolution)
@d1stortion, thanks a lot for taking out time to share your benchmarks.
i'm right now summarizing the whole issue and will update shortly.
i will just summarize,
possible reason 1: Zotac GT520 PCI card's PCI to PCI-E reverse bridge's compatibility issue with VIA chipset.
observation: the obvious difference between normal PCI graphics card and this card is the reverse bridge and is natively a PCI-E card. probably the transfers in both directions between the host and the gpu through the host bridge, pci bus, reverse bridge and the gpu to memory are unable to operate at optimum level due to the mismatch of generations. i tried tweaking the pci config registers of the the host bridge as well as gpu's reverse bridge and they were not fruitful. most of the changes i made in the northbridge's configuration registers that were related to PCI bus (like Buffer/Flow/Master/Arbitration) didn't enhance the gaming performance. as there is no separate chipset driver for my mainboard(only onboard vga, ide, agp & audio are availabe), i assumed that whatever changes i made in the chipset config registers(using wpcredit) will enhance/affect the performance in real-time. similarly the manufacturer of the gpu's reverse bridge(pericom) confirmed that no separate driver is required and the default windows driver is sufficient. i don't know whether the driver overrides my manual tweaking(less likely) or the tweaking itself does much. my progress through this troubleshooting path is nil as of now.
possible reason 2: VIA chipset's PCI bursting issue.
observation: i suspected this issue when i came across tecChannel's report "http://www.tecchannel.de/ueberblick/archiv/40 … down_pci_cards/". however iirc, VIA released a patch addressing the issue and an alternate third party patch was also made available. i tried both the patches, but they didn't address my issue. it also seems that those patches were made part of the windows xp service packs' default driver set and hence could not be the reason for the issue or else i must be missing something. my progress through this path will remain stuck unless i get any inputs about similar via chipset based mainboards.
possible reason 3: IRQ sharing between Audio, USB and Graphics card.
observation: i came across this when i visited viaarena forums' thread "http://www.viaarena.com/forums/showthread.php?t=25485". the OP was also complaining about poor performance in half-life 1. probably his problem is different than mine. however i'm tempted to try out the windows reinstallation with "standard pc" instead of ACPI PC to removing the IRQ sharing as a last option (reinstalling steam games and other programs always a pain).
possible reason 4: Modern Graphics card's driver optimization for Modern hardware.
observation: probably this modern card's driver is tweaked to work better with modern hardware than P3 class CPU's without SSE2 support. I didn't test thoroughly, but the card worked OK in a P4 system. it somehow seems less likely a driver issue, though only NVidia can confirm it. i also can't tweak the pci config registers of the graphics card as there is no datasheet to play with. i'm stuck in this path too and don't foresee any progress. if this is the real issue, then i'm doomed.
possible reason 5: cpu, motherboard or memory problem
observation: the ageing factor of the different components from 2000 might be the culprit. i also recapped the motherboard which probably was not upto the mark. however i don't suspect the branded memory (Kingston 512 MB PC133 CL3) for which i paid an obscene amount to get it shipped from abroad. the counter arguement to this suspected issue is that the synthetic benchmarks of the cpu/mainboard/memory didn't throw up any poor results.
if you are still reading this, my apologies for the lengthy post. it really took a load off my shoulders.
wrote:Those dips in flyby are normal, your scores are quite bad nevertheless. Here's what I get in Unreal Gold on my PIII 1100, Voodoo5, 440BX: 40.6 lowest, 170.1 highest, 102.2 average (1024x768, Glide)
And in software: 9.9 lowest, 23.9 highest, 17.9 average (same resolution)
@d1stortion, i suppose your Voodoo5 card is AGP or is it PCI? If PCI, i can atleast rule out PCI slot bandwidth limitation.
It's AGP, but as some in this thread have assured you, the PCI bus itself is not the fundamental issue here. Remember that Voodoo2 was the hottest card when Unreal came out, and that was PCI and played the game fine. Look at the first video I posted, there is a timedemo in there and you will see that Voodoo2 SLI outperforms your GT 520.
At this point I feel the situation is somewhat deadlocked. From what I understand the PCI bridge manufacturer has basically confirmed that it should work fine in your system, so I can only see either the CPU, some VIA issue or that IRQ sharing issue as the culprit. FWIW, I believe I do use ACPI on 98SE and have no problems with it.
Your best bet for further troubleshooting is to spare a few bucks for a vintage AGP/PCI video card I'd say...
wrote:It's AGP, but as some in this thread have assured you, the PCI bus itself is not the fundamental issue here. Remember that Voodoo2 was the hottest card when Unreal came out, and that was PCI and played the game fine. Look at the first video I posted, there is a timedemo in there and you will see that Voodoo2 SLI outperforms your GT 520.
At this point I feel the situation is somewhat deadlocked. From what I understand the PCI bridge manufacturer has basically confirmed that it should work fine in your system, so I can only see either the CPU, some VIA issue or that IRQ sharing issue as the culprit. FWIW, I believe I do use ACPI on 98SE and have no problems with it.
Your best bet for further troubleshooting is to spare a few bucks for a vintage AGP/PCI video card I'd say...
@d1stortion, I didn't contact the pci bridge manufacturer, only zotac & nvidia.
That aside, I made a silly mistake while benchmarking unreal. Vsync was not disabled. As I didn't find any settings in the renderer settings of advanced options and the highest fps being shown as 62, I assumed vsync must be off. When I forced it off from nvidia control panel, I got
Direct3d9: Lowest - 21.9, Highest - 107, Avg - 54.3
wrote:similarly the manufacturer of the gpu's reverse bridge(pericom) confirmed that no separate driver is required and the default windows driver is sufficient.
?
Does anyone know of a simple Windows graphics benchmark that will do a screen fill test and saturate the bus the graphics card is attached to? Something like vspeed for DOS maybe. I am wondering just what PCI performance he is getting.
