I sure HOPE they're available right after the launch (non-paper, if we're lucky
). Maybe not the next day, but within a matter of weeks, not months. I also hope they're available at MSRP and not price gouged (since the MSRP is already high enough as it is)....But what are the chances of that?
Edit: Though the reports of G70 already being in production do give me some hope. We just don't want a repeat of last year's madness.
biostud
Lifer
- Feb 27, 2003
- 19,951
- 7,049
- 136
Personally I'm not going to decide which yo buy before I know the specs and pricing of the cards. So even if nVidia get a head start in launch date, I will still wait to see what the r520 has to offer before buying. I would guess that most are going to do the same. So even if nvidia might get their card launched first, it'll probably not going to help them in selling lots of g70, before the r520 has been benched. This ofcourse depends on how large the speed margin is and how long we'll have to wait, before the cards of course is in stock.
I agree. If Nvidia releases a 24-pipe GTX and then (within a few months) ATI released a 32-pipe card and Nvidia has to bring out a 32-pipe Ultra to compete...then it will be better to have waited for the better card, assuming they're not charging $1200 for it - and assuming you don't NEED a new card immediately, obviously. (Since I already have a 6800GT and therefore don't need to upgrade too badly, might as well make the most of the upgrade when I get to that point)
Gamingphreek
Lifer
- Mar 31, 2003
- 11,679
- 0
- 81
Originally posted by: Pete
Gamingphreek, you're wrong about the OP's quote being biased one way or the other, you're wrong about nV not having released a "pipeline deficient" card in the past, you're wrong about the the FX's "microarchitecture" and compilers being the reason it lags so far behind in DX9 games, you're wrong about nV using FP32 with the FX in most games, you're wrong about nV always having had the edge in AA, you're wrong about nV having had worse AF until the GF6.
You're just wrong on so many counts, it'd be kind of incredible if you weren't doing this on purpose. My suggestion--at least with respect to nVidia's recent history and this thread in particular--is to post less and read more.
Since when is it biased to think ATI can cram more pipelines in with a smaller process than nVidia? Seems like common sense. Also seems like common sense that they may have yield problems with this new process, at least initially.
Surely synthetic benchmarks numbers and the 3DM03 fiasco taught you that nV must/will use FP16 precision to maintain reasonable performance with their cards?
Surely nV worked their butts off to optimise "DX9" games to their FX architecture. Why is it that Far Cry and HL2 and the like run so much worse on the FX series? In fact, why does the four-pipe FX series run about as fast as the four-pipe 9600 series in those games? It's possible game devs are inept and don't want to maximize FX owners' IQ with their cards. Or it's possible the FX just has some (micro)architectural flaws that can't be reasonably expected to be overcome in the real world of deadlines and "cross-platform" development.
ATI seriously debuted AF with the 8500 *after* nV did with the GF3. ATI's AF had almost no performance hit--compared to the GF3's rather large one--because ATI just used bilinear filtering and angle-dependency, while nV used trilinear filtering and "fully" filtered all angles. nV also went above and beyond with its filtering quality, whereas ATI stuck to the D3D bare minimum. 3DCenter had a good article on this, IIRC, which you should search for.
nV basically had "average" AA until the GF6 series, as did ATI until the 9700 series. 3dfx was first out of the gate with excellent (but at a huge hit) AA with the Voodoo 4/5 series. The 9700 moved ATI from supersample to multisample AA, which improved performance; adapted a jittered grid, which improved quality; integrated gamma-"correction", which also improved quality; and allowed for up to three passes, which allowed for higher max quality at a reasonable speed. The GF3 moved to MSAA, but kept the ordered grid for 4x, so it wasn't that hot (especially compared to 3dfx's pseudo-temporal rotated grid). The FX merely improved speed. The GF6 finally brought rotated grid to both 2x AND 4x, closing the gap to ATI's 2x and 4x modes considerably. Again, 3DCenter had a good article on this, IIRC, which you should search for, but most initial 9700P and 6800U p/reviews should have concise write-ups on their respectively improved AA.
