Unreal Engine 3 on R420 at GDC
- Thread starter SpellSinger
- Start date
- Status
Natoma said:Evildeus said:
Appears that way. He said the demo was running at 1024x768, not 640x480. That leads me to believe the rest is accurate. Unless, of course, he decides to clear it up.
The stuff he quotes covers: NV40 at 640x480; NV40 running very poorly; R420 at 1024x768; R420 running very well.
What else IS there? Just because people have chosen to focus on the resolution part of the quote, it doesn't mean he wasn't pointing out the inaccuracy in the speed interpretations too.
PaulS said:Natoma said:Evildeus said:
Appears that way. He said the demo was running at 1024x768, not 640x480. That leads me to believe the rest is accurate. Unless, of course, he decides to clear it up.
The stuff he quotes covers: NV40 at 640x480; NV40 running very poorly; R420 at 1024x768; R420 running very well.
What else IS there? Just because people have chosen to focus on the resolution part of the quote, it doesn't mean he wasn't pointing out the inaccuracy in the speed interpretations too.
No doubt you can interpret it that way. However, I took notice with the fact that he only mentioned resolution later on, and nothing else. Granted it was because SpellSinger specifically asked him, but if I were a developer, would I only speak about the resolution and not refute everything else if it were actually inaccurate?
Why would vogel come here and say "resolution" and not clear anything else up? That doesn't strike me as implying everything else was inaccurate. In effect it strikes me as saying "it was only the resolution you assumed that was inaccurate. everything else is fine."
Of course, I could be reading that incorrectly *cough* vogel clear it up *cough*.
Evildeus said:Yeah, you are right, and i really hope that you are
What other sources though? Are there any sources that have seen both the NV40 and R420 presentations, and have confirmed that it was only the former that was struggling?
Again, I've no doubt NV40 was struggling - but I need more than the original poster here (who's been proved wrong on at least the resolution aspect of his post) as proof that the R420 was flying along.
Bouncing Zabaglione Bros.
Legend
And the rumour pendulum swings back the other way!
PaulS said:Evildeus said:Yeah, you are right, and i really hope that you are
What other sources though? Are there any sources that have seen both the NV40 and R420 presentations, and have confirmed that it was only the former that was struggling?
Again, I've no doubt NV40 was struggling - but I need more than the original poster here (who's been proved wrong on at least the resolution aspect of his post) as proof that the R420 was flying along.
Epic ran the demonstration on a system with a 2GHz CPU from AMD using next-generation hardware from Nvidia (a system in another room ran the demonstration using next-generation ATI hardware). Even with that powerful silicon, the demonstration struggled at times. - Gamespot
Johnny Rotten
Regular
Good lord. Obviously Dan is under all sorts of NDA and cant comment. Im surprised we got THAT comment out of him to be honest. Trying to bait him into violating his NDA is a little weak.
could it just be that both were in the same res but the ati card didnt drop as many frames . Say it was running 20-30 fps it would look smoother than a 15-30 fps nvidia card .
Also if the r420 is just a modified r3xx card with more pipelines it would have much more stable drivers than a new nv40 card which could account for the speed diffrence .
Or of course nvidia didn't fix the 32bit fp speed in the cards
Also if the r420 is just a modified r3xx card with more pipelines it would have much more stable drivers than a new nv40 card which could account for the speed diffrence .
Or of course nvidia didn't fix the 32bit fp speed in the cards
Joe DeFuria
Legend
jvd said:
On that note, I'll make a related predcition.
When the first R420 vs. NV40 benchmark articles appear, with R420 having "expected" performance based on its architecture specs, and NV40 lacking, most web reviewers will say:
"And NV40's drivers are very immature because it's a brand spanking new super dooper advance architecture, so expect performance to improve dramatically sometime in the future, nVidia tells us."
That, instead of saying:
"Because NV40 is larger departure from its previous architecture compared to R420, R420's drivers will be much more mature and stable out of the box."
Well of course
I still feel we needed another gen of products with ati still very much in the lead and I'm really hoping that this will still happen. Because finally people were starting to see that nvidia is just a company and will do whatever it takes even if that is lieing to be on top and aren't any better than ati .
But if nvidia wins this gen all the lessons learned will be lost
I still feel we needed another gen of products with ati still very much in the lead and I'm really hoping that this will still happen. Because finally people were starting to see that nvidia is just a company and will do whatever it takes even if that is lieing to be on top and aren't any better than ati .
But if nvidia wins this gen all the lessons learned will be lost
hoom
Veteran
The apparent resolution difference noted by SpellSinger could be AA.
R3x0 6*AA is awesome & nv3x aa is decidedly not.
Based on that, you could fairly reasonably expect nv40 to have inferior aa and r420 to have even better aa (rumor is of 8*aa).
This can manifest itself in making the r420 look like its running higher res.
R3x0 6*AA is awesome & nv3x aa is decidedly not.
Based on that, you could fairly reasonably expect nv40 to have inferior aa and r420 to have even better aa (rumor is of 8*aa).
This can manifest itself in making the r420 look like its running higher res.
Everything above is complete bullshit.SpellSinger said:When visiting Epics meeting room at GDC I got the chance to see their new engine on NV40 and R420.
It looked like they were showing NV40 at 640x480 and it was very choppy with very low detail.
I was very impressed with R420 as it was obviously running with a much higher framerate and at 1024x768 or higher. The details in the models was much higher and it was very smooth. The only issue I notice were some minor artifacts which Epic put down to 24-bit FP.
It was very cool to see both parts but the R420 was by far the more impressive of the two. Epic claimed they are both equal in performance but would not show me their FPS counter. I believe this was NV marketing dollars speaking because I spent 20 minutes viewing each showing and the delta was clearly evident.
Anyone else here get the chance to see both? As far as I could tell ATI is gonna rule!!!
The ATI card looked like a 9800XT using the exact same board layout. I did not see the NVidia card but rumors around the show claimed it has two power connections, requires a 500W power supply and is extremely hot. Maybe NV should bundle a power supply with the board, instead of Doom3
I can't wait to get a chance to play with an R423 and PCI Express. Hopefully ATI will send boards out to developers soon!!!
ATI was also promoting hardware Normal Map Compression. What I heard sounded extremely cool and simple to use. No specs yet though.
This is also complete bullshit. There were no "green" computers and the only "branding" between the two systems were the AMD ones which had big fat Athlon 64 stickers on them. For the ATI cards we used systems provided directly by AMD and they were ass-kicker Athlon 64 systems built by ABS - very similar (although faster CPUs) to the great systems AMD provided for the Unreal University event. The systems we used from Nvidia were also Athlon 64 and spec'd out almost exactly the same as the systems from AMD. They came in standard off-the-shelf cooler-master silver cases. (Athlon 64 FX, lots of RAM, dual serial ATA raid hard drives, etc.). If anything the systems provided by AMD were probably faster than the ones provided by Nvidia because ABS does an incredible job tweaking their PCs for performance.SpellSinger said:
- Status
