Anandtech uses UPT on all current cards

[Reckless]

Apologies to Ant if this is not the correct forum (h/w, gaming & benchmarks)

Anandtech has updated the Unreal Performance Benchmark to reflect the latest version of the Unreal engine and tested against all current video cards. The Parhelia gets some reasonable review words but ultimately fails to deliver enough for the street price (which incidentely, is what I see as the problem is as well)

http://www.anandtech.com/video/showdoc.html?i=1647&p=1

The results show that the speed is OK for the forthcoming iteration of the Unreal engine but it doesn't provide a lot of headroom for future game releases. A cost price of $400 equates to a significant investment and thus it needs some form of in-built longevity (18-24 months min). It remains to be seen whether they are able to make that sort of improvement through driver updates (unlikely).

[nggalai]

Correction: Anand didn't use the UPT, but the latest build of the UT2003 Demo (28-06-2002). HUGE difference.

ta,
-Sascha.rb

[Ant]

I presume this is still a test that only he has got his grubby little hands on?

[nggalai]

Originally posted by Ant
I presume this is still a test that only he has got his grubby little hands on?

Jup--one of the perks of living some 30 minutes' drive off Epic's HQ, I presume.

ta,
.rb

[Stringy]

Gotta love his little "Memory Bench" test... LOL
I didn't know they have a 64Bit Bus P and ATi 8500 ?!?!

Looks awful slanted to these eyes...

the rest of the benches look ok... and do the GF4 series really perform... would have liked to see some FAA and FFSA in there, but that might not show the gap in performance he was going for...



Craig

[Ryu Connor]

$400 equates to a significant investment and thus it needs some form of in-built longevity (18-24 months min)

Personally, I think the days of video cards still being competitive in gaming 18-24 months away - especially at full detail - are over.

I think a year is really all you have anymore.

[mikeul]

Actually, lots of reviewers have this demo in their hands - they just have to write to Mark Rein like I did - but they need to get hands on a Parhelia first or find which parhelia drivers are good enough to review on.

[103er-Fan]

Originally posted by Ryu Connor


Personally, I think the days of video cards still being competitive in gaming 18-24 months away - especially at full detail - are over.

I think a year is really all you have anymore.

I have to agree here. I you want to buy a graphics card which offers maximum speed at full details over the next two years, you can forget it.
But if you just want to play games and don't care about the detail settings, then you can still have some hopes there. The v5 5500 for example is 2 years old, and it still plays the most things fine, you just have to reduce some detail settings here and there. Unfortunatly you'll run into driver problems with this card nowadays

[Cheesekeeper]

At least the Parhelia was true to its word - it did manage to maintain playable framerates in almost every test, but as usual, the GF4 was holding a major lead. Once some benchmarks based on Doom3 come out, that will give us a much better idea of how much life the Parhelia has in it. The tests involving large (ie 8+) light sources seem to be one of the Parhelia's biggest strengths - the advanced lighting in Doom 3 may tip the scales a bit more towards the Parhelia. I suppose a lot will also ride on just what ATI have up their sleeves (since Carmack seems rather keen on them)........

[Ryu Connor]

Once some benchmarks based on Doom3 come out, that will give us a much better idea of how much life the Parhelia has in it.

Since Doom III is about eighteen to twenty-four months away, you'll probably want something other than Parhelia to run it. Either a refresh from Matrox or a card from a different vendor.

[Electric Amish]

Originally posted by Ryu Connor


Since Doom III is about eighteen to twenty-four months away, you'll probably want something other than Parhelia to run it. Either a refresh from Matrox or a card from a different vendor.

I agree. You shouldn't be looking for a card now to play a game that comes out in over a year.

I wish Anand had thrown some tests in there with Aniso and AA.

amish

[thop]

The Parhelia gets beaten by the 8500 and even the GF3 in most tests (edit: P wins in High Detail, the others in Medium). However UT2k3 still seems to be very playable with the P at higher resolutions.
But why has the P so inredibly low memory bandwith? sth. not right here.

[nggalai]

Originally posted by thop
The Parhelia gets beaten by the 8500 and even the GF3 in most tests (edit: P wins in High Detail, the others in Medium). However UT2k3 still seems to be very playable with the P at higher resolutions.

Exactly--Parhelia seems to be more than fast enough to play UT2003 at the highest in-game settings. And that's what counts.

Bummer though it seems to be immensely crippled regarding bandwidth usage. Let's hope that it's a drivers issue.

ta,
-Sascha.rb

P.S. hey, you might be right about fans on MURC and 3DC. .rb

[typedef enum]

I think my worst fear of Parhelia has seemingly come true...In essence, I'm reconfirming what John Carmack has already stated: going into it, I though the Parhelia would be the final nail on the necessity for extreme bandwdith...a-la Bit Boys. Instead, we're seeing a bit of reality, combined with immature drivers.

That is to say, all the R&D that nVidia has put into increasing their bandwidth efficiency is really showing itself. For the life of me, I cannot fathom just how badly Parhelia gets nailed on this bandwidth test...But, we also know that nVidia has been tinkering with the whole Memory-Controller/occlusion/Z-Optimizations for a good deal of time.

[Parhelia Pro256]

No, you're wrong