bloodem wrote on 2024-02-08, 15:35:

Remove everything that you don't need and try again. 😀

BitWrangler wrote on 2024-02-09, 02:55:

Got any USB cards in it? That's a frequent bugbear.

Ok the first issue I eliminated... Turns out the large HDD support from @DenizOezmen isn't working properly with a SATA to IDE bridge and a 120gb SSD. Switching to a 16GB SSD solved the issue where Quake would lock up when loading the map...

The second issue is Quake locking up with everything maxxed out at 1024x768 when applying the maxxed out settings upon automatic restart. I think this is an AGP power delivery issue with my boards (P2B, P2B-S, P3B). The P3B was ok with an MX440 8x, but other cards (64-bit MX420, first gen MX440) exhibited this issue.

I'll try with an FX5200 and report back.

By the way, even an MX440 is limited by CPU in Quake3. At 1024x768, you'll be stuck with an average of 90FPS with a PIII 650E, and it goes down from there, the less powerful the CPU. I would have tested the C3 on the P3B-F, but either it needs a unicorn Slotket to work, or it doesn't work at all with that board.

EDIT: Ok, good news. I finally found a combination where everything works... The GeForce FX5600SE is completely stable on the P2B-S, it must have a low power draw on the AGP... It performed comparably to the other cards with the P3 650E, (90 or so FPS). With the Ezra-T at 1Ghz, performance drops to around 60fps -- still quite acceptable. Ok, I am happy with this arrangement and can finally proceed with my build... The problems with this system have been dogging me for a couple of years already.

I'm requesting another sanity check please:

I'm testing my C3 Nehemiah on my Gigabyte GA-6BXC with the Powerleap BIOS (for that matter - is it normal that there's no power monitoring in this BIOS version?), and I cannot loop 3DMark 2000 for more than 24 hours... It always crashes either back to desktop with a hard lock or just crashes back to desktop.

OTOH, I placed a Katmai PIII in there and it's been looping for more than 24 hours no issues. To be fair, I did test a PIII Coppermine on my P3B-F and that also crashed to desktop eventually with 3DMark, but I don't trust that board for stability, long-term.

I tried downclocking the Nehemiah all the way to 6x100 (600mhz), that didn't help...

So if you can, would you please loop 3DMark2000 and report whether it can survive more than 24 hours of looping?

Videocards I tested were all GeForce 4 MX variants.

Thanks

EDIT (01/12/2025 12:28 EST): I did some more testing. For some reason Katmai is 3DMark 2000 infinite loop stable on the GA-6BXC, but not Coppermine (or C3, for that matter). I had a P2B run the same Slot 1 Coppermine (500E) with the same test, and it is stable. All my boards and PSUs have new capacitors so that's ruled out. Last night I put the C3 in the P2B and it has been running fine until now and I presume it will be completely fine looping for however long. So I'll be abandoning the GA-6BCX in favor of the P2B (nope. just checked, it's crashed to desktop -- next up is my P2B-S) The only difference will be the inability to switch back to a higher FSB with SMB after going down to 50Mhz. I can live with that sacrifice if it means a completely stable system.

As to what is causing the instability on some boards is probably AGP power delivery. My P2B is modded to take 3.3V straight from the PSU.

its an old post but it remember testing an early version of viac3-466(probably samuel1 without any l2 cache) underclocked to 233 and compared against socket7 cpus:

it was tested with asus cubx-e, geforce2ultra and windows98se.
as you can see, a c3-233 is most comparable to a k5-100.

bloodem wrote on 2022-02-16, 06:33:

EDIT (March 30th, 2022): This thread's original purpose was to show how a VIA C3 build compares to a high performance SS7 platfo […]
Show full quote

EDIT (March 30th, 2022): This thread's original purpose was to show how a VIA C3 build compares to a high performance SS7 platform. It has since evolved into a benchmark marathon (and maybe something more in the future).
Further down, you'll find a list of links to different posts in this thread that contain video benchmarks, conclusions and more information.
I plan to add more videos in the future, so this list will be constantly updated.

As a follow-up to a discussion that started here, I've decided to create a separate thread where I (we) can make an accurate comparison (based on empirical data) between VIA C3 and other platforms.

A couple of rules which I plan on following when performing these comparisons:
- record a ~15 minute video where the exact software & used settings are perfectly visible (this way we can have a real apples to apples comparison). Also, will try and do all tests using the same settings (with just a bit of variation for entertainment purposes).
- whenever possible, use the exact same platform for testing (i.e.: All Slot 1/Socket 370 CPUs will be tested using the same GA-6BXC motherboard, the same GeForce 2 Ti card which is overclocked at GeForce 2 Ultra speeds, and the same drivers: 7.76).
- for different platforms, use the same GeForce 2 Ti card / same drivers (7.76)
- overclocks are totally fine (and recommended, actually), since they allow us to better understand what a CPU's true potential is, but I will stick to clocks which should be achievable with most/all CPUs of that particular type (of course, the overclock MUST be 100% stable).

