How fast is the C3 1200? Is it as fast as a Tualatin at the same clock? Does it work in Coppermine motherboards?

sliderider wrote:

How fast is the C3 1200? Is it as fast as a Tualatin at the same clock? Does it work in Coppermine motherboards?

From what I understand, it works fine in many high quality coppermine motherboards. However, it's more aligned with a 800mhz coppermine than a 1.2ghz Tualatin. Maybe even a bit slower.

sliderider wrote:

How fast is the C3 1200? Is it as fast as a Tualatin at the same clock? Does it work in Coppermine motherboards?

It works at a real 1.2Ghz, and it's the only VIA C3 core which has the FPU at full speed instead of just half speed (like Samuel and Ezra), but according to the articles I've read, it's still considerably slower then a real P3, even slower then a Celeron.

I never benchmarked one, at least now yet, because I haven't had these chips very long yet and haven't made a s370 rig since I got them.

They are supposed to run very cool though, requiring less cooling then a P3 or Celeron at the same speed.

They are low heat because they are such simple CPUs. They are similar to a Pentium (but with SIMD) and that's why a P6 or K7 dusts them. I think they were most popular as cheap CPUs for embedded and industrial products.

Can you use them in SMP boards?

sliderider wrote:

Can you use them in SMP boards?

You would be the first one to do it... I can't find a single instance of a C3 SMP rig.

Apparently even the Nehemiah chips suck. A 1ghz Nehemiah gets beaten down by a 600mhz celeron according to some forum chatter I just read. That's painful...

sliderider wrote:

Can you use them in SMP boards?

I read the C3 datasheet and it mentions Nehemiah currently not supporting this, so probably not.

But that doesn't mean it can't be done though (like the early Celeron).

Tetrium wrote:

sliderider wrote:

Can you use them in SMP boards?

I read the C3 datasheet and it mentions Nehemiah currently not supporting this, so probably not.

But that doesn't mean it can't be done though (like the early Celeron).

The difference there is that Intel inadvertently left that functionality in when they created the P-III derived version of the Celeron. They disabled it later on when motherboards like the ABit BP6 were released with two 370 sockets and instructions for running dual Cellies in SMP mode.

Wasn't that the Mendocino Celeron? I thought Intel fixed this SMP oversight with the Coppermine Celerons?

At any rate, afaik the C3's aren't supposed to be run in dual-CPU, but I can't exclude that there still is "some" way to get them running in dual-CPU systems? I don't really know as I never build a dual-CPU rig myself (only ever experience is trying out 2 Katmais in a dual Slot 1 board, had it post once, and that was it 🤣!).

There was one user here with a lot of dual-CPU rigs, maybe he knows more?

You've given me an idea- how cool would it be to build a wierd gaming rig with a C3 cpu and a Matrox or S3 vid card and then tweak it to perform as well as a P3 + Voodoo or Geforce combination?

sgt76 wrote:

You've given me an idea- how cool would it be to build a wierd gaming rig with a C3 cpu and a Matrox or S3 vid card and then tweak it to perform as well as a P3 + Voodoo or Geforce combination?

Lol, thats kinda the same thinking I've been having 😜

I was thinking C3 1.0Ghz or 1.2Ghz + a Voodoo 2 (and see if theres any CPU bottleneck 😜) + Windows 9x (which in my case is Windows ME in like 99% of the cases 😜).

Personally I'd expect at least Windows 9x performance to be quite good on such a rig. No idea about gaming though, will depend for a large part on the FPU performance of the CPU. Samuel and Ezra have the (already weak) FPU at only half the clock speed of the CPU anyway, so they should be comparable to (thinking.......) say, a K6-2/300 when at 800/900 Mhz??
I'm guessing the Nehemiah would make an awesome Super 7-ish rig, but without jumpers, with all memory cached and with silent cooling (as big heatsinks won't be blocked by caps and other stuff), and with AGP 4x...and ATA-66 or ATA-100...

Tetrium wrote:

But wait! Theres more!! […]
Show full quote

swaaye wrote:

Oh yeah for some reason I got caught up on the 6x86 and MII.

But I see even less reason for a VIA C3! 😁

But wait! Theres more!!

With the C3 you can finally get Super Socket 7 performance without having all the troubles with setting voltage and having to deal with crappy AGP implementation.
And theres more! It will actually work on an Intel chipset! Stability! (if you don't take into account the C3's bugs of which I know very little, but am not really worried about anyway 😜)

Even if it has bugs, it's probably related to for that time new features. Building it as somekind of Super 7-ish rig will probably circumvent those problems as it'll run older code mostly...at least that's the hypothesis 😜

Edit:And theres 1 more advantage the C3 has over Coppermine. It's just darnedly looks so much better!

