"2009 in graphics, ATI wins"

[thilanliyan]

http://www.fudzilla.com/content/view/16988/1/

Has anyone seen a recent discrete GPU sales breakdown?

 

[Lonyo]

No, because one's not due until middle of February.
The most recent by JPR was from mid-November, which was for Q309, so another month and a half for Q4.

http://www.jonpeddie.com/publications/add-in-board-report/

AMD is due for their end of year report any day now, while NV don't file anything again until end of January.

 

[evolucion8]

The article its a bit inaccurate, stating that the GTX 260 were launched early in 2008 when it was launched in June 16, stating that the ATi's HD 4870 wasn't competitive enough when in reality, nVidia had to create a GTX 260 216 to remain competitive against the HD 4870 1GB in which traded blows depending of the game scenario.

I know that ATi could be in a better position if it isn't because of the issues with TSMC.

 

[Ackmed]

Well, obviously they did. Had a much better year, with better cards.

 

[Arkaign]

Is the article written by Yoda? The hell, what?

 

[scooterlibby]

Should have been: "2009 in graphics, consumers lose "

Seriously, ouch! No competition from Nvidia and ATI's shortages have left my butt sore (don't ask). I really thought I would have something different sitting in my rig by now.

 

[Arkaign]

Well, at least there's nothing that SLI'd GTX260's can't play really well, particularly at 1920x1200

 

[soccerballtux]

Well, at least there's nothing that SLI'd GTX260's can't play really well, particularly at 1920x1200

yeah but you gotta pay an arm and a leg for SLI capability.
CrossfireX came with my mobo. Totally value added, and now I will be buying AMD/ATI cards, 1 now and probably another one later.

 

[scooterlibby]

Well, at least there's nothing that SLI'd GTX260's can't play really well, particularly at 1920x1200

Very true. I'd still like to max out Crysis Warhead and Stalker Clear Sky, though, and be ready for DX11 BF Bad Company 2.

 

[lifeblood]

Maybe it's me but I think Nvidia is taking crap they don't deserve. The guys at Nvidia are not idiots, but they do make mistakes (GeForce FX). It is possible they made another with Fermi. However, from everything I've read its TSMC that dropped the ball on this. They are the ones who can't seem to make 40nm work. Nvidia is hurting more than ATI because they have the larger, more complex chip, but both companies are suffering pain because of TSMC.

If Fermi shows up and it sucks, we should feel free to blast nVidia as idiots. Until then, we should be blasting TSMC as idiots, not Nvidia.

BTW, is it Nvidia or nVidia?

 

[Lonyo]

Maybe it's me but I think Nvidia is taking crap they don't deserve. The guys at Nvidia are not idiots, but they do make mistakes (GeForce FX). It is possible they made another with Fermi. However, from everything I've read its TSMC that dropped the ball on this. They are the ones who can't seem to make 40nm work. Nvidia is hurting more than ATI because they have the larger, more complex chip, but both companies are suffering pain because of TSMC.

If Fermi shows up and it sucks, we should feel free to blast nVidia as idiots. Until then, we should be blasting TSMC as idiots, not Nvidia.

BTW, is it Nvidia or nVidia?

You say that, but all I see is both AMD/ATI and also Nvidia shipping 40nm parts.
I don't see how TSMC can be the only one to blame when both companies have parts shipping. It's a problem with the NV design philosophy that they are in their current situation. It's their fault they don't have any 40nm DX11 parts, it's their fault their high end chip isn't shipping yet, it's their fault if excessive delays have been caused by TSMC (since AMD have managed to get out not one but two DX11 product lines).

You can try and lay the blame at TSMC all you want, but both companies have had 40nm shipping for a while now, so any failure can only lie with NV. They also haven't been helping themselves by 1) complaining about TSMC and 2) saying they will not use Globalfoundries.

Here's a question for you:
If Fermi has been delayed by 5+ months, and isn't particularly able to be scaled down as-is, where are the rest of the chips?
One would assume that the basic architecture is down, and can be used to design other chips, but with an already 5 month delay, where is the rest of the NV DX11 product line news? Can it be that despite a big delay on Fermi they don't really have anything for the rest of the market except respins and renames of existing products?
ATI have already launched the HD58xx cards, and the HD57xx cards, and there are more launches planned for January (HD56xx and maybe something else?), so what have NV been doing in the meantime? They've launched a very low end DX10.1 part called the GT310 (or something), which is... 40nm! And OEM only, and has pretty much zero performance.
Where is the news of a full top-to-bottom DX11 lineup? Lets blame TSMC!

 

[v8envy]

All that aside, I'd love to see the GTX260 216 with DX11 on 40nm sporting a street price of around $160 to match the 5770. I'd certainly consider one or two.

I wonder what the 8800GTS aka 9800GTX aka GT250 will be named in the 3 product line. 340? 350? It'll be some kind of record to have essentially the same GPU last four years & generations while having only moved from upper midrange to lower midrange in the lineup.

 

[lifeblood]

If Fermi has been delayed by 5+ months, and isn't particularly able to be scaled down as-is, where are the rest of the chips?

Looking at it that way, you do have a point. However, isn't nV's strategy to build a single monolithic GPU for the top end then scale it down from there? I think that's what I read on this website. So if they can't get that single monolithic chip working, then the rest of the lineup is also delayed. Of course, that would still make this Nvidias fault as it is a fault of their design philosophy.

I accept your point, sir.

 

[Lonyo]

Looking at it that way, you do have a point. However, isn't nV's strategy to build a single monolithic GPU for the top end then scale it down from there? I think that's what I read on this website. So if they can't get that single monolithic chip working, then the rest of the lineup is also delayed. Of course, that would still make this Nvidias fault as it is a fault of their design philosophy.

I accept your point, sir.

Well, for that it would be helpful if someone like IDC was here to explain respins and what they impact.

Another question is... what exactly are the midrange Fermi based cards going to be, if they even exist? No one really knows.
Fermi is architected to make up for the GT200 shortfalls in HPC work, right? As well as (obviously) improve GPU performance.
Do mid-range cards (certainly one would think the low end doesn't) require a lot of those features? I mean, sure, make Fermi optimised for HPC workloads and also good at gaming, but does that mean they intend to carry the same architecture over to the mid-range?
The Digitimes story which came out today (http://www.digitimes.com/news/a20091228PD207.html) implies that NV will just keep using some existing products (plus the GT310 or 305, whatever it's called), to fill out the rest of their product range, meaning no DX11 anywhere except the high end/top of the line, and only then in Q2 when "GF104" comes out. That's going to be over 6 months behind ATI, and will only cover about half the market (by revenue) with DX11 cards.

For the OEMs at least, there could be some issues with that, although at least NV already have the very low end covered with something cheap in the 300 nomenclature.

 

[scooterlibby]

You say that, but all I see is both AMD/ATI and also Nvidia shipping 40nm parts.

True they are shipping, but it's a dribble and that is a partial cause of the above MSRP prices the 5 series cards are commanding. Also, it's impossible to get a 5970. When you search Newegg now only one comes up and it's out of stock at $679. I think there's plenty of blame to go around, and TSMC deserves some of it.