For once: a decent Parhelia writeup...

[Dr Mordrid]

http://www.extremetech.com/article/0,3396,s=1017&a=26865,00.asp

Not too bad and rather balanced, even if they have presumed one or two things incorrectly (sorry, can't tell you which items )

Dr. Mordrid

[PXR5]

excellent read---thanks for the link
seems he uses graphics cards for more than games
and his brain for more than dumbing down press releases.
I'm surprised at how few of the reviewers so far seem to have asked for clarification or expansion on the info in the press packet.
thanks again

---e

[chaoliang]

"Considerable differences obviously remain, but Parhelia's intricate design and 80 million transistors in a .15 micron process show that GPU complexity is headed in only one direction: North. "

What does "north" mean? Can someone explain? Thanks.

[chaoliang]

Could it possibly mean the north bridge?

[JF_Aidan_Pryde]

Haha.. north as in *UP*
Future transistor count will keep rising.

[chaoliang]

Aha! These games of words. Thanks.

[moreau]

Originally posted by PXR5
seems he uses graphics cards for more than games
and his brain for more than dumbing down press releases.

[rugger]

I just noticed

It seems that for the first time in a very long time,

The performance of the memory will finally be perfectly matched to the performance of the graphics chip. (instead of typical nvidia chip speed is vastly greater than memory bandwidth)

We obviously havn't seen anything yet if this is true!

[rugger]

And this is probably why matrox simply don't need any conventional bandwidth saving technology

Damn, this is just so damned good

[rugger]

And I have my foot in my mouth again,

Please disregard my last 2 posts, they seem wrong

[Snake-Eyes]

The thought has popped into my head after reading the Parhelia tech specs the first time (and again every time I see a site complaining about it), that Parhelia probably has some bandwidth conservation features that nobody has bothered to ask about. For one thing, all those memory interfaces (the 512bit bus running throughout the chip) seem to have the potential to be MUCH more effective than even the nV crossbar is right now (and we know from various reviews that the crossbar has made a big dent when comparing GF4 and R8500's, despite the 8500 having higher theoretical bandwidth). Seems to me that this is a major point that not many have asked more information about.

Not that I expect it to be the only thing that hasn't been looked into. (Wish I knew more too, hehe. I really believe Parhelia is going to knock some socks off when it's true performance is seen in reviews).

[bsdgeek]

<i>Originally posted by Snake-Eyes<br></i>(Wish I knew more too, hehe. I really believe Parhelia is going to knock some socks off when it's true performance is seen in reviews).

Mmmm delicious performance...

[dneal]

Extremetech is a great review site. The two primary reviewers, Loyd Case and Dave Salvator, are veterans to the industry. I was so happy when I saw this site open. These guys used to write for Computer Gaming World, by far my favorite magazine (at least up to 1998 or so.) They've also done some stuff for PC Mag and Computer Shopper.

Tom's Hardware is gaggingly awful these days. It's sad that some people may actually look at the reviews posted there and rely on it. Don't get me started on his 100% BS Lightwave 7b benchmark. He uses like 1/10 of the rendering process to judge processors.

Here are some other sites that are usually pretty good with what they say about hardware.

Aces Hardware
Anandtech
Tech Report
Extremetech
RealWorldTech