Just compare this boring pale piece of green plastic to this golden beauty! 😁

Edit2:And the advantages just keep on coming, yeah, haha!

I tested using the CUSL2, a motherboard that doesn't support Tualatin.
This means the fastest chip it normally supports is either the P3-1000 (good luck finding the 1133 part) or the Celeron Coppermine 1100.
But!...this board works perfectly fine with a 1.2Ghz Nehemiah!
Yup, VIA C3 is the fastest clocked CPU supported by boards that won't work with Tualatin!!11 😁 😜

Somehow reading this post made me think of these

http://www.youtube.com/watch?v=tGvHNNOLnCk

http://www.youtube.com/watch?v=_-8IufkbuD0&NR=1

🤣

Wow...that guy gives me a headache 🤣!

Tetrium wrote:

Wow...that guy gives me a headache 🤣!

Be glad you didn't live in the NY-NJ area in the 70's and 80's. We got hammered with ads like this on TV every few minutes.

http://www.youtube.com/watch?v=fO9XC3tAbkQ

And here's a special one for Vogoners

http://www.youtube.com/watch?v=ivhRPiQueaE&feature=related

SavantStrike wrote:

Then I got brainy when I saw Tetrium talking about the Via C3 chips in another thread. Some of them are unlocked from the factory (the earlier ones are, 600 or 700mhz plus are often locked).

What about getting an early C3 and running it at around a 1x multiplier, especially on a slotket residing in a 440BX based motherboard (guaranteed ISA slots, and 440BX hotness). Seeing as they're already not as fast as a PIII clock for clock, it might be possible to get that thing to run really slowly, especially if there's a way to disable the cache (is there)?

Such a system should be capable of running all but the oldest of old games. If the cache can be disabled, I would imagine it might even get as slow as a Pentium 1 without cache.

I have spend some time to come to know early VIA Cyrix III/C3 CPUs. Yes, early C3 CPUs are all unlocked: all CPUs with Samuel (C5A) core are unlocked, but later CPUs with Samuel 2 (C5B) core (and as I know Ezra too) have locked multiplier. You cannot identify it that is Samuel (C5A) or later by CPU printing (logo, name), but you can identify it about the voltage: Samuel (C5A) core is 1.8-2.0 V but laters are 1.65 V or less.
I have collected these information (about CPU photos with both sides of the chips) here:
http://en.wikipedia.org/wiki/List_of_VIA_micr … chnology_design

I have the fastest Samuel (C5A) the 733MHz one and I can underclock, I can change the multiplier. I have tested with DOS games and the "Runtime error 200" speed limit. On 650MHz and lower there is no "Runtime error 200" but 667MHz and higher the "Runtime error 200" is present. I have tested all the DOS games on my videocard compatibility matrix and I have not found problem. Also I have test Grim Fandango (Win9x Direct3D game) that crash on PII 400 and faster, and not problem on VIA 733. The MHz can be deceptive because these VIA CPUs are really slow (for example the PII 350 is outperform the C3 733 and the Pentium 233 MMX much fester with DOS games than the C3 650MHz). So C3 is compatible because really slow.
But using VIA CPU with Intel BX board can be problemous. I have tested three of them and I have problems with multiplier. The motherboard compatibility with VIA CPU is not well documented. I think Samuel (C5A) with FSB 100MHz would not problemous on these boards:
https://web.archive.org/web/20140912050048/ht … 3/boardlist.jsp
(that list is last updated on Dec. 2003).
But if someone want a more powerful DOS gaming machine beyound the "Runtime error 200" speed limit the AMD K6-III+ 500MHz (or slower) the best choise (but nowadays a bit expensive).

Some DOS games/programs make "divide by zero" error message on all 486 and laters CPUs (because the CPU cache). On lot motherboards you can enable/disable CPU L1 cache in bios, but DOS pragrams can do it too and you can use these games/programs easily on your 486 and later systems.

Great info! You just answered most of the questions that remained regarding those mysterious VIA/Cyrix Socket 370 CPUs. No mistake they are very slow per clock. What is the lowest multiplier you can set? What BX board did you use?

You say you have problems with the multiplier. Do you mean that the jumperless BIOS multiplier setting does not match the observed multiplier? (That is to be expected. A pentium II for example has a different interpretation of the multiplier BF signals compared to a Tualatin core. It works best to use a jumpered board and try+note all settings for that CPU type.)

