schneiderguy
Lifer
- Jun 26, 2006
- 10,801
- 91
- 91
Why not? The rv670 die is way smaller than G92 and a bit smaller than G94 so ATI is making more money off of each card sold than nvidia
Too bad that they don't sell these cards well; AMD has 25% market share of discrete graphics cards. They reported one week ago that their sales on each and every one of their businesses have been smaller than expected.
According to that link, ATi had a ~50% edge in marketshare over nV in Q4 '04, now nV has a 50% edge over ATi. That is very close to a complete collapse.
In realistic terms, ATi's highest end GPU, its' most bleeding edge part released can not compete with 2 year old nV technology. They very well may make a comeback, but the industry hasn't seen this kind of utter domination since the days of 3Dfx.
In realistic terms, ATi's highest end GPU, its' most bleeding edge part released can not compete with 2 year old nV technology. They very well may make a comeback, but the industry hasn't seen this kind of utter domination since the days of 3Dfx.
Sylvanas
Diamond Member
- Jan 20, 2004
- 3,752
- 0
- 0
If by 2 year old technology you are referring to the G80, it's not meant to compete with it....the 3800 series are competitive on price- if you want a single card with better performance than a G80 (GTX) then thats the 3870X2- which is in the same price bracket. Note the market share difference in Q4 2004 and Q4 2007 between the two is about the same as it was evident back in the 9700pro days, so we have seen it before and will probably see it again.
I just hope something new comes out from either ATI or NVIDIA this year that is substantially faster than my overclocked 8800GTX. Aside from dual-GPU solutions, there doesn't seem to be a big difference in performance - at least, not enough to justify a new card. It's so strange that a couple of years ago, I was left with an underperforming card and no money to buy the new hotness. Now I've got a very good card and the money to buy something better, but no such product exists. C'mon ATI and NVIDIA, I actually *want* to give you my money for a significantly better GPU.
apoppin
Lifer
No .. wishful thinking .. even though it is more elegant perhaps than Gx2, 3870x2 is still a compromise "sandwich card" .. and with the release of Gx2 , you need two x2s .. which is no real "solution"
and if we see r700 launch in May-June, we see GT200 grab AMD by the throat and again with GT100 [or G94 shrink] on the bottom
that leaves AMD to compete on "price"
how nice
i would not like to be in AMD's position - not IF i had a choice and not UNLESS r700 is a monster which is doubtful
the *only thing* that may save AMD is that Intel is evidently not "finishing them off" .. the DeathStar after giving what they thought is a Mortal Blow to their Old Enemy, has Surprisingly Broken Off Engagement with AMD .. and has moved off on a Wild Tangent [in their Supreme Arrogance] to turn their PR Ray-GunTracing Disrupters on NVIDIA ...
NVIDIA's beating AMD Graphics will not Kill AMD .. Intel can only really do that .. and they have somehow decided NVIDIA is more of a threat.
Interesting times
Sylvanas:
That was Q4/2007. AMD has already warned that it won't get good results in Q1 since it's sales have been dissapointing on all fronts.
---
About HD3870 X2:
It's around 65% more expensive than 8800 GTS 512. It's also 28% more expensive than 2x9600 GT which already wins HD3870 X2. HD3870 X2 is, at least in Europe 40% more expensive than 8800 GTX; performance difference depends on do you cherish more timedemo or game performance.
That was Q4/2007. AMD has already warned that it won't get good results in Q1 since it's sales have been dissapointing on all fronts.
---
About HD3870 X2:
It's around 65% more expensive than 8800 GTS 512. It's also 28% more expensive than 2x9600 GT which already wins HD3870 X2. HD3870 X2 is, at least in Europe 40% more expensive than 8800 GTX; performance difference depends on do you cherish more timedemo or game performance.
