GT300 taped out

Whoa, if this all is true then this is quite a beast. Wondering about the power requirements also.
Well TDP is pretty much the same as with GTX 295 in WORST CASE SCENARIO. GTX295 has TDP of 289W and since PCI-E standarts dictate that 300W is the limit..well.
 
Same performance ? You mean close performance and games. For anything else, NVIDIA GPUs win by a landslide. Just look at folding@home numbers where a single GTX 280 is twice as fast as a single HD 4870. NVIDIA doesn't market their cards for games only and the architectural specs reflect that.
Oh please. The number of people who give a shit about F@H is minuscule compared to the people who buy cards solely for games.

And PhysX barely has any support. You have Mirror's Edge, Cryostasis and to an extent, Unreal Tournament 3 - that's pretty much it.

Gaming performance is what sells cards regardless of what NVIDIA would like you to believe.
 
ATI isn't going to be able to compete without their own big chip. Don't expect any less than 2 billion transistors and die area > 400 mm2, for something that can compete with GT300, if these rumors are true.

At least how many percent would this card be faster than the ATI card? ;)
 
Oh please. The number of people who give a shit about F@H is minuscule compared to the people who buy cards solely for games.

And PhysX barely has any support. You have Mirror's Edge, Cryostasis and to an extent, Unreal Tournament 3 - that's pretty much it.

Gaming performance is what sells cards regardless of what NVIDIA would like you to believe.

This has hardly any thing to do with what I believe. It's what the cards can be used for. If you use the potential or not, is entirely your decision. And FYI, gaming performance is what sells card to GAMERS and the market is minuscule for GAMERS, since the bulk of graphics sales come from mid-low end cards sold to OEMs, in pre-built systems. Plus, Graphics cards are used for much more than that nowadays (GPGPU) and even before with 3D applications other than games (CAD tools, etc) and the market for that isn't that "minuscule"

I didn't even mention PhysX, because PhysX is not relevant for this discussion. PhysX is a physics API and can be used by anyone, not just GeForce owners.
And another FYI, games are not developed over night. PhysX already got support from big names in the gaming industry (EA, 2K, etc), but the games themselves won't come out tomorrow...
 
Oh please. The number of people who give a shit about F@H is minuscule compared to the people who buy cards solely for games.
Correct.
When a card can do both well, that's a win in my book. Go to any tech site's DC section and ask for advice on a folding/gaming setup. It's a market that adds to total cards sold.
 
I didn't even mention PhysX, because PhysX is not relevant for this discussion. PhysX is a physics API and can be used by anyone, not just GeForce owners.
And another FYI, games are not developed over night. PhysX already got support from big names in the gaming industry (EA, 2K, etc), but the games themselves won't come out tomorrow...
PhysX has been out for a couple of years now. NVIDIA acquired it Feb 08, and stand alone boards had been out for at least a year before that. It's had plenty of time to be implemented, and at this point I don't hold much hope that it will be (bear in mind I have a 9600GT solely for PhysX so I would sure as hell be glad if it was, I just don't think it's going to happen.)
 
PhysX has been out for a couple of years now. NVIDIA acquired it Feb 08, and stand alone boards had been out for at least a year before that. It's had plenty of time to be implemented, and at this point I don't hold much hope that it will be (bear in mind I have a 9600GT solely for PhysX so I would sure as hell be glad if it was, I just don't think it's going to happen.)

You said it. NVIDIA acquired it in 2008. PhysX's problem was the fact that AGEAI didn't really promote well enough for developers to use instead of Havok. That changed with NVIDIA and even though not many, there are already games that support PhysX right now and more are coming. They can't all show up in just one year. Again, this has nothing to do with the discussion (which is why I didn't understand why you brought it up), since PhysX is a physics API and its use is not directly related with graphics cards and their sales.
 
http://www.hardware-infos.com/news.php?news=2954&sprache=1
They are claiming that GT300-A1 samples are working with following clocks:
Core 700MHz
Shaders 1600MHz
Memory 1100MHz (4400MHz effective)

Single precision performance 2,457 teraFLOP/s (GT200 0,933)..double precision performance of G300 should be higher than GT200's single precision?? Still too hard to believe
 
Last edited:
http://www.hardware-infos.com/news.php?news=2954&sprache=1
They are claiming that GT300-A1 samples are working with following clocks:
Core 700MHz
Shaders 1600MHz
Memory 1100MHz (4400MHz effective)

Single precision performance 2,457 teraFLOP/s (GT200 0,933)..double precision performance of G300 should be higher than GT200's single precision?? Still too hard to believe

And 281.6 GB/s of memory bandwidth...If these turn out to be true...what a monster graphics card...
 
