Benchmarking Win9x in a VM like QEMU or PCem is cheating, IMO. If you're going to do that, why limit yourself to a 7900 GTX? The rest of the system is nowhere near period-correct, so I'm curious to know the logic behind your choice of a semi-period-correct GPU like the 7900 GTX. Why not use the most powerful GPU you've got available to really juice up that score?

the correct graphics card is detected on 3dm01, isn't this some kind of pass-trough with proper drivers and usage of the videocard?
if that is the case, it's not really being emulated

SPBHM wrote on 2023-07-17, 00:57:

the correct graphics card is detected on 3dm01, isn't this some kind of pass-trough with proper drivers and usage of the videocard?
if that is the case, it's not really being emulated

If you need to use an emulator for cross-compilation and cross-debugging to get Win9x to see and use the physical hardware via a translated or interpreted layer, then yes, it's emulated.

And it's cheating.

Assuming he hasn't used any LOD cheats in 2001 SE, this puts him well ahead of the previous world record from 2016:
https://hwbot.org/submission/3228042_zafiropo … gtx_69807_marks

Though his Game 4 - Nature score is a lot higher, probably due to the LOD cheat...

I suppose scores from QEMU and Real Hardware should be treated separately...

With GPU passthrough, SoftGPU and [that project], those not interested in real hardware do have some options for going the Win9x VM route, which I think is fine, as long as there's a clear separation between VM and Real Hardware builds.

KT7AGuy wrote on 2023-07-16, 22:51:

Benchmarking Win9x in a VM like QEMU or PCem is cheating, IMO. If you're going to do that, why limit yourself to a 7900 GTX? The rest of the system is nowhere near period-correct, so I'm curious to know the logic behind your choice of a semi-period-correct GPU like the 7900 GTX. Why not use the most powerful GPU you've got available to really juice up that score?

Because there aren't drivers for anything better than a 7900 GTX. I've checked.
I'm not emulating a GPU using a 7900 GTX, I'm passing a 7900 GTX through to the VM.
And why? Because I doubt there is anyway to use or emulate anything faster. I mean, you can't pass through a GTX 1080 to 98 because there aren't drivers, there aren't drivers AFAIK for something like a virtio gpu, and any of the emulated cards in PCem or QEMU that are supported are (or should be based on their description) slower then a 7900 GTX.
I'd love to know if I'm wrong. It'd be very exciting for me if there is a way to harness the full horsepower of, say, an RTX 4090 in Windows 98.

KT7AGuy wrote on 2023-07-17, 01:14:

If you need to use an emulator for cross-compilation and cross-debugging to get Win9x to see and use the physical hardware via a translated or interpreted layer, then yes, it's emulated.

And it's cheating.

What do you even mean by that? I'm just using the vfio-pci device in QEMU. Truth be told, I just followed a tutorial from the Gentoo wiki. For transparency, I'll just go ahead and post my QEMU config.

MrMateczko wrote on 2023-07-18, 15:40:

Though his Game 4 - Nature score is a lot higher, probably due to the LOD cheat...

Also, what is the "LOD cheat"?
edit: Googling "LOD cheat" brought nothing useful (which is why I asked), but googling "LOD cheat 3dmark" brought me right to it. It all makes sense now.

Loganius wrote on 2023-07-16, 22:14:

Might have just set the Win98SE 3DMark01 world record. Setup: VM in QEMU with: 1 core from a Ryzen 5 5600x @ 4.8GHz 16GB of RAM […]
Show full quote

Might have just set the Win98SE 3DMark01 world record.
Setup:
VM in QEMU with:
1 core from a Ryzen 5 5600x @ 4.8GHz
16GB of RAM (2GB of which is usable even with patchmem... yay)
NVIDIA 7900 GTX PCIe @ 695/875MHz (core/mem)

what I've learned from this is that 3DMark 01 is insanely CPU bottlenecked.

It's nice to see some more exotic configurations in this thread. 👍
That's a fantastic score! My gosh. As crazy CPU-limited as 3D'01 is, I'm guessing you're actually hitting the limits of the 7900 GTX there.

Just ran it on an RTX 4070/i7-8700 because I don't think I've benched this particular machine yet: 97130. So that's only 23K points more with a much faster GPU, slightly slower CPU, and no crazy VMs or GPU passthrough to potentially trim a couple thousand points off.

Edit: Also tried my 5GHz Ryzen 5900X. Surprisingly, it didn't score much higher than the 4.4GHz i7-8700. "Only" 101170. Was actually expecting a bit more, but I suspect that 3DMark01 doesn't like the dual CCD arrangement of the 5900X. A single CCD Ryzen would probably do a little better, especially one equipped with V-cache like the 5800X3D or 7800X3D.

Edit 2: Sheesh, this thread could use a clean up! Looks like most of my older pictures are either missing or have an ugly watermark over them. I still have most, if not all, on my file server, so I may try and re-up them when I get the time.

Edit 3: Actually, nope. 40 pages, and I was active on most of them. That'd just take way too much time.

P3 650 Slot1 @ 670Mhz
512Mb Reg ECC RAM
Radeon 9800 Pro

Actually i play Auquanox on it, in FullHD

matze79 wrote on 2023-09-28, 20:28:

P3 650 Slot1 @ 670Mhz 512Mb Reg ECC RAM Radeon 9800 Pro […]
Show full quote

P3 650 Slot1 @ 670Mhz
512Mb Reg ECC RAM
Radeon 9800 Pro

Actually i play Auquanox on it, in FullHD

p3.jpeg

Actually it has to be shown around 18k score with proper cpu anyway its interesting combination

Asus P4P800SE
Pentium M 1.86 ghz - OC 2 ghgz
5900XT 390mhz/700mhz - OC 450mhz/800mhz
DDR400 OC 500mhz

18702

very nice, I couldn't help but compare to my Celeron 450 PC!

Motherboard - Asrock P4i65G
CPU - Pentium IV Prescott 3.2Ghz HT (PGA478)
RAM - Kingston 512Mb x2 ddr (Dual Channel)
GPU - Ati Radeon x1950 Pro AGP 512mb GDDR3 (Slightly OC with ati overdrive autotune)
Windows xP Home Edition SP3 32bit
[Hello! this it's my first post in VOGONS, i will appreciate opinions about the score... i'm newbie and i guess the gpu has bottleneck or is okay]

AMD Phenom II X4 980 stock at 3,7Ghz
2GB DDR2 pc1066
AM2NF3-VSTA AM2 mainboard.
Ati Radeon X1950Pro AGP 8x.
Windows XP

Alan Nolath wrote on 2023-10-17, 02:21:

Motherboard - Asrock P4i65G CPU - Pentium IV Prescott 3.2Ghz HT (PGA478) RAM - Kingston 512Mb x2 ddr (Dual Channel) GPU - Ati R […]
Show full quote

Motherboard - Asrock P4i65G
CPU - Pentium IV Prescott 3.2Ghz HT (PGA478)
RAM - Kingston 512Mb x2 ddr (Dual Channel)
GPU - Ati Radeon x1950 Pro AGP 512mb GDDR3 (Slightly OC with ati overdrive autotune)
Windows xP Home Edition SP3 32bit
[Hello! this it's my first post in VOGONS, i will appreciate opinions about the score... i'm newbie and i guess the gpu has bottleneck or is okay]

It seems about right to me. That's about on par with scores I achieved with a Northwood 3.2. I have a HD3850 which only really gets higher scores due to the Nature Benchmark.

indeed, clearly the graphics card can do a lot more, but 3dm01 quickly becomes a CPU benchmark and 18.5k seems high for a 3.2GHz P4, which makes sense given the graphics card

Imperious wrote on 2023-10-20, 00:14:

Alan Nolath wrote on 2023-10-17, 02:21:

Motherboard - Asrock P4i65G CPU - Pentium IV Prescott 3.2Ghz HT (PGA478) RAM - Kingston 512Mb x2 ddr (Dual Channel) GPU - Ati R […]
Show full quote

Motherboard - Asrock P4i65G
CPU - Pentium IV Prescott 3.2Ghz HT (PGA478)
RAM - Kingston 512Mb x2 ddr (Dual Channel)
GPU - Ati Radeon x1950 Pro AGP 512mb GDDR3 (Slightly OC with ati overdrive autotune)
Windows xP Home Edition SP3 32bit
[Hello! this it's my first post in VOGONS, i will appreciate opinions about the score... i'm newbie and i guess the gpu has bottleneck or is okay]

It seems about right to me. That's about on par with scores I achieved with a Northwood 3.2. I have a HD3850 which only really gets higher scores due to the Nature Benchmark.

Thanks for the opinion guys. I'm not expert in benchmarks

I did these a while back. P4 northwood OC'd to 3.6ghz, 7800 GS and HD3850

ayy

ciornyi wrote on 2023-09-29, 10:30:

matze79 wrote on 2023-09-28, 20:28:

P3 650 Slot1 @ 670Mhz 512Mb Reg ECC RAM Radeon 9800 Pro […]
Show full quote

P3 650 Slot1 @ 670Mhz
512Mb Reg ECC RAM
Radeon 9800 Pro

Actually i play Auquanox on it, in FullHD

p3.jpeg

Actually it has to be shown around 18k score with proper cpu anyway its interesting combination

Yeah i switched the CPU for fastet 100Mhz Coppermine.
Now its running 1,1Ghz and 7853 3Dmark 2001's.

Here's a 4090 with my 12900. It scored lower than the 3090 with the same CPU, im sure its just driver differences and possibly new windows code that causes the difference. At this point only the CPU really seems to matter.

Ryzen 7 5800X3D OC on WinXP :