Sylvanas
Diamond Member
- Jan 20, 2004
- 3,752
- 0
- 0
That depends on what you subjectively classify as a 'solution'. Yes, some people don't like the concept of SLI/CF on a single card and don't think it a *real* *solution* but the fact I can plug it in and get higher frame rates than a single GPU (in the vast majority of games...emphasis on majority) is enough for me to deem it as a 'solution' to a problem of getting better performance. (insert 'but theres input lag' comment here....I haven't noticed it).
Will be interesting to see how things roll out over the next few months.
apoppin
Lifer
Your analysis might be correct except you are forgetting GX2
- it trounces 3870x2 and requires QuadCrossfireX to match or beat it.
So for GX2 it is true, some people don't like the concept of SLI/CF on a single card and don't think it a *real* *solution* but the fact I can plug GX2 in and get higher frame rates than ANY single GPU (in the vast majority of games...emphasis on majority) - including that *other* sandwich compromise - is enough for me to deem GX2 as the 'solution' to a problem of getting better performance. (insert 'but theres input lag' comment here....I haven't noticed it).
it goes two ways ... and AMD doesn't perform
Sylvanas
Diamond Member
- Jan 20, 2004
- 3,752
- 0
- 0
Originally posted by: apoppin
Your analysis might be correct except you are forgetting GX2
- it trounces 3870x2 and requires QuadCrossfireX to match or beat it.
So for GX2 it is true, some people don't like the concept of SLI/CF on a single card and don't think it a *real* *solution* but the fact I can plug GX2 in and get higher frame rates than ANY single GPU (in the vast majority of games...emphasis on majority) - including that *other* sandwich compromise - is enough for me to deem GX2 as the 'solution' to a problem of getting better performance. (insert 'but theres input lag' comment here....I haven't noticed it).
it goes two ways ... and AMD doesn't perform
Sure, but I was responding to:
When in fact ATI's highest end product does infact compete and perform better than 2 year old technology.
I'm not disputing the GX2's better performance, nor am I comparing it to a 3870X2- which it is significantly faster than, although which where I am a 3870X2 is at least $250 cheaper than a GX2- and thus occupies a different price bracket which ATI have not fulfilled this generation.....much like when the 2900 launched it was to compete with the G80 GTS not higher priced GTX.
Sylvanas:
There's no $250 price difference between HD3870 X2 and 9800 GX2. Yes they have big price difference, but it's $170 at NewEgg..at least (Cheapest X2: $360, GX2: $530). You can still get 9600 GT SLI with $110 less money (even now when that Palit 9600 GT's price sky rocketed)..and 9600 GT SLI already wins HD3870 X2 (Because G94 seems to scale better for dual GPU solutions than RV670)
There's no $250 price difference between HD3870 X2 and 9800 GX2. Yes they have big price difference, but it's $170 at NewEgg..at least (Cheapest X2: $360, GX2: $530). You can still get 9600 GT SLI with $110 less money (even now when that Palit 9600 GT's price sky rocketed)..and 9600 GT SLI already wins HD3870 X2 (Because G94 seems to scale better for dual GPU solutions than RV670)
Sylvanas
Diamond Member
- Jan 20, 2004
- 3,752
- 0
- 0
As i said where I am there is a $250 price difference- I'm not in the US but over the ocean in Australia....nonetheless there still is a large gap in price wherever you go between the X2 and the GX2. Sure 9600GT SLI is an option but that requires the use of an SLI motherboard which people using the popular Intel chipsets don't have and thus the use of onbaord Xfire is a better option than spending money on changing mobo + buying new cards.
Sylvanas:
OK.. I just wondered when I had prices from Europe and North America.
Now that I checked with Shopbot.au..cheapest HD3870 X2 is 509$ and cheapest 9800 GX2 is 670$. That's more like 160$
Well 9600 GT SLI + Pretty good SLI motherboard (MSI P7N SLI) would cost 594$.
OK.. I just wondered when I had prices from Europe and North America.
Now that I checked with Shopbot.au..cheapest HD3870 X2 is 509$ and cheapest 9800 GX2 is 670$. That's more like 160$
Well 9600 GT SLI + Pretty good SLI motherboard (MSI P7N SLI) would cost 594$.
apoppin
Lifer
Originally posted by: Sylvanas
Originally posted by: apoppin
Your analysis might be correct except you are forgetting GX2
- it trounces 3870x2 and requires QuadCrossfireX to match or beat it.
So for GX2 it is true, some people don't like the concept of SLI/CF on a single card and don't think it a *real* *solution* but the fact I can plug GX2 in and get higher frame rates than ANY single GPU (in the vast majority of games...emphasis on majority) - including that *other* sandwich compromise - is enough for me to deem GX2 as the 'solution' to a problem of getting better performance. (insert 'but theres input lag' comment here....I haven't noticed it).
it goes two ways ... and AMD doesn't perform
Sure, but I was responding to:
When in fact ATI's highest end product does infact compete and perform better than 2 year old technology.
I'm not disputing the GX2's better performance, nor am I comparing it to a 3870X2- which it is significantly faster than, although which where I am a 3870X2 is at least $250 cheaper than a GX2- and thus occupies a different price bracket which ATI have not fulfilled this generation.....much like when the 2900 launched it was to compete with the G80 GTS not higher priced GTX.
Hang on .. are you trying to tell me you are *proud* of 3870x2 - finally, after nearly 2 years barely beating a SINGLE GPU?
-that is what *i* did basically .. got a 2900xt and played equally well to a 8800GTS THEN i got another 2900 and barely beat an Ultra .. for the same money as an Ultra if i got it last MAY
anyway, after 3870x2 finally beats an ultra .. and then a few weeks later it is blasted off it's shaky throne?
yeah, it is improvement considering 2900xt vs GTX Ultra .. but .. but ..
.. but .. .. i don't get it .. do i need ANOTHER 3870x2 - yeah, then i will get GX2 performance - for more than GX2 money
huh?
you better explain it
Sylvanas
Diamond Member
- Jan 20, 2004
- 3,752
- 0
- 0
No, you are putting words in my mouth. I was merely saying ATI's top offering does outperform the G80- Nothing more, I wasn't commenting if that's *good* or *bad*, note that I mentioned the *performance* difference between the two....yes it performs better, but if ATI had the same performance with a single GPU that would indeed be a BETTER *solution*(and a more than overdue one). If you want my opinion however, simply put, I'd say that ATI dropped the ball this round in relation to the high end- thats not a revelation however.
@ Rusin
Thats indeed a better price- I had no idea of Shopbot's existence :thumbsup: . As for the MSI board and 9600GT's- if i were building a new rig that would be my choice.
apoppin
Lifer
no "word putting" .. i simply did not understand what you were saying ..
... and it barely outperforms G80 ultra.
- i do understand now .. it is SO obvious i missed it .. but it is such a small plus for AMD, i just thought there was "more"
sorry
I hate having to quote myself, but I feel it was required here as if I was trying to say GPUs then I would have typed GPUs instead of GPU. If you want to compare multi GPUs then the x2 should be put up against 8800GTXs in SLI- point would still stand. ATi's highest end GPU struggles hard against nVidia's low-mid range 9600GT in settings people use(w/AA) for a higher price point. They are using a more advanced build process already, they are using a bigger transistor budget already, they are consuming more watts already then their competitors when looking at comparable performance. They already have more raw power and more shading power, the design simply sucks, and sucks badly. nVidia made a horrific release with the FX series of cards, one they tried to put behind them fairly quickly- AMD doesn't seem to get that they released a complete dud and move past it, they keep finding new ways to put in people's faces that they can not compete. I was running ATi parts in my system from the R9500Pro up until a very short while ago, right now ATi's offerings are, at best, pathetic.
Maybe they will get things turned around with their next generation part and return to, at the VERY least, a competitive level, as of right now they look an awful lot like 3dfx did at the end.