Kind of makes me feel glad that I don't have the cash for a GT200-based card at this point. GT300 is going to be awesome if even half of those specs are true.
 
You said it. NVIDIA acquired it in 2008. PhysX's problem was the fact that AGEAI didn't really promote well enough for developers to use instead of Havok. That changed with NVIDIA and even though not many, there are already games that support PhysX right now and more are coming. They can't all show up in just one year. Again, this has nothing to do with the discussion (which is why I didn't understand why you brought it up), since PhysX is a physics API and its use is not directly related with graphics cards and their sales.
Because PhysX is a selling point of the video cards exactly like CUDA and other GPGPU features like Folding@Home are, so if you are going to argue that one of them is a selling point why would you discount the other? I would argue that PhysX is a lot more relevant to how most people use their graphics cards than the GPGPU functionality, as it ties into gaming performance.
 
Because PhysX is a selling point of the video cards exactly like CUDA and other GPGPU features like Folding@Home are, so if you are going to argue that one of them is a selling point why would you discount the other? I would argue that PhysX is a lot more relevant to how most people use their graphics cards than the GPGPU functionality, as it ties into gaming performance.

Your claim was that PhysX is irrelevant at this point due to it not being used much. It was then pointed out that PhysX didn't get marketed by a major company until last year, as well as that last and this year there was a major uptake of PhysX by big game companies. Considering the development time of games to be 1-2 years, if not more, it would be reasonable to assume that there won't be many PhysX-based games yet.

It doesn't mean that PhysX isn't a big selling point, it just shows the power of marketing (Aegia was a nobody selling something new, never an easy feat) and the inertia in the game market.
 
Your claim was that PhysX is irrelevant at this point due to it not being used much. It was then pointed out that PhysX didn't get marketed by a major company until last year, as well as that last and this year there was a major uptake of PhysX by big game companies. Considering the development time of games to be 1-2 years, if not more, it would be reasonable to assume that there won't be many PhysX-based games yet.

It doesn't mean that PhysX isn't a big selling point, it just shows the power of marketing (Aegia was a nobody selling something new, never an easy feat) and the inertia in the game market.
Maybe I just pay more attention than other people, but here is the HardOCP "editorial" on PhysX from just about 3 years ago: http://www.hardocp.com/article.html?art=MTA1MiwsLGhlbnRodXNpYXN0 I've known about these cards for a while, I just never had one because they were expensive and there was little need.

Since that time, how many big name titles have come through? I realize that AEGIA certainly doesn't have the marketing sway that NVIDIA does, but it's also not like this technology is brand new. With the direction that game programming is going, I would not be entirely surprised to see a newer, more open physics implementation take it's place.

Anyways, my entire point was that the market for PhysX is about the same as the market for Folding@Home - small. Most people buy video cards for video performance, regardless of how they are marketed, with the exception maybe being the professional FireGL/Quadro market. Knocking ATI because their cards fold slower is ultimately pointless in the grand scheme of things because probably a fraction of a % in sales is due to F@H performance. If folding performance is meaningful to you, then by all means - speak with your wallet, and buy the other brand that does it better. But I think the majority of consumers are going to go after cards that offer them cheaper, speedy gaming performance, and that is the market ATI has been targeting.
 
Anyways, my entire point was that the market for PhysX is about the same as the market for Folding@Home - small. Most people buy video cards for video performance, regardless of how they are marketed, with the exception maybe being the professional FireGL/Quadro market. Knocking ATI because their cards fold slower is ultimately pointless in the grand scheme of things because probably a fraction of a % in sales is due to F@H performance. If folding performance is meaningful to you, then by all means - speak with your wallet, and buy the other brand that does it better. But I think the majority of consumers are going to go after cards that offer them cheaper, speedy gaming performance, and that is the market ATI has been targeting.

similar things could have been said some time ago about some CPU features. how many users even now make use of virtualization? yet it's considered a standard feature in both camps, and i have to imagine a reviewer would ding a new CPU that didn't support it.

your original point was "Gaming performance is what sells cards regardless of what NVIDIA would like you to believe." by selling cards i take it you mean in volume, and the cards that sell in volume aren't usually flagship cards anyway. usually they're either a generation behind or a lower-end card.

flagship products in general only attract a small fraction of consumers in the first place, so in a way your argument falls in on itself. the F@H and physX consumer bases might be only small parts of the general graphics card market, but not so much in the top-tier market.

video game performance itself isn't what sells cards, OEM partners sell cards, and i can't help but think having the performance crown and getting the enthusiast community buzzing helps to secure those partnerships. i don't know if they will take the crown, but i can't say these rumors don't look very promising.
 
