Well what we do have is the eveidence that this card has 64MB of ram no doubts about that as far as I can tell

because a 3D mark difference of 160 Points should be a descent difference to see the odds right 8)

And yeah so far it is the only 64MB PCI 4500 in the world since we have never seen such a thing with a standard single GPU PCI 3dfx card to be honest there!

what about games like Unreal with as without FSAA what differences do you get from that game or Quake3 and so on, it might be interesting to test on how it would perform in Glide apps as OpenGL apps since it was designed to do Glide in the first place

Hmm but would that explain the performance difference in 3D Mark 2001SE B330 ?

Since the difference between the 4500 AGP with 32 as 64MB show the same type of difference with the 32MB as 64MB PCI and that the markings on the Ram seem to have the worng discriptions

The 3D Mark tests give enough proof for that even with FSAA x2,the test results show a 160 point difference, which imho alot compared to the 32MB model, a difference too big for the same amount of ram @ default speeds.

Well I Don't think that ssstjy would cheat us on the results, I think you are being a little bit sceptical there

it still remains an unclear situation because 160 points difference with the same card @ the same speeds doesn't sound normal, that one has 32MB is allready known and that the other may have 64 would make sence for such a difference look what happens with the AGP 32MB as 64MB boards the difference between them is also about the same a bit more because they run @ true AGP x4 speeds instead of PCI heh.

If it is 32MB like you believe it is, how can we be so sure of that, then not by going to the decails on the ram chips

Well, now... hold your horses, dudes... There is no point to argue; there still are hard facts to be adressed.
First, there is no question about it: no matter of the default size of the RAM chip, if it is larger than 8Mb, may it be 16Mb, 32Mb or any higher per module, if there is no connection made from the memory controller of the GPU to the A11 line adress of the chip, the card will only see up to 8Mb present or it will not identify the memory at all. The presence of a dedicated A11 adress line trace onto the PCB or of a rework is REQUIRED for 16Mb RAM chips opperation; it is not phisicaly possible without it to function at full size. But the trace does not have to be visible at a "eyes only" inspection. It can run through an inner layer of the PCB and rise under the chip to be hooked up.
Second, if we accept that there is a chance that the memory modules soldered onto Ssstjy's Voodoo4 PCI were printed with the wrong code, but are in fact 16Mb memory modules each, then there is an easy way to test if the A11 adresses are connected to theyr designated signals at the VSA100. So easy that in fact there should be no reason for Ssstjy not to do the test by himself if his card is a real 64Mb one and he wants to confirm that fact without doubts.
If you remember, years ago, Hank Semenek had removed the GPU's out of a Voodoo5 AGP to search for the A11 connection points required for a future rework and he posted the results over the internet.
But because I was interested into modding a Voodoo4 PCI to 64Mb and nobody else researched that PCB for the A11 connection points, I decided to do it myself and I found them after un-soldering the VSA100. A self-explaining picture was posted by me on this forum. Here it is again:



For Ssstjy to find if he has onto the PCB a connection to the A11 adresses of the chips will be just to use a multimeter. The A11 line is pin 21 on TSOP2-86 memory modules, like the ones soldered on the Voodoo4's. He has to set the multimeter to measure electrical resistance -ohms- or to beeper if it has such function. Then he has to place first one cable of the multimeter to the MA_A11 point on the PCB of the card in the position marked in the picture. Next, he has to place the other cable on the pin 21 of the two chips located on the top side of the Voodoo4, relative to the PCI connector; of course not for the two chips at once, but one by one. If the multimeter beeps or displays 0 resistance then there is a trace going from the memory controller to the A11 adress. If it does not... well, then the claim for a fully working 64Mb Voodoo4 PCI cannot be valid. Here is a picture showing where the A11 pin is on the chip. For the two "top" modules it should be tested for the MA_A11 point on the PCB.



Next, he has to move one conductor of the multimeter to the MA_B11 point on the PCB and to check for connection to the A11 at the pin 21 of the two RAM chips located on the lateral side of the Voodoo4. The same consideration apply as before. Here is a picture with the two modules soldered on the lateral side of the PCB and the location of the pin 21 that should be tested for connection to the MA_B11 point that is marked on the first pic.



I hope that Ssstjy will agree to do this test to remove the justified doubts regarding his claimed 64Mb Voodoo4 PCI. I do not know what you dudes think, but I say that it cannot be more easy than so.
Good luck, Ssstjy, and get back soon with news...

Quote:


I'm not even reading this thread regularly - and in the last five minutes I followed it perfectly!

Resistance is futile!

As for the other question...

http://en.wikipedia.org/wiki/Occam's_Razor

Quote:


If you don't understand how electronics work, you should not give forth opinions on them.

The sun rises in the East- Debating it *could* rise in the West is pointless when in fact it can only rise in the East.

I have already pointed out TWICE that if the address line isn't there, then in fact the VSA-100 WILL NOT access 64 mb of ram. I'm offended I have to point this out, again to you. You seem to disregard facts when they don't fit what you wish to be true.

The PCB looks unmodifed, has a production PCB date and number and has no external line added, not to mention that the ram chips carry production markings and sizes.

Quote:


No bud, it's that I do read your posts for more than like 7 times it's that they just don't explain why the card get's more points in 3D Mark, that's all I was asking maybe you didn't understand mine properly?

So it's still odd why there is a difference in 3D Mark that part especially with FSAA x2 enabled That still hasn't been explained yet, a shame though.

Maybe the wrong size ramchips were placed during production with the wrong ram printings who knows but that might be possible that the wrong text was printed on the card's memory chips, it still remains unclear.