Maybe you don't have DMA enabled harddrive transfers for some reason. That generally shits up performance as well.

The HDD is in Ultra DMA Mode 5. The speeds I am getting are quite good for a laptop and the overall system performace. It is also slow when not accessing the disk.

http://cpu.rightmark.org/products/rmma.shtml - this might do for examining cache - that stepped speed graph shows the breakdown of primary, secondary and RAM - not sure how a 64 I / 64 D / 64 secondary would show up - if the 64 secondary is full speed, then there may not be much of a break point between primary and secondary.

I'll tell you that the chipset doesn't suck but there is very little that you can tweak to make that cpu run any better. Laptops that used that cpu were more or less the very very back of the economy section with the only selling point besides cost was battery life. VIA could have done a lot better and even now they got a Quad core that is almost on par with AMD but no still not on the market.

Whoa those Everest benchmark results are bizarre...

Everest can test cache levels. Tools -> Cache and Memory Benchmark.

Could the CPU be overheating and throttling? Though I don't even know if C3 has that capability.

You might find this thread of interest, i have a 1.5ghz esther desktop, even with DDR2 ram and SATA it is still dog slow, bottom of the page i posted superpi scores against a P4 1.8ghz.
Has anyone used modern VIA CPUs?

Nehemiah is the last core revision of the viac3 series, and somewhat faster than the older ones.
however i just can't understand that since they are all based on idt centaur design, how can the viac3 be even slower than the winchip clock to clock?

Thank you Matth79. RMMA will be a useful program for me. The cache was working. DDR RAM is installed.

Swaaye is probably correct. After reseating the heatsink and (loud) fan, the performance is now in the expected range of around 500 MHz PIII and significantly worse in floating point, in line with the other benchmarks. I haven't had anything that doesn't spew out hot air in normal operation overheat before. Apparently the CPU does have throttling; it remained stable in that condition and didn't crash even once.

As a games machine it is not very useful. The UniChrome graphics can't render any games with decent framerate with the screen resolution of 1024*768. At least it is a standard one. GLQuake and Half-Life were almost playable.

Via chips are going to be a lot slower than a comparable speed P-III. Here's some benches

http://www.tomshardware.com/reviews/VIA-s-C3- … -GHz,472-8.html

Slower than a 667mhz Celeron in many tests.

I know its a laptop, and things like sleep-mode/hibernation are usually important for quick access, but that system is so slow that my suggestion would be to get a hold of a copy of Windows XP SP1 or earlier because that system would fare better with a no-ACPI install, no hibernation, no system-restore. It would mean that in order to power it down, you would have to shut it down, manually, but the performance gain is worth it IMO.

After that I would run services.msc and disable anything not of absolute importance.

noshutdown wrote:

Nehemiah is the last core revision of the viac3 series, and somewhat faster than the older ones.
however i just can't understand that since they are all based on idt centaur design, how can the viac3 be even slower than the winchip clock to clock?

AMD K5 was faster than K6 on a per clock basis. My guess would be that like this example the intent was to improve clock speed scaling. Centaur didn't have the R&D capabilities to build a state-of-the-art processor and the compromises aren't pretty when compared to their competition with much more money.

My observations are the same now, sliderider.

The heatsink was slightly dirty. But after removing it I saw a shiny pad underneath, which consisted of a foil and a dark grey thermal compund (presumably). I noticed that the foil had an uneven surface and a fingerprint in the middle. The CPU package was completely clean. The foil came right off. I scraped and rubbed off the old thermal interface material and applied some MX-2 paste before putting the HS back. The computer is now working stable at its expected speed for half a day. Could it be that the foil should have been removed during assembly?

I removed hibernation, system restore, system file protection, firewall, security center, etc., from the ISO using "nLite". I didn't expect to make this into a games machine, for which SP1 would indeed work better. Every now and again I come by a small utility that either complains about missing EncodePointer or GetLogicalProcessorInfo functions, hence SP3. Actually most 90s games run okay on it. I was just used to having settings on the max and set my expectations too high. I was disappointed that Rally Championship 2K had a low framerate. Anything more recent works on modern cards anyway.

