R300 the fastest for DoomIII, John Carmack Speaks Again
- Thread starter multigl2
- Start date
- Status
Doomtrooper said:
Hrm, I don't really like UT as a test, as it is somewhat unreliable. But, let's do it. And I don't have RTCW or Serious Sam 2, so those aren't options.
Game options:
Renderer: my modified version of Vogel's OpenGL renderer (only difference: UseSmoothAlpha switch...if you'd like a copy, let me know)
VSYNC: Disabled
DetailTextures: Disabled
UseSmoothAlpha: Disabled
Trilinear filtering: Disabled (to make it more even, of course)
Sound: Enabled
If you'd like a full list of the graphics options, let me know.
Here we go, 1600x1200x32, no FSAA:
no aniso: 83.90 fps
8x aniso: 78.46 fps
1600x1200x32, 2x FSAA:
no aniso: 54.69 fps
8x aniso: 50.06 fps
1024x768x32, 8x aniso (This time I'll list the min/max:avg):
no FSAA: 52.83/179.67 - 87.84
2x FSAA: 53.25/176.83 - 87.50
4x FSAA: 43.20/128.92 - 76.90
Btw, I was going to do 1280x1024x32 w/ FSAA, but 4x FSAA proved to be too much for my 64MB of RAM (4x FSAA at this res uses 60MB for the framebuffer).
Lastly, a benchmark of the settings I actually play at:
(1600x1200x32, 2x FSAA, 8x aniso, trilinear and UseSmoothAlpha enabled)
Min 19.08 Max 84.13 Avg 48.83 fps.
And yes, I can do very well at these settings. In fact, I can usually win out in DM against Inhuman bots now.
demalion said:I'm a bit astounded by the resistance to the possibility of the R300 being clearly better than the NV30 by some. A lot of posts seem to be saying "it can't be the NV30 because it just couldn't be that the R300 beats it", nevermind what the text that has been quoted states. The alternative (which is, of course, possible) is that JC being sloppy and misleading in his comments...though possible, I thought automatically assuming such was the case instead of a simple interpretation of what was stated clearly was reserved for "Swedish swear word" sites.
Resistance? Dude, get your head out of your A$$. If ATI's next generation this fall blows away Nvidia's NV30, then so be it.
But he isn't talking about the NV30 at all, he does not even have the hardware in his hands, he only knows the specs of what is coming. Nvidia still has the emulation being done in software right now and the generation after the xbox is the Geforce 4.
I fail to see why you guys don't get it, are you guys really that dense?? While Nvidia was working on the xbox, ati was making a better product, this is true, why is this so hard for you to comprehend?????
The next generation after the xbox would be Geforce 4 and not NV30. If he was speaking about the NV30, he would have said the next generation beyond the Geforce 4.
The Geforce 4 and Xbox are close, but the Geforce 4 is more optimised than that of the xbox and the geforce 4 was delayed because of the xbox architecture (Geforce 4 was to come out in the fall, but it was late because of xbox), allowing ATI to move past Nvidia in the dev cycle.
This has nothing to do with the NV30, read his past coments and all of this will fit together.
The NV30 is under NDA right now (read his last coments for an understanding) and he can't say anything about it. He only knows about it, he doesn't have the actual hardware, the specs are on paper and performance is also on paper only and nothing real world.
What I find so sad is that you guys are supposed to know your stuff, but you have a severe lacking of comon sense.
Nvidia isn't the only game in town, but this NV30 performance talk is just an embarrassing missunderstanding.
That depends on what you term a 'generation' in products. I wouldn't class GF4 as a generational cycle, but a refresh of the NV20 generation architecture, much the same as GeForce2 was a refresh of GeFore256.
OpenGL guy
Veteran
Kristof said:
Uh, what are your sources for this information? Do you work at ATI? Have you signed an NDA with ATI? Have you seen the R300? I'll answer the last three for you: No, no and no.
I see now. JC hasn't worked closely enough with nvidia, obviously, that's why ATI "won" the demonstration at E3. Sounds logical to me!
I find this incredibly amusing.
I think Kristof is claiming to be a psychic.
Go on believing what you want, it'll make my job that much easier.
P.S. Your "DoomIII drivers" comment reminds me of that wacko a few months back claiming that ATI had some "extra-special-benchmark-cheating-drivers released only a few select websites".