In this initial video, I'm testing the 1 GHz VIA C3 "Ezra-T" (default FSB: 100 / multiplier: 10.0), which - based on my experience with 7 identical CPUs - can be easily overclocked to at least 1.2 GHz (FSB: 133 / multi: 9.0). I went with 1.26 GHz for this test (FSB: 133 / multi: 9.5), because I was able to hit this frequency with all my CPUs (some required a very small voltage bump to be 100% stable).
Sorry for the quality, it's not particularly good. I do have a VGA capture device, but it only captures at a maximum of 30FPS/Full HD resolution, so I decided to just record the screen with an iPhone 12 @ 4k/60FPS. The sound is recorded separately and synced in post.

FULL SYSTEM SPECS:
MB: Gigabyte GA-6BXC rev 1.9 (VRM mod)
CPU: VIA Ezra-T 1 GHz (OC @ 1.26 GHz / FSB133 & 9.5 multi) using an MSI MS-6905 "Master" slotket V2 (works fine even with other no-name slotkets, though).
RAM: 3 x 128 MB SAMSUNG SDRAM PC133 (using 3 sticks was a bit faster than with 1 or 2 - might be because memory bank interleave is automatically enabled when all memory slots are populated, though I'm not even sure if the 440BX supports memory bank interleave).
VIDEO: Asus V7700Ti GeForce 2 Ti (OC @ GeForce 2 Ultra clocks)
SOUND: Creative Sound Blaster Live 5.1 SB0220
SOUND2: ESS AudioDrive ES1688F (non-PNP / irrelevant for this test).
HDD: Seagate 40 GB IDE/PATA

YOUTUBE VIDEOS:

1. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & VIA C3 EZRA-T 1 GHz OC @ 1.26 GHz (first benchmark), see also the final VIA C3 Ezra-T benchmark, graphs and conclusion in this post
   
   
2. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & VIA C3 Nehemiah 1.2 GHz OC @ 1.46 GHz (and additional information)
   
   
3. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & PENTIUM 3 KATMAI @ 600 MHz (and additional information)
   
   
4. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-2+ 550 OC @ 633 MHz using a GeForce 2 Ultra (and additional information)
   
   
5. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 2 Ultra (and additional information)
   
   
6. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a Voodoo 3 3500 (and additional information)
   
   
7. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 2 MX400 (and additional information)
   
   
8. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 4 MX440 AGP8X (and additional information)
   
   
9. WIN98 GAMING BENCHMARKS: REVISITED: Gigabyte GA-6BXC & VIA C3 Ezra-T with/without OC (+ graphs, conclusion and additional information)

NEXT UP: Will switch to MS-DOS. Stay tuned.


A few pics with the Ezra-T test system:

Hi, I am very impressed with your work on the GA-6BXC +VIA C3 project. I have a question regarding the Ezra-T processor. I am using the GA-6BXC REV board. 2.0 with the latest powerleap bios and an EZRA-T 1000 FSB 100 MHz processor. On startup it reports the processor as VIA C3 231MHz (77x3.0). I mount NEHEMIAH 1,2 everything is correct. Greetings. Piotr

piokum77 wrote on 2025-02-07, 19:18:

Hi, I am very impressed with your work on the GA-6BXC +VIA C3 project. I have a question regarding the Ezra-T processor. I am using the GA-6BXC REV board. 2.0 with the latest powerleap bios and an EZRA-T 1000 FSB 100 MHz processor. On startup it reports the processor as VIA C3 231MHz (77x3.0). I mount NEHEMIAH 1,2 everything is correct. Greetings. Piotr

Hello! I don't have any GA-6BXC available for testing right now, they're all in storage, but if I remember correctly, indeed, it doesn't recognize the Ezra-T properly. Can you confirm that other than this cosmetic issue, everything else works as it should and that you see the correct speed in Windows / MS-DOS?

piokum77 wrote on 2025-02-07, 19:18:

Hi, I am very impressed with your work on the GA-6BXC +VIA C3 project. I have a question regarding the Ezra-T processor. I am using the GA-6BXC REV board. 2.0 with the latest powerleap bios and an EZRA-T 1000 FSB 100 MHz processor. On startup it reports the processor as VIA C3 231MHz (77x3.0). I mount NEHEMIAH 1,2 everything is correct. Greetings. Piotr

This is expected behavior for this motherboard and BIOS... My Nehemiah posts at "180mhz FSB" which is obviously spurious. Maybe someone will fix that BIOS for us one day.