No disrespect intended, but so much of what you posted is wrong. It's been pointed out by other people, but maybe spelling it out in detail will clarify your errors. Don't take it personally, just learn and help the next guy out, like the rest of us try to.
Edit: My first two paragraphs seem more inflammatory than I mean them to be. I'll leave my post as is, just know that I didn't mean to come across so, well, mean. We all have to learn somehow, and I hope you learned more from my post than that I'm easily exasperated.
Maybe i confused the AA and AF scenarios with this generation. I know that this generation however the AF was fixed however still suffers a large performance hit due to the loss of an ALU.
I believe there was a thread on this not too long ago (made by mean)... something like "Whats wrong with Nvidia". In that thread that is where i first saw this information i posted.
-Kevin
Ronin
Diamond Member
For those that were interested in seeing what it looks like in a box (as I said, since the images are already out there, I am only providing these based on what you have already seen).
http://server.counter-strike.net/images/misc/exteriorflash.jpg
http://server.counter-strike.net/images/misc/exteriornoflash.jpg
http://server.counter-strike.net/images/misc/interiorflash.jpg
http://server.counter-strike.net/images/misc/interiorflash2.jpg
http://server.counter-strike.net/images/misc/interiornoflash1.jpg
Take them for what they're worth.
http://server.counter-strike.net/images/misc/exteriorflash.jpg
http://server.counter-strike.net/images/misc/exteriornoflash.jpg
http://server.counter-strike.net/images/misc/interiorflash.jpg
http://server.counter-strike.net/images/misc/interiorflash2.jpg
http://server.counter-strike.net/images/misc/interiornoflash1.jpg
Take them for what they're worth.
This is just my take on it. So don't 'prove me wrong'. I never said these were facts. This is just my prediction on what will happen.
ATI will have trouble with its new process and 32 pipes, so should a 24-piped G70 come ahead, ATI will be in trouble because of an inferior architecture. The next-next generation, the situation will be in reverse because ATI already has a 90nm part.
NVIDIA thinks they can get ahead by releasing first, and ATI thinks it's better to wait and release a better card later.
NVIDIA's card will show up a month after its 6/21 "launch", at MSRP (or remotely reasonable) prices. Before that month later, the cards will all be on pre-order at seriously inflated prices, with just a couple available. Since NVIDIA already started mass-producing them, they are preparing for the worst. So I think this coming generation will still have a paper launch, just not nearly as bad. The exception will be the G70 GTX, which would be analogous to a 6800 Ultra Express for a couple of months.
(All these are single-slot.)
Month after 6/21 "launch":
7800 GTX PCI-E (G70, 24 pipes) - $700+
7800 GT PCI-E (G70, 24 pipes) - $550
7800 PCI-E (G70, 20 pipes, unlockable) - $425
Another month thereafter:
7800 GTX PCI-E (G70, 24 pipes) - $600+
7800 GT PCI-E (G70, 24 pipes) - $485
7800 PCI-E (G70, 20 pipes, unlockable) - $400
Later on, 4 months after the 7800 is announced, a 7600 will be announced, and a month later:
7600 GTX PCI-E (G72, 16 pipes) - $325
7600 PCI-E (G73, 12 pipes, not unlockable) - $275
Hey, I can even make up fake codenames.
Note that I'm a pessimist. Let's hope my predictions don't come true.
ATI will have trouble with its new process and 32 pipes, so should a 24-piped G70 come ahead, ATI will be in trouble because of an inferior architecture. The next-next generation, the situation will be in reverse because ATI already has a 90nm part.
NVIDIA thinks they can get ahead by releasing first, and ATI thinks it's better to wait and release a better card later.
NVIDIA's card will show up a month after its 6/21 "launch", at MSRP (or remotely reasonable) prices. Before that month later, the cards will all be on pre-order at seriously inflated prices, with just a couple available. Since NVIDIA already started mass-producing them, they are preparing for the worst. So I think this coming generation will still have a paper launch, just not nearly as bad. The exception will be the G70 GTX, which would be analogous to a 6800 Ultra Express for a couple of months.
(All these are single-slot.)