Docwiz said:demalion said:I'm a bit astounded by the resistance to the possibility of the R300 being clearly better than the NV30 by some. A lot of posts seem to be saying "it can't be the NV30 because it just couldn't be that the R300 beats it", nevermind what the text that has been quoted states. The alternative (which is, of course, possible) is that JC being sloppy and misleading in his comments...though possible, I thought automatically assuming such was the case instead of a simple interpretation of what was stated clearly was reserved for "Swedish swear word" sites.
Resistance? Dude, get your head out of your A$$. If ATI's next generation this fall blows away Nvidia's NV30, then so be it.
This leaves me with the question...if my comment doesn't apply to you, why did you reply? If it does, how is my head in my a$$? This colors your post right from the start as pretty useless.
Nice of you to jump in after we have some clarification to try and, apparently, boost your ego. Why did you feel to pretend you know exactly what stage of development the NV30 is? Also, why are you replying to me about something I never discussed? What this makes it look like to me is you got all riled up at people "dissing" your brand and decided to spew in response without bothering to differentiate between them.
Upon further perusal, it seems you are trying to take a superior stance to the entire board, not just me? In any case, it is because of these opinions that I don't respond to the rest of your post, and instead take the opportunity to:
1) Thank Chalnoth for the clarification (which I believe, in case there is any doubt in your mind), yet point out that any of my comments and observations about the nature of the reaction in general on the board, and his/her posts, still stand, if he/she wishes to comment on them.
2) Point out to typedef (in a friendly way) that picking the right path out of two choices with your eyes closed (or blind, as the case may be) is luck, not wisdom.
That said, I'll say I'm surprised at how carelessly he made his statements in the MSNBC interview (I'm still curious if they mangled or dropped significant bits), but atleast I've learned better how far to assume in terms of clarity in quoted statements as opposed to a complete synopsis (as in a .plan or a complete message completely authored by the source).
Why the sudden spurt of trolling tone in posts here? I guess it means the site is becoming popular again...?
Hellbinder[CE said:
I love it when somebody so opinionated, egotistical and just plain patronising is proved wrong.
Of course as typedef enum points out, this does point to the nv30 being delayed, which isn't exactly good news for nVidia. I'm sure this is what certain people will concentrate on now given that other arguments seem to be lost.
Just decided to register now to post this link, but you guys beat me to it by asking Carmack himself
Check this out... a guy who looks to me like a journalist from a print magazine says that Carmack compared NV25 and R300, NV30 not being functional at the time in question.
http://www.shacknews.com/funk.y?id=3756458
This means that currently, the R300 is faster (and obviously better feature-wise) than the GF4 refresh.
It also means that R300 is quite a bit ahead of NV30. It's pretty obvious now that NV30 probably DID take a lot more refreshes than normal to get up to shape, and it still doesn't look done.
When Carmack speaks, he sure says a lot
Check this out... a guy who looks to me like a journalist from a print magazine says that Carmack compared NV25 and R300, NV30 not being functional at the time in question.
http://www.shacknews.com/funk.y?id=3756458
This means that currently, the R300 is faster (and obviously better feature-wise) than the GF4 refresh.
It also means that R300 is quite a bit ahead of NV30. It's pretty obvious now that NV30 probably DID take a lot more refreshes than normal to get up to shape, and it still doesn't look done.
When Carmack speaks, he sure says a lot
Joe DeFuria
Legend
I have to say, this does raise the question:
Will we see NV30 at all this year? Or will the fall part from nVidia be the GeForce4 refresh? I think people are starting to make the wrong arguments now: it may not be NV30 vs. R300 this fall, it may very well be the R300 vs. NV28....
Will we see NV30 at all this year? Or will the fall part from nVidia be the GeForce4 refresh? I think people are starting to make the wrong arguments now: it may not be NV30 vs. R300 this fall, it may very well be the R300 vs. NV28....
Well its not fair and its an old review I did but I do have 8500 Q3 numbers handy at those same settings:
My system
KT7A-Raid with KT764 BIOS
2 - 256 MB Kingston CAS2.5 SDRAM
AMD 1.333 Ghz Sock CPU
2 - IBM 7200 30 GB HDD in RAID 0 Array
Sound Blaster Live
Win2k w/SP2
ATI 8500 w/3286 Drivers
1600x1200x32
No Aniso
88.6
"High Aniso"
81.8
"Highest Ansio"
75.4
Again not a fair as I had a faster CPU. Also I can not equate the High Aniso to any n-Tap from nVidia. However it shows a very small drop going from no aniso to max aniso where as the GF4 Ti200 drop by a rather large amount. This data was record from the Lat Nov 2001 time frame. As we know lots changed so maybe someone can post more uptodate numbers....
