Parhelia : Reference board clocked higher?

[stmok]

I sense a disturbance in the Force...

Remember the initial test Parhelia cards (Reference boards)?
Do you remember the size of the heatsink and fan?
(It was quite large)

And wasn't it the same card used in initial testing which resulted in details being published by an Asian source? You know, the "faster than the GF4 Ti4600 by 20% in 3DMark2001".
(That's what got everyone thinking...GF4 Killer)

Clearly, that is a higher clocked version. You don't stick a massive HSF on a reference card to make it look good.

Now note the retail/bulk version...Look at the size of heatsink and fan.

Why did Matrox cut back on speed?

Was it too unstable?

What if a reviewer got their hands on this particular reference board? (How would this change?)

[MrGaribaldi]

stmok> it has been discussed quite a few times before...
haig told us that they clocked the card down due to heat problems...

they could've kept the core speed as it was, but then they'd need an hsf which'd violate AGP specs... And for some strange reason, they didn't want to violate those specs...

[Cobos]

But what is a good point is why not Matrox spent a little bit more on getting a better heatsink, which still doesn't violate specs... Like a bit larger copper thingy... covering also the ramchips, not because you need to overclock the ram but because you then get a serious area covered by a heatsink. That way they might have upped the card another 10-20Mhz perhaps and then easier sold it at the 399 pricepoint.

Cobos

[MrGaribaldi]

Originally posted by Cobos
But what is a good point is why not Matrox spent a little bit more on getting a better heatsink, which still doesn't violate specs... Like a bit larger copper thingy... covering also the ramchips, not because you need to overclock the ram but because you then get a serious area covered by a heatsink. That way they might have upped the card another 10-20Mhz perhaps and then easier sold it at the 399 pricepoint.

That's a very good point...
If you add in a silver heatspreader, you should've been able to hit even higher speeds, without increasing the cost too much...

Silver is expensive, but with the amount needed for each hs being small...

Well, at least we'll be able to hit 240mhz without too much trouble

Come to think of it, it looks like Matrox tried to save as much as possible on the hsf they've used... Espescially if you take a look at what LeadTek is offering with their cards...

Cobos, do you have time to be on this board now, with Arcon so close?

[Cheesekeeper]

Take a look at this thread:

http://forums.murc.ws/showthread.php?s=&threadid=34090&perpage=15&pagenumber=3

We've been discussing some alternatives here. Haig has been nice enough to confirm for us that by improving the cooling, the Parhelia chip should happily run 20MHz higher. He also said that to run it any faster than that would require increasing the chip voltage........

[Greebe]

That's a very good point...
If you add in a silver heat spreader, you should've been able to hit even higher speeds, without increasing the cost too much...

And what would a silver heat spreader have over the copper one that is already on P?!

Effectively nothing... maybe 0.1 C/W improvement if that.

[Jon P. Inghram]

Matrox is just plain cheap! Everyone knows diamond is better than copper or silver, so why didn't they use it instead?

[MrGaribaldi]

Originally posted by Greebe
And what would a silver heat spreader have over the copper one that is already on P?!
Effectively nothing... maybe 0.1 C/W improvement if that.

Depending on how you make the hs... If you were going to for a hughe flat hs, it thin layer of silver would help spread the heat throughout the entire hs...

but if you're only going for a regular cooler, then sure, it wouldn't have any noticeable effect...

.02$

[Yannick]

Originally posted by Jon P. Inghram
Matrox is just plain cheap! Everyone knows diamond is better than copper or silver, so why didn't they use it instead?

Despite your smiley, I don't think that this idea is so strange. I don't know what the industrial prices are, but I think that artifical diamond coating is not unreachable in terms of cost. Still, I don't know the order of magnitude of the price, and even then it would only serve for better thermal coupling between chip and heatsink...

[Greebe]

Nasa is experimenting with Diamond for use in semiconductor/substrate design seeing that it can handle temps upto something like 500c verses sillycon's rather limited 100c top end.

This is much more important for use in space seeing that the extremes are like from +/- 250c

While still experimental, someday once it's viable we'll see it trickle down into mainstream designs (barring cost limitations).

[[Ch]amsalot]

Greebe,

I thought NASA was still hunting for old 286's

-[Ch]ams

[Oboy]

Originally posted by [Ch]amsalot
Greebe,

I thought NASA was still hunting for old 286's

-[Ch]ams

If there RAD Hard.

Oboy