Last edited:
Ageia was a tiny, tiny company, with barely the resources to produce the PPU hardware, let alone promote their technology. No sane game company would use it for a big title, or even any title at all. Heck, my own company probably wouldn't be using PhysX right now if nVidia or similar wouldn't be backing it, with the expensive PPUs replaced with the same GPUs already installed in many people's systems.

With all the money, time and resources already invested in PhysX and Havok, I really don't see how they could be replaced in less than 5 years time. What isn't open enough in your opinion about PhysX/Havok, by the way?

Physics is rapidly becoming an essential part of modern games. Now that the graphics look quite realistic already, it's time we get rid of clipping, skating, lack of mass and all those millions of things which make it hard to take the real-time animations in games serious. For that we need hardware-accelerated physics, being PhysX at this point.
 
And 281.6 GB/s of memory bandwidth...If these turn out to be true...what a monster graphics card...

HD 2900 XT 128 GB/s memory bandwidth
HD 4890 125GB/s memory bandwidth

guess which is a more "monster" card.
 
HD 2900 XT 128 GB/s memory bandwidth
HD 4890 125GB/s memory bandwidth

guess which is a more "monster" card.
didnt the 2900xt have 1650mhz memory on a 512bit bus for 105.6 gb/sec? either way it wouldnt have mattered because memory bandwidth was not where the 2900xt was lacking.
 
HD 2900 XT 128 GB/s memory bandwidth
HD 4890 125GB/s memory bandwidth

So because the 2900XT had lots of bandwidth and was slow you're implying that another card will lots of bandwidth will also be slow? Nice logic there.
 
So because the 2900XT had lots of bandwidth and was slow you're implying that another card will lots of bandwidth will also be slow? Nice logic there.
I think he was just replying to someone else that seem to be saying the card would be a monster because of the memory bandwidth. his point appeared to be that you dont judge a card based on one aspect of its design. Im sure he knows that a better overall card can actually utilize that memory bandwidth.
 
ATI isn't going to be able to compete without their own big chip. Don't expect any less than 2 billion transistors and die area > 400 mm2, for something that can compete with GT300, if these rumors are true.

Yeah and what happened when RV770 came out with 800 SP's and nearly 1 billion transistors packed into a die half the size of GT200?

I'm sorry, but even if the specs are true, you're plain wrong about what they would need to compete. GT200 had twice the specs of the G80 and it didn't provide 2x the performance, now did it?

This is why I'm going to wait rather than drool over a product who's specs aren't even confirmed. Everyone saw GT200 as the next G80 because: "OMG, 240 shaders is nearly 2x 128 shaders, it's got 1.5x more bandwidth, ROPS, and TMU's so it must be much faster! ATI has no chance, it's got a tiny die and can't possibly be more efficient! It's going to need 1000 shaders!" etc. etc.

I'm sure Nvidia isn't going to rest on their laurels, but given that we don't even have confirmation of how much this has changed architecturally (or even if it is new) I'm not going to jump on a hype bandwagon that hasn't even gotten a real set of wheels yet.
 
Yeah and what happened when RV770 came out with 800 SP's and nearly 1 billion transistors packed into a die half the size of GT200?

I'm sorry, but even if the specs are true, you're plain wrong about what they would need to compete. GT200 had twice the specs of the G80 and it didn't provide 2x the performance, now did it?

This is why I'm going to wait rather than drool over a product who's specs aren't even confirmed. Everyone saw GT200 as the next G80 because: "OMG, 240 shaders is nearly 2x 128 shaders, it's got 1.5x more bandwidth, ROPS, and TMU's so it must be much faster! ATI has no chance, it's got a tiny die and can't possibly be more efficient! It's going to need 1000 shaders!" etc. etc.

I'm sure Nvidia isn't going to rest on their laurels, but given that we don't even have confirmation of how much this has changed architecturally (or even if it is new) I'm not going to jump on a hype bandwagon that hasn't even gotten a real set of wheels yet.