My system
KT7A-Raid with KT764 BIOS
2 - 256 MB Kingston CAS2.5 SDRAM
AMD 1.333 Ghz Sock CPU
2 - IBM 7200 30 GB HDD in RAID 0 Array
Sound Blaster Live
Win2k w/SP2
ATI 8500 w/3286 Drivers
1600x1200x32
No Aniso
88.6
"High Aniso"
81.8
"Highest Ansio"
75.4
Again not a fair as I had a faster CPU. Also I can not equate the High Aniso to any n-Tap from nVidia. However it shows a very small drop going from no aniso to max aniso where as the GF4 Ti200 drop by a rather large amount. This data was record from the Lat Nov 2001 time frame. As we know lots changed so maybe someone can post more uptodate numbers....
Joe DeFuria said:
I doubt whether it's quite as severe as that, but I'd guess that the R300 will have a good head-start on any next-gen nVidia offering. At a guess? 3 months.
What ATI needs to make sure of is that they have relatively stable drivers available from the get-go when they launch the product. It's not gonna take a lot for sites to jump on their backs again if they find missing/broken features. (NB This statement in no way insinuates that NV drivers are perfect).
I don`t want to stir things up again too much, since a lot of you folks seem to get very emotional during these nvidia vs. ati debates, but imho it will be interesting to see the difference between ati and nvidia in their abillity to produce decent drivers for their respective next-gen parts this fall (if both parts actually do get released in this timeframe). Up until now, nvidia has always led ati generational releases by about half a year which obviously led to a clear driver dev. time advantage. (I bought nvidia tnt, geforce 256, geforce 3 immediately after release and I do remember lots of issues (at least with the first two) and lots of speed increases through the respective first 4 months of driver developement). So all in all i don´t expect nvidia to be lots faster this fall.
Joe DeFuria said:
Nope. When nVidias CEO recently stated that the next generation will come fall, it'll come fall. These guys don't mess around crucial statements like that. I'm sure that he only said this because nVidia at that point knew that the NV30 chip would be make it into the fall timeframe.
The high clock GF4 makes sense to me as the leaks about this product (NV25 on 0.13 ) apparently was the reason why some thought that NV30 would be delayed. In fact nVidia seems to be moving at lot to the 0.13 procent (record number of tape outs).
It just makes perfect sense. And we knew that R300 would be working before the NV30 - didn't we?
Regarding how the R300 and NV30 will compete... I have no idea of course.
Well, they are leaving it a little late, if there really are no working NV30's (to a reasonable level) about - remember dev's had NV20's 6 months before they were released.
Joe DeFuria
Legend
Well, a 3 month head start in the environment of "6 month" product cycles can be significant.
The problem for nVidia as I see it, is that if they don't have working silicon at this point, they are really cutting it close if they wish to have shipping silicon in any significant quantity by Sept / October. nVidia has been very "strict" with their product launches over the past years: always "launching" something in late August/Early September, and then always launching something around March. Key word there: something.
It would be very odd (but not completely unreasonable) for nVidia to push back a NV30 launch 3 months to say Nov / December. (Shipping quantity in say Dec / January). IMO, if NV30 is not ready for launch in September (Shipping Oct/Nov), nVidia will launch the "Super Dooper GeForce4 ti xl plus ultra" this fall, and hold off on the NV30 until the Spring.
Another alternative would be to launch an extremely limited supply of NV30 chips this fall. (Assuming that it is the production end that is having problems, not chip design probs.) This would be more to save face as the market leader than anything else. Such a move would probably piss off many vendors though, who have already been complaining about supply and distribution issues from what I've heard.
The good news for nVidia, is that it appears Doom3 is a bit further off than any of us anticipated. I'm not expecting it until the latter half of '03. If Doom3 were to come out this fall, without Nv30 on the market, that would be a major coup for ATI. As it stands, we will probably be looking at the successors to R300 and NV30 as the "top of the line" cards available when Doom3 ships.