I have tested three BX board:
Abit AB-BX6: not work at all;
something Lucky Star: Boot as 550MHz but writes Celeron;
ASUS P3B-F: Not boot with jumperless, jumpered to FSB 100 will boot as 550MHz whatever is the multiplier.
But the CPU is not multiplier locked because with my EPoX EP-693A (chipset: VIA Apollo Pro 133) can change the multiplier. Yes, multiplier work in different way as PII (might Tualatin too), the EP-693A have a "VIA Cyrix III CPU Ratio" submenu in the bios. The lowest multiplier that EP-693A allow is 4x (both for VIA and Intel CPUs) so the slowest is 4x66=266MHz. I believe that the Samuel (C5A) CPU on 266MHz is far the slowest "family 6" CPU performance.
Later I have bought a Samuel (C5A) 650MHz (FSB 100 CPU) to test the "Runtime error 200" thing on both Apollo Pro 133 and BX. The BX chipset is faster with VIA CPU too than Apollo Pro 133 but "Runtime error 200" is not occur. With the VIA 650 CPU also I cannot change the clock multiplier on BX boards.
Pentium II 233MHz also have no "Runtime error 200" problem (I have run my 350 on 233 by FSB 66), but VIA 650MHz is faster than PII 233 both on DOS (about 30%) and Win98 (about 15%) too. So the best CPU for a Slot1/Socket 370 DOS retro machine is the VIA CPU 650MHz (likely with the BX will be the fastest), but if someone want to build a DOS only gaming machine, a lot of Socket 7 CPU can outperform the VIA CPU 650MHz thats still below the "Runtime error 200" limit. For example my P1 MMX 233 and my Cyrix MII 433 (300MHz) are faster than Samuel (C5A) 650MHz on Intel BX. The best performer below the "Runtime error 200" limt is K6-III family on 500 MHz (with VIA MVP3) or below the second is K6-2 family also on 500 MHz (with VIA MVP3) or below (might K6-2+ is better than K6-2, I have not tested because I have no K6-2+ CPU). After AMD K6-III and K6-2 families goes the MMX 233 or the Cyrix MII 433 depends on games, some is better with Intel MMX some better with Cyrix. Cyrix with VIA MVP3 under DOS needs register optimalisation (6x86opt) but ALI Aladdin V no need 6x86opt for Cyrix 6x86/MII CPU-s. On Socket 7 in DOS ALI Aladdin V is faster in low resolutions (like 3DBENCH) but VIA MVP3 is faster than Aladdin V in higher resolutions with all CPUs. I have not tested/compared in Win9x the VIA Samuel (C5A) 650MHz to Socket 7 CPUs, maybe VIA can compete in this OS, but under DOS underperforms lot Socket 7 systems.

It's amazing that a Pentium MMX 233 MHz is faster than C5A 650 MHz. That level of performance per clock seems similar to or worse than Centaur's own WinChip 2!

I was amazed too. The C5A 650 MHz by DOS performance is somewhere between the MMX 200 and the MMX 233 MHz. Unfortunately I have no WinChip 2 to compare them but I would be curious. 😀

Tetrium wrote:

But wait! Theres more!! […]
Show full quote

swaaye wrote:

Oh yeah for some reason I got caught up on the 6x86 and MII.

But I see even less reason for a VIA C3! 😁

But wait! Theres more!!

With the C3 you can finally get Super Socket 7 performance without having all the troubles with setting voltage and having to deal with crappy AGP implementation.
And theres more! It will actually work on an Intel chipset! Stability! (if you don't take into account the C3's bugs of which I know very little, but am not really worried about anyway 😜)

Even if it has bugs, it's probably related to for that time new features. Building it as somekind of Super 7-ish rig will probably circumvent those problems as it'll run older code mostly...at least that's the hypothesis 😜

Edit:And theres 1 more advantage the C3 has over Coppermine. It's just darnedly looks so much better!

Just compare this boring pale piece of green plastic to this golden beauty! 😁

Edit2:And the advantages just keep on coming, yeah, haha!

I tested using the CUSL2, a motherboard that doesn't support Tualatin.
This means the fastest chip it normally supports is either the P3-1000 (good luck finding the 1133 part) or the Celeron Coppermine 1100.
But!...this board works perfectly fine with a 1.2Ghz Nehemiah!
Yup, VIA C3 is the fastest clocked CPU supported by boards that won't work with Tualatin!!11 😁 😜

Are you related to Steve Ballmer?

http://www.youtube.com/watch?feature=player_d … e&v=tGvHNNOLnCk