The 3870 X2 may be dual-GPU, but that is just the way the market is heading. It is no longer possible (in general) to sustain GPU advancement with a single GPU setup. Look at the size of G80.... ~480-530mm^2 and that's not even including the display chip which was separate. No company wants to produce a chip that big, it's not good for business. Now look at the 3870 X2.... it is made up of 2 small chips 192mm^2 in size, allowing for excellent yields and cheap cost. R700 will be the same idea, it will be two relatively small chips that are not too expensive to make. Not only does this make manufacturing easier, it makes design easier; the same RV770 core in the high-end parts can be used in the midrange cards.
Now nVidia might be going for a last hurrah with GT200, which will probably be a similar size to G80. But it's just not sustainable, process technology cannot keep up with GPU advancement. A few years ago, a top of the line GPU was ~200mm^2. Now we're looking at 500mm^2 on much more advanced processes.
R700 will be a success against G92b chips, the question is of course how well it will compete against GT200. But with 50% more shading processors, 2X texture units (major bottleneck in R600 design), higher clockspeeds, and GDDR5 memory, I think it should be a success.
That's 530mm on a 65nm process, though. There's still 55nm, 45nm and 32nm in the foreseeable future. People are more than happy to fork over a premium for a single card vs. the sandwich nightmare. So in the coming 2-3 (4?) years, it's still feasible to increase transistor counts with process shrinks keeping die sizes somewhat sane.
Put it another way. Give me a 4 way sandwich for $200 and a single GPU for $300, with comparable performance and I'll take the $300 single GPU. The math of a multi-GPU AFR solution dictates that you will NOT have latency any shorter than the amount of time a single GPU can render a frame, no matter what you do. If you're rendering 8 fps x 4 for 32 FPS in the benchmark and fraps you've still got a 1000/8 = 125 ms latency between acting and seeing the result when actually playing! I don't care about benchmarks as much as smooth, lag-free gameplay. If it costs me 50% more to get it, so be it. 100% is also within the realm of "I'd have to think about it very hard."
Until SFR improves to the point of being indistinguishable from a single GPU I'm not going to eat the crap sandwich both companies would LOVE to feed me.
Put it another way. Give me a 4 way sandwich for $200 and a single GPU for $300, with comparable performance and I'll take the $300 single GPU. The math of a multi-GPU AFR solution dictates that you will NOT have latency any shorter than the amount of time a single GPU can render a frame, no matter what you do. If you're rendering 8 fps x 4 for 32 FPS in the benchmark and fraps you've still got a 1000/8 = 125 ms latency between acting and seeing the result when actually playing! I don't care about benchmarks as much as smooth, lag-free gameplay. If it costs me 50% more to get it, so be it. 100% is also within the realm of "I'd have to think about it very hard."
Until SFR improves to the point of being indistinguishable from a single GPU I'm not going to eat the crap sandwich both companies would LOVE to feed me.
Depends on how good their performance is. G80 was huge success for Nvidia because it sold pretty well compared to AMD's counter part.
That R700 will use larger cores than R680; People are saying that RV770 would need like 200-250 million extra transistors which would make it 850-900 million transistor monster. In comparison GT200 is said to contain 1100-1200 million transistors. If Nvidia won't change architecture with GT200 it would mean that with 65nm this chip should be the same size as G80 was..most likely slight smaller if they have made any changes for architecture..which is very likely. It's not out of the question will this new chip come in 55nm since Nvidia is now launching already 55nm versions of G90-series.
GT200 will be quite a bit larger than G80. It consumes 225-250 watts vs. 177 watts and it is 65nm compared to 90nm. There have been reports that it barely fits on a PCB and is over 1 billion transistors. So what is that a 40%+ increase in transistors but a 28% decrease in process? It should be close to a 600mm^2 GPU. I think 40nm is after 55nm so we may be at the limit late next year.
G92:
754M transistors
324mm^2
G80:
680M transistors
484mm^2
-----
This rumoured 1100M transistor number would take us, with G92's transistor density, to around 470mm^2 if they haven't updated their architecture anyway (which should increase slightly that density and make that chipsmaller. If it's 55nm it could be around same size as R600 was.
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