I agree with the comments here about drivers. The R300 and NV30 represent "new cores" for both ATI and nVidia, and will finally show a more level playing field on which to judge the respective driver teams. I do expect both parts to have more issues than we're used to seeing when it comes to firing up a variety of older games.
The problem for nVidia as I see it, is that if they don't have working silicon at this point, they are really cutting it close if they wish to have shipping silicon in any significant quantity by Sept / October. nVidia has been very "strict" with their product launches over the past years: always "launching" something in late August/Early September, and then always launching something around March. Key word there: something.
It would be very odd (but not completely unreasonable) for nVidia to push back a NV30 launch 3 months to say Nov / December. (Shipping quantity in say Dec / January). IMO, if NV30 is not ready for launch in September (Shipping Oct/Nov), nVidia will launch the "Super Dooper GeForce4 ti xl plus ultra" this fall, and hold off on the NV30 until the Spring.
Another alternative would be to launch an extremely limited supply of NV30 chips this fall. (Assuming that it is the production end that is having problems, not chip design probs.) This would be more to save face as the market leader than anything else. Such a move would probably piss off many vendors though, who have already been complaining about supply and distribution issues from what I've heard.
The good news for nVidia, is that it appears Doom3 is a bit further off than any of us anticipated. I'm not expecting it until the latter half of '03. If Doom3 were to come out this fall, without Nv30 on the market, that would be a major coup for ATI. As it stands, we will probably be looking at the successors to R300 and NV30 as the "top of the line" cards available when Doom3 ships.
I agree with the comments here about drivers. The R300 and NV30 represent "new cores" for both ATI and nVidia, and will finally show a more level playing field on which to judge the respective driver teams. I do expect both parts to have more issues than we're used to seeing when it comes to firing up a variety of older games.
Joe DeFuria
Legend
They don't INTENTIONALLY mess around, but they also cannot predict the future. Givin the fact that NV30 is not up and running at the moment (let alone when those statements were made what, a month or two ago?), there is simply no way to know for sure that they CAN ship this fall. The CEO can have a best guess, but he cannot guarantee anything.
That CEO comment was from some interview, not an official PR.
True.Joe DeFuria said:
Not true.Joe DeFuria said:
The statements were made in the quarterly conference call just this week, which is much more official than any PR. The SEC gets very angry when misleading statements are made as part of a financial statement of a publicly traded company.
Also, we have no proof that NV30 is or is not up and running at the moment. The silicon may be back from the fab or it may not be. The drivers may be up to snuff or not. All we know is that at E3, it wasn't the demo platform for Doom3, though we can assume it was not for good reason.
We could interpret Carmacks statements plenty of ways, but to draw conclusions from them and base future conclusions on them is treading on thin ice.
Typedef Enum
Regular
Let me clarify why I said...err...what I said (not that it really makes much difference)...In some ways it was a guess...but I was basing it on what I would call common-sense/logic.
Joe DeFuria
Legend
I'm pretty sure the "most significant contribution" statements were made earlier.
EDIT: here's the link about that statement, MAY 1:
http://news.com.com/2100-1040-896850.html?legacy=cnet&tag=lthd
In any case, this is what I dug up on that conference call:
So yes, as of the conference call, the "PLAN' is to have NV30 out this fall. THis is no surprise. However, note that they are telling the story that they are pretty much DEPDENDENT on TSMC's 0.13 process being up to scratch with their new part. That is a big IF.
Of course, and there is no indication that this is a misleading statement, even if it turns out NV30 isn't ready this fall. I'm sure you are well aware of the "Forward looking statements" disclosure that is always part of these shareholder calls.
BTW, is there anywhere in that conference call where nVidia states that their planned and previously announced nForce 615/620 chips were cancelled? You would think THAT would be an important disclosure for such a call, wouldn't you?
Well, that's good. Because nVidia has PRs about the 615/620 chipset product, but has yet to have any PR or other official disclose about it's cancellation from what I can see.
No one is drawing conclusions here, Russ...we're all speculating. And my speculation tells me that NV30 "might" launch this August, but if it does, we'll see extremely limited quantities. And it would NOT surprise me if it were pushed off to spring.
I think that we can be certain that at least as of last week, it was not up and running to any appreciable extent. SURELY, nVidia would tried to get the Doom3 demo on that part. The fact that nVidia was apparently trying to wring out as much performance on a GeForce4 variant, should clue us all in to the status of NV30 at the moment.
- Status