Sure a C3 Nehemiah has to be rated half its MHz when compared to a P-III Coppermine.
Still, at 1200MHz for example it runs Quake in 640x480 at 71 FPS.
Considering it is multiplier adjustable on the fly, it is like an AMD K6+ but able to plug in a i440BX motherboard. An interesing option for retro systems IMO.

j7n wrote:

My observations are the same now, sliderider.

The heatsink was slightly dirty. But after removing it I saw a shiny pad underneath, which consisted of a foil and a dark grey thermal compund (presumably). I noticed that the foil had an uneven surface and a fingerprint in the middle. The CPU package was completely clean. The foil came right off. I scraped and rubbed off the old thermal interface material and applied some MX-2 paste before putting the HS back. The computer is now working stable at its expected speed for half a day. Could it be that the foil should have been removed during assembly?

I removed hibernation, system restore, system file protection, firewall, security center, etc., from the ISO using "nLite". I didn't expect to make this into a games machine, for which SP1 would indeed work better. Every now and again I come by a small utility that either complains about missing EncodePointer or GetLogicalProcessorInfo functions, hence SP3. Actually most 90s games run okay on it. I was just used to having settings on the max and set my expectations too high. I was disappointed that Rally Championship 2K had a low framerate. Anything more recent works on modern cards anyway.

I wonder if there is a way to make a patch to fake EncodePointer atleast. Pretty stupid for a program to fail to run for something as useless as "EncodePointer"

And yes, VIA CPUs are *very* slow

I have the C3 800Mhz reviewed in this article.

Since it's apparently slower than a Celeron or P3 at half the clock speed, I've never tried using it for anything. But it's gorgeous, and heavy. It's nice to have on my desk. 😀

It was designed for low power consumption and low price, so it has a much lower transistor count than a P3 or Celeron. Comparing it to a P3 is like comparing an Atom processor to a Core i7.

swaaye wrote:

noshutdown wrote:

Nehemiah is the last core revision of the viac3 series, and somewhat faster than the older ones.
however i just can't understand that since they are all based on idt centaur design, how can the viac3 be even slower than the winchip clock to clock?

AMD K5 was faster than K6 on a per clock basis. My guess would be that like this example the intent was to improve clock speed scaling. Centaur didn't have the R&D capabilities to build a state-of-the-art processor and the compromises aren't pretty when compared to their competition with much more money.

k5 is faster than k6 per clock, but its only because k5 was designed by amd while k6 wasn't. amd's own team has been working on k7 after k5 was finished, and k6 was developed by nexgen's team purchased by amd. so they had different tech sources, unlike winchip and viac3 were both from idt centaur.

noshutdown wrote:

k5 is faster than k6 per clock, but its only because k5 was designed by amd while k6 wasn't. amd's own team has been working on k7 after k5 was finished, and k6 was developed by nexgen's team purchased by amd. so they had different tech sources, unlike winchip and viac3 were both from idt centaur.

I'm sorry. That aspect slipped my mind for some reason. P3 and P4 are a better example of clock scaling compromise I suppose. C3 has fairly impressive power consumption though AFAIK. The Centaur simplicity approach plus low power consumption would seem to preclude IPC gain.

AidanExamineer wrote:

I have the C3 800Mhz reviewed in this article.

Since it's apparently slower than a Celeron or P3 at half the clock speed, I've never tried using it for anything. But it's gorgeous, and heavy. It's nice to have on my desk. 😀

Ah, I have been reading that before. 😀
That 800 MHz C3 is probably a Samuel 2 core. My take on C3s older then the Nehemiah is that they are either bugged (way to go VIA) or incompatible with the mainboards I tried: It seems the internal caches are not working properly, for as far as they are present. IMO C3 Samuel and Ezra core are therefor only interesting for systems that are slow on purpose.
The C3 Nehemiah has caches that just work, and is TWICE as fast in Quake compared to its predecessors. At the same clockspeed that is.