That coming from you is a real treat, since you were one of the first jumping on the hype bandwagon for the 800 SPs rumor, which I was the first to share in this forum btw.

No one is jumping on any hype, except you. We are talking about rumors, just like we were talking about rumors in the past. It seems you and others have a hard time grasping a simple concept called conditional praise. IF it has these specs, it will be a monster graphics card. How it will perform, can only be defined WHEN it is released and tested. You praised RV770's performance based on perlin noise tests, which were only rumors at the time there was still much doubt on the Stream Processor count (which most rumors pointed to 480 and you just refused to accept it).

So don't come here with that kind of speech, because everyone with memory remembers what you said and if not, can easily verify what I said above with a simple search. Rumors are rumors until verified. There's no hype, just speculation based on the rumors. Learn to live with it and don't lecture anyone, with something you do yourself, got it ?
 
Ok guys no more fighting...lets just imagine ourselves for now with 2xGT300 in our Core i7 system getting over 130FPS on Crysis and GTA IV everything maxed out running on a pair of 30inch Dell LCD's:D:D :drool:
 
ZerazaX said:
GT200 had twice the specs of the G80

No it didn't. But you're right, there's no way to tell right now what AMD will need to counter GT300 or vice versa. But it's pretty clear that these rumours imply a much bigger jump over GT200 than the latter was over G80.
 
I'll stick with ATI. Why lets face it Monolithic chip design is dead end. Simple as that. Look at CPU's.
I can't imagine that the G300 will have a good yield and I am sure it will cost hell of a lot to make.

ATI was going to make simpler smaller cores and then stick them on to one die. Since GPUs are already perellel beasts this will not matter. I am sure that ATI with implement even better Bridge for crossfire as well. Heck 48XX has a good implementation already they can improve on it more for larger scaling.

As for intel Larabee I think its going to be mostly hype and a big FAIL.
 
I'll stick with ATI. Why lets face it Monolithic chip design is dead end. Simple as that. Look at CPU's.

Actually, most CPUs are monolithic designs. They are the whole chip, including all the cache, on one piece of silicon. All AMD chips are this way, and most Intel chips are. The ones that buck the trend are the Core 2 Quads and Pentium Ds which are two separate chips in one physical package.

Don't confuse cores with chip design. You are correct that all the chips have gone multi-core, but they are still monolithic pieces of silicon. GPUs actually went this way long before CPUs, and to a much greater extent since graphics are so highly parallel. That's why you'll see lower end components that are just cut down version of high end units. You have something like an 8800 that has 8 sets of 16 shader processors. So, you cut of 75% of those and you get an 8600GT. No major redesign, you just scale down the number of units and thus size and cost.

So monolithic chips are still very popular. I'm not saying it is the right or wrong way for the future of GPUs, but it is the way most things are done. There's a good chance the processor in your system is monolithic, and a good chance the next one will be too.
 
I'll stick with ATI. Why lets face it Monolithic chip design is dead end. Simple as that. Look at CPU's.
I can't imagine that the G300 will have a good yield and I am sure it will cost hell of a lot to make.
What do you care how much it costs them to make it? The only thing that should matter to a consumer is the end price. Last I checked the 260 is usally a few dollars less than a 4870 1 GB. But lets not look at that, lets repeat viral marketing ads. :rolleyes: Oh, and see above me for a response to your "look at CPUs".

ATI was going to make simpler smaller cores and then stick them on to one die. Since GPUs are already perellel beasts this will not matter.
As long as it's on one peice of silicon it is a monolithic chip. If it's not, welcome to latency.

I am sure that ATI with implement even better Bridge for crossfire as well. Heck 48XX has a good implementation already they can improve on it more for larger scaling.
Well I know Nvidia will make SLI scale over 100% in the next gen cause they're uber! :rolleyes:
 
ATI have DX11 in their cards now. NV is dragging feet and doesnt even have DX10.1
NV cant get their process shrink to work.

when Nv actually have a working 40nm card that doesnt suck power down like homer in a beer factory... I'll consider looking.

SLI is tiny market and does not scale well once you hit 3 or 4 cards.
 
ATI have DX11 in their cards now. NV is dragging feet and doesnt even have DX10.1
NV cant get their process shrink to work.

when Nv actually have a working 40nm card that doesnt suck power down like homer in a beer factory... I'll consider looking.

SLI is tiny market and does not scale well once you hit 3 or 4 cards.
and everything at ATI is working just right and crossfire scales perfectly?
 
when Nv actually have a working 40nm card that doesnt suck power down like homer in a beer factory... I'll consider looking.

If you want to talk about sucking check the power consumption of a 4850/4870. :p

Seems a bit early for the ATI fans to already be afraid of this card and want to downplay it.

Myself I will enjoy the summer and hopefully read some interesting reviews by Halloween.
 
ATI have DX11 in their cards now. NV is dragging feet and doesnt even have DX10.1

That's right cause the 4890 is DX11 ready.

when Nv actually have a working 40nm card that doesnt suck power down like homer in a beer factory... I'll consider looking.
Your right, ATI used half the power of Nvidia cards. And that extra <1$ a month on an electric bill is the difference between me being able to afford a video card and not. :rolleyes:

SLI is tiny market and does not scale well once you hit 3 or 4 cards.
Your right, 4870x2x2 scales over 100%!

ugh. I'm going to have to go to the store, I'm out of sarcasim.
 
Scales better than SLI :p

Moot point anyway as SLI/Xfire make up a tiny fraction of the market.

My point is that if you are investing cash in a card for at least 1-2yrs then makes sense to have the latest tech in it.

Also given Nv's current trend of remarketing old as new and other pr stunts... I really dont think we're going to see this "awesome" card anytime soon. It also doesnt make sense how they are going about it. But we'll see i guess. Maybe Nv has a rabbit in the hat.... instead of another flashy rebrand.
 
If you want to talk about sucking check the power consumption of a 4850/4870. :p

Seems a bit early for the ATI fans to already be afraid of this card and want to downplay it.

Myself I will enjoy the summer and hopefully read some interesting reviews by Halloween.

I'm not afraid of it. I just dont think they are going to have it any time soon nor that it will do what is promised.

Are they going to follow DX11? or whine to MS again and get most of it pulled as its "too difficult to do" again?
 
Still this better be the can of whoopass Jen-Hsun Huang promised a while ago.
 
uh.. nvidia being first to market dx11? huh? ATI already started marketing it months ago with their news releases..

though i would read the articles.. but that website refuses to load for me..

though id still say bs on the 512bit memory.. nvidia must be on crack if they think they can make a profit off the 300 series using it.. 512bit gddr5 is the equivilant of 1024bit gddr3.. now the problem with that is, the fastest gddr5 that is in production is only 5ghz.. now why in any ones right mind would bother wasting all that money to create a 512bit memory controller for memory that isnt even close to its maximum potential? gddr5 has already been taken up and passed 7ghz.. its just not in mainstream production yet.. and the 5ghz chips just started going in production a couple months ago.. so in the end.. the question is.. will the gt300 have the old 3.6-4ghz chips.. or will it have the new 5ghz chips.. and since amd/ati designed and pushed the development of gddr5.. id take a good guess that they already bought out the entire supply of 5ghz chips..

until there are some actual hard facts and not a bunch of bogus rumors floating around to make people think nvidia's actually pulled their heads out of their butts.. its all just rumors.. so dont get your hopes to high when you find out that whats been plastered all over the news isnt even close to the final product... and yes.. i could be wrong.. but to do what everyone is saying nvidia's going to do is a suicide run to try and beat ATI.. aka.. welcome back the 600+ dollar nvidia cards..

fanboy alert :p
 
Still this better be the can of whoopass Jen-Hsun Huang promised a while ago.

only if GPU computing takes off. rumor has it (take with a BIG grain of salt) that they have allocated a large portion of their die space to this. I can't see this being good for actual game play. phsyx yes, encoding, maybe if they make one that works well (and isn't competing with free encoders that work better if slower), but for rendering? Maybe the ray tracing will be a big thing to come but right now I don't understand why they are doing that.
 
There is an old saying that goes "A fool and his money are soon parted" :D

ATI has been behind NVIDIA for years now, I don't see that changing anytime soon.

I mean common man, can we be a little realistic here, ati and nvidia are pretty much the same performance wise right now, and this comes from a gtx 275 owner here, I am tired of people fighting over graphic cards. I will be honest and say that while I do believe that the specs might be true but reports of nvidia having sample card is highly unlikely, I mean they don't even have any 40nm card right now on the market, I have an open mind about it because who knows its not like I have any sources, but looking at their current market situation, I think they will be behind in release to ati.
 
Back
Top