Month after 6/21 "launch":
7800 GTX PCI-E (G70, 24 pipes) - $700+
7800 GT PCI-E (G70, 24 pipes) - $550
7800 PCI-E (G70, 20 pipes, unlockable) - $425
Another month thereafter:
7800 GTX PCI-E (G70, 24 pipes) - $600+
7800 GT PCI-E (G70, 24 pipes) - $485
7800 PCI-E (G70, 20 pipes, unlockable) - $400
Later on, 4 months after the 7800 is announced, a 7600 will be announced, and a month later:
7600 GTX PCI-E (G72, 16 pipes) - $325
7600 PCI-E (G73, 12 pipes, not unlockable) - $275
Hey, I can even make up fake codenames.
Note that I'm a pessimist. Let's hope my predictions don't come true.
Ronin
Diamond Member
Conjecture?
It's highly possible you won't see one. I'd venture it's going to depend on the adoption rate with this release (and with the PCIe variants of the 6xxx cards), but if you do see any AGP releases, it will probably be only one card from either/both companies, and not much more.
It's highly possible you won't see one. I'd venture it's going to depend on the adoption rate with this release (and with the PCIe variants of the 6xxx cards), but if you do see any AGP releases, it will probably be only one card from either/both companies, and not much more.
I just bought the ATI x800xl! And I don't even have the rest of the parts for my computer. I'll probably be ordering the other stuff this month. So... should I sell the ATI card that I haven't even used yet in preperation to buy a new nVidia card? How much will the G70 debut at?
Gamingphreek
Lifer
- Mar 31, 2003
- 11,679
- 0
- 81
On the last image, what chip is the top one. I assue the bottom is the G70, however what is the top one. I cant make out the name on the core but it looks like 6600GT or at least something GT.
-Kevin
Ronin
Diamond Member
Gamingphreek
Lifer
- Mar 31, 2003
- 11,679
- 0
- 81
Oh that would make sense. If the game isn't heavily stressing the shaders it isn't really stressing the ALU's and can afford, without much performance loss, to lose one. Thanks!
I can't make any predictions for R520 and am not really interested in doing so. I will note that I am reasonably sure that R520 won't be based too heavily on the old R300 design (fanboys and ATi engineers have suggested all that is needed is to bolt-on SM3.0 to the existing design, but it isn't that simple IMO. ATi's current GPU's lack many many things that are mandatory for SM3.0 like branching/looping at take many hardware shortcuts. I think the entire pipeline has had to be redesigned as a consequence and this is the main reason behind why R520 has taken so long to appear).
On nVidia's side, my predictions are that we will see a full gamut of filtering and AA options for the floating point rendering (OpenEXR) modes.
We should also see see a performance increase in shader operations due to further improvement in the onchip resource scheduling (this was a major factor in NV3x performing as poorly as it did and was corrected for nV40 allowing the driver JIT compiler to extract the most performance possible from the chip each cycle). A modest increase in registers could yield significant performance gains. nV40 does not have many more registers per pipe than NV3x did.
Other things that could possibly happen (but I don't necessarily expect to happen until the 90 nm version of G70 hits - if at all) would include adding TMU capabilities to the second large ALU in each pipeline and making both large ALU's capable of performing all math operations. It's also possible that 128 bit floating point rendering will be introduced (the playstation3 GPU features it), but I don't think G70 will be fast enough until the 90nm refresh personally.
On nVidia's side, my predictions are that we will see a full gamut of filtering and AA options for the floating point rendering (OpenEXR) modes.
We should also see see a performance increase in shader operations due to further improvement in the onchip resource scheduling (this was a major factor in NV3x performing as poorly as it did and was corrected for nV40 allowing the driver JIT compiler to extract the most performance possible from the chip each cycle). A modest increase in registers could yield significant performance gains. nV40 does not have many more registers per pipe than NV3x did.
Other things that could possibly happen (but I don't necessarily expect to happen until the 90 nm version of G70 hits - if at all) would include adding TMU capabilities to the second large ALU in each pipeline and making both large ALU's capable of performing all math operations. It's also possible that 128 bit floating point rendering will be introduced (the playstation3 GPU features it), but I don't think G70 will be fast enough until the 90nm refresh personally.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter