Parhelia projected clockspeed

[xortam]

Originally posted by Jammrock
I think all you people are smoking crack and not sharing

Jammrock

What do you base that opinion on Jammy?

O.K. Ben, you've got me interested enough to go see what you've got to say in your report ...

[Pace]

Ben: I got to the site ok, but my end (modem) is very slow! I'll check it tomorrow from Uni.

P.

[ben2207]

Let me be clear on something. Matrox PR initiated the contact, I had asked the clockspeeds at E3 , but they wouldn't be pinned down. After my report was posted on the E3 meeting with Matrox, they sent me a clarification on the clockspeeds. Pace you might try http://www.pcrave.com as that site is not clogged with hits....

[Indiana]

To end this, here is the (mostly very positive) post made by Ben, so noone can claim he couldn't read it:

Matrox Meeting Notes. No benchmarks but - Benjamin Sun
I'm not going to gvie any guesses on performance , as the boards I played on this morning are beta. However here's what I got from my meeting (excluding anything that might be considered um off the record)

1. Displacement mapping absolutely rocks . Adding arbitrary geometry to figures or in outdoors terrain is a serious step in the right direction.
2. FAA absolutely also rocks. Swimming textures in buildings in FS 2002? GONE.
3. 10bit color was a little bit of a disappointment. Because the LCD monitors that were used in the demo, it downsampled to 8bit color. While there is a difference in banding, it was less apparent than I expected. When I test the card with 3 CRTs I'll take a deeper look into it.
4. It will not be a Quake3 Geforce4 Ti killer with the way most reviewers run the benchmarks. Of course with 80 million transistors on a .15um process, there's likely to be a lot of heat, and it's unlikely that the Parhelia 512 will have a more than 300mhz gclock.
5. Having said that, Matrox suggests, and I agree with them 100% that performance in Quake3 is fast enough today. Turn on FSAA (games with stencil buffers have issues with FAA) anisotropic filtering and Surround Gaming , and it should be completely playable
6.Surround gaming absolutely rocks. I got to play Quake3 and on 3 monitors it was absoletly smooth. One little note, while playing , I didn't consciously look to my right or left, but in my peripheral vision caught movement on the extra monitors and I was able to react without thinking
7. Matrox suggests, and after walking around the show floor, I agree, that DX 8 games will be more prevalent this Christmas. The number of DX9 games will be very limited if any at all. Thus their decision to go with DX8.1
8. If you have a chance and are at E3 go to the AMD/Matrox/Alienware lot (last year's G.O.D) lot , take it. Soldier Of Fortune II and F1 2002 makes great use of surround gaming . The HUGE monitors for the racing game are the way to play a racing game.
9. Parhelia 512 can multipass render . 8 textures per pixel is possible on a Parhelia
10. Review cards are likely to be sent out in early June. WOOHOO!!! Final retail boards will likely be out in late June/early July.
11. Did I mention Displacement Mapping totally rocks ??? Watching a character be N-Patched , then displacement mapped, makes a totally different look
12. To give you a sense of how powerful the pixel shaders are capable of being. The ocean demo with a shark, the bass, a school of fish? Everything is bumpmapped and using 4 textures per pixel. The Bass consists of 8 pixel shaders with 100 pixel shader instructions.
13. The anisotropic filtering did sharpen Quake3. The Parhelia is capable of dynamically allocating 64 texture samples in a pass . In other words 1 pixel with 64 sample anisotropic filtering, or alternatively 4 pixels with 16 sample aniso in a single clock.
14 Performance hit with 16x FAA in one application (no numbers, sorry) was about 20% with their 4x FSAA around 50% . Sorry , can't say more.
15Only one mode of FAA 16x at the moment.

[xortam]

I've seen those meeting notes posted here before and I would say that those comments seem reasonable.

P.S. Those comments were posted from Ben's Beyond3D post as reported in this thread.

[ben2207]

darn you Indiana now I have to come up with more unique content for people to click my site!!!

[cjolley]

ben2207,
The reason for so much skepticism is that Matrox is so quiet about clock speeds that they didn't even put them on their G400 overclock utility.
I don't think they have ever said an official word word about clock speeds for any chip before.
I wonder if someone is getting chewed out even now about your email.
chuck

[Indiana]

I just wanted to lower the amount of all those MURCers accessing your site and crashing you server

[ben2207]

No , xortam, I posted simultaneously on PCRave, Beyond3d and NVNews LOL . Almost 5000 hits on one little thread on Beyond3d

[cjolley]

PS Can you imagine the OC wonder this thing might be if the only thing holding it back from a 50% OC is heat?
cwj

[Kosh Naranek]

Originally posted by xortam
ben2207 informed me that Matrox had to trim the P down to a 256-bit GPU due to heat problems. Plans are to ship the 512-bit version when they go to 130 nm later this summer.

I simply don't believe this !!!

BUT if it's true then Matrox made a VERY BIG mistake in claming the worlds first 512 bit GPU.

And I wouldn't be buying a Parhelia any time soon™

[ben2207]

Kosh uh I never said anything of the sort...

[xortam]

Note the emoticon that I used Kosh.

[djroberts]

OH MY GAWD !!!!

They are cutting the chip in half!!!!

Were only gettin half the Paranoia!!!!

Matrox PR said were only getting Half the Chip at 220 MHurtz!!!!


OH MY GAWD !!!!

dj

[xortam]

Originally posted by ben2207
No , xortam, I posted simultaneously on PCRave, Beyond3d and NVNews LOL . Almost 5000 hits on one little thread on Beyond3d

I was just pointing out that this was the source for the first post here concerning your meeting notes. There are some interesting discussions in that Beyond3D post (still reading it through) and I noticed several MURCers had posted there.