Thank you very much, colleagues, for your reply. Apart from this minor ‘cosmetic problem’ everything works under DOS/ Windows 98 SE. On the Asus P2B 1.12 Ezra-T is detected correctly, the only thing missing is support for FSB 50 MHz. Greetings

piokum77 wrote on 2025-02-07, 23:28:

Thank you very much, colleagues, for your reply. Apart from this minor ‘cosmetic problem’ everything works under DOS/ Windows 98 SE. On the Asus P2B 1.12 Ezra-T is detected correctly, the only thing missing is support for FSB 50 MHz. Greetings

Issue this command in DOS for 50Mhz FSB: smb /sp 10 50

You can download the program here.

Asus P2B rev. 1.12 does not support FSB 50 MHz.
66, 100 and 133 using Rayer SMB work perfectly.

piokum77 wrote on 2025-02-10, 22:55:

Asus P2B rev. 1.12 does not support FSB 50 MHz.
66, 100 and 133 using Rayer SMB work perfectly.

Revision 1.12 of the ASUS P2B has the ICS9248-39 PLL clock generator chip like my P2-99. Using the 50 MHz FSB parameter with RayeR's SMB utility causes it to run at 124 MHz FSB with the PCI bus clocked to 44.3 MHz and the official parameters to set the FSB to 133 MHz is the one having the PCI bus clocked at 41.6 MHz.

Check my cheat sheet if you are going to use that utility to change the FSB on the fly.

Hello, so I have 2 C3 cpus:
Via C3 800 mhz 133x6 v1.25 CPUID: 000698 (this one seems quite a weird cpu)
Via C3 1200 mhz 133x9 v1.45 CPUID: 00069A.
Picture: https://imgur.com/a/RxtyAnC

In the end there is any performance difference between a the different cores?

Nemo1985 wrote on 2025-02-11, 11:34:

Via C3 800 mhz 133x6 v1.25 CPUID: 000698 (this one seems quite a weird cpu) Via C3 1200 mhz 133x9 v1.45 CPUID: 00069A. <snip> In […]
Show full quote

Via C3 800 mhz 133x6 v1.25 CPUID: 000698 (this one seems quite a weird cpu)
Via C3 1200 mhz 133x9 v1.45 CPUID: 00069A.
<snip>
In the end there is any performance difference between a the different cores?

I would guess the C3 800 is an Ezra core and the C3 1.2Ghz is a Nehemiah core. Yes, the Nehemiah is faster clock per clock and has more features (CMOV, SSE).

mockingbird wrote on 2025-02-11, 14:02:

Nemo1985 wrote on 2025-02-11, 11:34:

Via C3 800 mhz 133x6 v1.25 CPUID: 000698 (this one seems quite a weird cpu) Via C3 1200 mhz 133x9 v1.45 CPUID: 00069A. <snip> In […]
Show full quote

Via C3 800 mhz 133x6 v1.25 CPUID: 000698 (this one seems quite a weird cpu)
Via C3 1200 mhz 133x9 v1.45 CPUID: 00069A.
<snip>
In the end there is any performance difference between a the different cores?

I would guess the C3 800 is an Ezra core and the C3 1.2Ghz is a Nehemiah core. Yes, the Nehemiah is faster clock per clock and has more features (CMOV, SSE).

Thank you, actually both of them are identified as Nehemiah and the cpu layout seems the same. I should try to test them at the same frequency?

Just for some perspective on some 'real-world' Nehemiah performance, with a PIII 650 Coppermine at 100Mhz FSB, I get around 85FPS for Quake III 1.32c DEMO001 with everything maxxed out at 1024x768 (32bpp, 32bit textures, trilinear, quality high,...) using an FX5200 clocked at 270/275(550). With the Nehemiah 1.2Ghz at 100Mhz FSB (12 x 100), I get around 94FPS, so I'd put it on par with a P3 Coppermine 650. At 133Mhz FSB at 866Mhz, the Coppermine scores around 115FPS.

I tested my 3 cpus on my qdi advance 10T
Via c3 Samuel 2 800 mhz v1.65:
CPU-ID Signature: 000673
Quake 640: 21.6
Fastdoom: 529 ticks

Via c3 Nehemiah 800 mhz v1.25:
CPU-ID Signature: 000698
Quake 640: 23.1
Fastdoom: 472 ticks

Via c3 Nehemiah 1200 mhz v1.45:
CPU-ID Signature: 00069A
Quake 640: 24.6
Fastdoom: 449 ticks

I also tested the last cpu at 800 mhz but the performance despite the different cpu signature were the same of the first nehemiah, but with the second one I was also able to reach 1463 with setmul with some added performance.

Good morning,
I have been browsing your thread on Vogons regarding the GA-6BXC and VIA C3 - I am very impressed, a piece of great work, congratulations. You wrote that you are in possession of several pieces of the VIA C3 EZRA-T 1000 MHz processor. Would you resell one piece ?
Greetings
Piotr - retro collector from Poland

bloodem wrote on 2022-02-16, 06:33:

EDIT (March 30th, 2022): This thread's original purpose was to show how a VIA C3 build compares to a high performance SS7 platfo […]
Show full quote

EDIT (March 30th, 2022): This thread's original purpose was to show how a VIA C3 build compares to a high performance SS7 platform. It has since evolved into a benchmark marathon (and maybe something more in the future).
Further down, you'll find a list of links to different posts in this thread that contain video benchmarks, conclusions and more information.
I plan to add more videos in the future, so this list will be constantly updated.

As a follow-up to a discussion that started here, I've decided to create a separate thread where I (we) can make an accurate comparison (based on empirical data) between VIA C3 and other platforms.

A couple of rules which I plan on following when performing these comparisons:
- record a ~15 minute video where the exact software & used settings are perfectly visible (this way we can have a real apples to apples comparison). Also, will try and do all tests using the same settings (with just a bit of variation for entertainment purposes).
- whenever possible, use the exact same platform for testing (i.e.: All Slot 1/Socket 370 CPUs will be tested using the same GA-6BXC motherboard, the same GeForce 2 Ti card which is overclocked at GeForce 2 Ultra speeds, and the same drivers: 7.76).
- for different platforms, use the same GeForce 2 Ti card / same drivers (7.76)
- overclocks are totally fine (and recommended, actually), since they allow us to better understand what a CPU's true potential is, but I will stick to clocks which should be achievable with most/all CPUs of that particular type (of course, the overclock MUST be 100% stable).

In this initial video, I'm testing the 1 GHz VIA C3 "Ezra-T" (default FSB: 100 / multiplier: 10.0), which - based on my experience with 7 identical CPUs - can be easily overclocked to at least 1.2 GHz (FSB: 133 / multi: 9.0). I went with 1.26 GHz for this test (FSB: 133 / multi: 9.5), because I was able to hit this frequency with all my CPUs (some required a very small voltage bump to be 100% stable).
Sorry for the quality, it's not particularly good. I do have a VGA capture device, but it only captures at a maximum of 30FPS/Full HD resolution, so I decided to just record the screen with an iPhone 12 @ 4k/60FPS. The sound is recorded separately and synced in post.

FULL SYSTEM SPECS:
MB: Gigabyte GA-6BXC rev 1.9 (VRM mod)
CPU: VIA Ezra-T 1 GHz (OC @ 1.26 GHz / FSB133 & 9.5 multi) using an MSI MS-6905 "Master" slotket V2 (works fine even with other no-name slotkets, though).
RAM: 3 x 128 MB SAMSUNG SDRAM PC133 (using 3 sticks was a bit faster than with 1 or 2 - might be because memory bank interleave is automatically enabled when all memory slots are populated, though I'm not even sure if the 440BX supports memory bank interleave).
VIDEO: Asus V7700Ti GeForce 2 Ti (OC @ GeForce 2 Ultra clocks)
SOUND: Creative Sound Blaster Live 5.1 SB0220
SOUND2: ESS AudioDrive ES1688F (non-PNP / irrelevant for this test).
HDD: Seagate 40 GB IDE/PATA

YOUTUBE VIDEOS:

1. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & VIA C3 EZRA-T 1 GHz OC @ 1.26 GHz (first benchmark), see also the final VIA C3 Ezra-T benchmark, graphs and conclusion in this post
   
   
2. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & VIA C3 Nehemiah 1.2 GHz OC @ 1.46 GHz (and additional information)
   
   
3. WIN98 GAMING BENCHMARKS: Gigabyte GA-6BXC & PENTIUM 3 KATMAI @ 600 MHz (and additional information)
   
   
4. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-2+ 550 OC @ 633 MHz using a GeForce 2 Ultra (and additional information)
   
   
5. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 2 Ultra (and additional information)
   
   
6. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a Voodoo 3 3500 (and additional information)
   
   
7. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 2 MX400 (and additional information)
   
   
8. WIN98 GAMING BENCHMARKS: ASUS P5A & AMD K6-3+ 570 OC @ 633 MHz using a GeForce 4 MX440 AGP8X (and additional information)
   
   
9. WIN98 GAMING BENCHMARKS: REVISITED: Gigabyte GA-6BXC & VIA C3 Ezra-T with/without OC (+ graphs, conclusion and additional information)

NEXT UP: Will switch to MS-DOS. Stay tuned.


A few pics with the Ezra-T test system: