Astounding Year-to-Year Growth in PC Graphics

[exar333]

I'm sorry, I guess I'm reading it wrong but that's not what I'm seeing. Nvidia has lost overall marketshare (24.3% this year vs 30.6% last year) while Intel and AMD have both increased overall marketshare.

It does appear that Intel, ATI, & Nvidia have all increased the total number of units shipped, or am I reading that wrong?

Maybe I'm being a fanboy, but the Intel numbers really annoy me. They have traditionally shipped crap integrated GPU's and get away with it because they currently have they better CPU. I know it's capitalism, but it still is annoying.

No, your just flat out wrong. Integrated graphics is enough for a majority of PC users.

 

[SirPauly]

<post was removed>

Sometimes it is wise to read instead of react with a cheap-shot:

AMD gained in the notebook integrated segment, but lost some market share in discrete in both the desktop and notebook segments due to constraints in 40nm supply. Nvidia picked up a little share overall. Nvidia&#8217;s increases came primarily in desktop discretes, while slipping in desktop and notebook integrated.

Most of the discussion in video card forums is actually video cards or discrete desktop cards.

 

[Genx87]

Hard to imagine why Nvidia wants to get into HPC and mobile markets. Intel's craptacular integrated GPU is steadily eating the market.

 

[GaiaHunter]

Intel was the leader in Q4’09, elevated by Atom sales for netbooks, as well as strong growth in the desktop segment. AMD gained in the notebook integrated segment, but lost some market share in discrete in both the desktop and notebook segments due to constraints in 40nm supply. Nvidia picked up a little share overall. Nvidia’s increases came primarily in desktop discretes, while slipping in desktop and notebook integrated.

Most of the discussion in video card forums is actually video cards or discrete desktop cards.

Although that statement doesn't go in line with the numbers presented.

I don't see how NVIDIA could have picked up a little share overall if the overall share of NVIDIA shrunk from 25.3% in 3Q09 to 24.3% 4Q09.

So either the statement is wrong or the numbers are wrong.

Most interesting to AMD is the increasing in revenue and in operating incoming. Considering the information is accurate, it is clear the problems that plagues(plagued) the TSMC 40nm hurt the potential sales of AMD. On the other hand they seemed to be having much larger margins - unless the integrated notebook was the real mover for AMD (wich is doubtful).

 

[heyheybooboo]

MCP6xx still lives! Long live the nForce 430 south bridge!

(not really)

It has even transitioned (limped?) to AM3. I'd also guess that quite a few s775 7xxx chips have shipped in the last year. OEMs eat them all up. Go cheap FTW.

I imagine nVidia has seen the writing on the wall. That gravy train is coming to an end in a slow death, and they will squeeze every last penny of profit out it while they still can ...





--

 

[ViRGE]

In case you didn't read the entire PR, keep in mind with these numbers is that NVIDIA doesn't use a typical calendar system. They use fiscal quarters, which are 11 months ahead of the real world (i.e. Fiscal September 2010 is really October of 2009), which means that they are basically a month behind everyone else. So NVIDIA's "Q4" is November-January, which obviously isn't done yet, so JPR's numbers are from NVIDIA's Q3 (August-October) while they're from everyone else's Q4 (October-December).

Anyhow, the point is that NVIDIA's numbers are for the late summer through the fall, and do not include Christmas. Christmas of course is a huge time for computer shares, so NVIDIA's actual market share is likely a bit higher. But this won't show up until Q1'10 when NVIDIA's Christmas sales will show up against everyone else's weak post-Christmas sales.

 

[SirPauly]

Although that statement doesn't go in line with the numbers presented.

I don't see how NVIDIA could have picked up a little share overall if the overall share of NVIDIA shrunk from 25.3% in 3Q09 to 24.3% 4Q09.

So either the statement is wrong or the numbers are wrong.

Most interesting to AMD is the increasing in revenue and in operating incoming. Considering the information is accurate, it is clear the problems that plagues(plagued) the TSMC 40nm hurt the potential sales of AMD. On the other hand they seemed to be having much larger margins - unless the integrated notebook was the real mover for AMD (wich is doubtful).

I don't understand why -- nVidia lost more over-all share than AMD but gained in discrete marketshare according to JPR. The reason why this may be surprising to some is based on ATI's new 5XXX line of GPU's but it's obvious that volume was the reason -- not the quality of product.

 

[Aaluran]

snip

Indeed. It went right by my eyeballs without registering in my brain. What does that do besides create confusion? Give a(n illusionary) bump to the figures of the slow season? Idiotic.

 

[SirPauly]

Indeed. It went right by my eyeballs without registering in my brain. What does that do besides create confusion? Give a(n illusionary) bump to the figures of the slow season? Idiotic.

That was in the context of revenues and JPR couldn't offer nVidia's revenue numbers based on their fiscal fourth quarter hasn't ended yet. JPR's numbers offered are GPU's shipped/sold for the calender year/quarter from all parties, imho.

 

[Wreckage]

I don't understand why -- nVidia lost more over-all share than AMD

This is mainly because they have halted chipset sales thanks to a cockblock by Intel.

but gained in discrete marketshare according to JPR.

The reason why this may be surprising to some is based on ATI's new 5XXX line of GPU's but it's obvious that volume was the reason -- not the quality of product.

DX11 is not a real motivator and with the 58xx so high priced and the rest of the line performing poorly, people went with another option, clearly.

 

[Aaluran]

So market share quarters are the same for everyone and NV's doesn't match their revenue quarter. Am I getting this right this time?

Very confusing.

I was wondering what the general purpose of using a different fiscal calendar was. Besides create a headache of reading financial information.

 

[SirPauly]

This is mainly because they have halted chipset sales thanks to a cockblock by Intel.


DX11 is not a real motivator and with the 58xx so high priced and the rest of the line performing poorly, people went with another option, clearly.

I disagree, and believe that if there were more 5XXX volume, nVidia may of lost some discrete market share. DirectX 11 and tessellation are important and one can see that it is important for nvidia as well, moving forward.

 

[Genx87]

In case you didn't read the entire PR, keep in mind with these numbers is that NVIDIA doesn't use a typical calendar system. They use fiscal quarters, which are 11 months ahead of the real world (i.e. Fiscal September 2010 is really October of 2009), which means that they are basically a month behind everyone else. So NVIDIA's "Q4" is November-January, which obviously isn't done yet, so JPR's numbers are from NVIDIA's Q3 (August-October) while they're from everyone else's Q4 (October-December).

Anyhow, the point is that NVIDIA's numbers are for the late summer through the fall, and do not include Christmas. Christmas of course is a huge time for computer shares, so NVIDIA's actual market share is likely a bit higher. But this won't show up until Q1'10 when NVIDIA's Christmas sales will show up against everyone else's weak post-Christmas sales.

The only thing in the article that has to do with Nvidia's calendar is revenue. This article is using the traditional calendar year cycle for shipments.

 

[Lonyo]

I disagree, and believe that if there were more 5XXX volume, nVidia may of lost some discrete market share. DirectX 11 and tessellation are important and one can see that it is important for nvidia as well, moving forward.

Speculatively based solely on the units shipped and AMD's revenue figures, I am not sure it would have made that much difference if there had been more shipping HD5xxx units.
Sure it might have been slightly beneficial and prevented some desktop share erosion, but based on the increase in AMD revenue vs increase in units shipped, and the fact that NV increased discrete marketshare, it seems reasonable to assume that NV gained a lot in the low end (where there are no HD5xxx cards), so HD5xxx shipments would not be hugely relevant.
That being said, the upcoming launch of low end DX11 products for AMD might help them in the low end to get back some marketshare, since they will have a technically superior product in terms of API support and (probably) a lower power and cheap to make product which will be welcomed by OEMs.
At the moment AMD has no real advantage over NV in the low end high volume market, which makes a strengthening NV position there quite possible/probable, and would account for increased overall desktop discrete share.

Of course, nothing can be confirmed without either NVs figures for revenue growth, or just simply a better breakdown of numbers (anyone want to pay the $1000 for the full report?).

 

[evolucion8]

This is mainly because they have halted chipset sales thanks to a cockblock by Intel.


DX11 is not a real motivator and with the 58xx so high priced and the rest of the line performing poorly, people went with another option, clearly.

Please, the HD 5800 series is faster than anything nVidia has currently in their respective price point, the HD 5770 competes well with the GTX 260+ and is cheaper, the HD 5750 is slighly faster than the GTS 250 and costs the same, the HD 5670 is faster than the GTS 240 and GTS 220 and yet, is cheaper. Which kind of people will go to another option with outdated tech? No DX11, no Eyefinity, higher power consumption?

Please, the choice is clear, nVidia will be more relevant when it launches its lineup with Fermi derivatives and hopefuly it will perform good and will lower prices, because the money is in the midrange and low end cards.

 

[akugami]

The only thing I'd note is that it's easy to try to say the Radeon 5 series didn't help AMD in sales. It would also be true. The unbiased will note, and even the article does so, that AMD was supply constrained by 40nm parts and we all know that the parts in question were really the fact that they could not get enough Radeon 5 series cards out.

The reality is that a lot of people were waiting for Radeon 5 cards in October and November but most couldn't buy any until at least the middle of December. Supply has been better in January onwards so it'll be interesting to see how AMD does in the next quarter with better supplies and then compare the following quarter after that when Fermi gets released. I think this will give a truer sense of how well the Radeon 5 series is doing for AMD.

 

[Wreckage]

Please, the HD 5800 series is faster than anything nVidia has currently in their respective price point, the HD 5770 competes well with the GTX 260+ and is cheaper, the HD 5750 is slighly faster than the GTS 250 and costs the same, the HD 5670 is faster than the GTS 240 and GTS 220 and yet, is cheaper. Which kind of people will go to another option with outdated tech? No DX11, no Eyefinity, higher power consumption?

Please, the choice is clear, nVidia will be more relevant when it launches its lineup with Fermi derivatives and hopefuly it will perform good and will lower prices, because the money is in the midrange and low end cards.

Well your opinion clearly does not represent the majority of video card buyers.

Eyefininty? That represents maybe .01% of the market. DX11 as we noted in another thread was around 10% of the market. Right now neither of these fluff options are worth dropping coin on a 5xxx series.

If anything 3D and PhysX are more compelling options and have broader industry support. This is probably why NVIDIA still maintains a marketshare advantage.

 

[Aaluran]

Well your opinion clearly does not represent the majority of video card buyers.

Eyefininty? That represents maybe .01&#37; of the market. DX11 as we noted in another thread was around 10% of the market. Right now neither of these fluff options are worth dropping coin on a 5xxx series.

If anything 3D and PhysX are more compelling options and have broader industry support. This is probably why NVIDIA still maintains a marketshare advantage.

First, what he meant was that (as he clearly said and you completely ignored), Radeon 5000-series beat or equal their nVidia competitors for performance AND are cheaper and less power-hungry AND have more features at each price point. All at the same time. The features was really the smallest part of his argument.

Also, if DX 11 is fluff why is nVidia working very hard to deliver it? And as a bonus question, will you still think it's not worth the $$ drop after Fermi comes out?

 

[v8envy]

Neither PhysX nor DX11 are relevant for the sub-$100 value cards which make up the bulk of sales. IMO the $70-120 market was all NV in Q4 09. Nicely priced G92 cards as well as the heavily marketed if underperforming 2XX 40nm cards. The 4850 and 4870 supplies were scarce, pushing prices well over $100.

 

[Wreckage]

First, what he meant was that (as he clearly said and you completely ignored), Radeon 5000-series beat or equal their nVidia competitors for performance AND are cheaper and less power-hungry AND have more features at each price point. All at the same time. The features was really the smallest part of his argument.

A 5770 is generally slower than a GTX260 and the 5750 is slower than a GTS250
http://www.techreport.com/articles.x/17747/1

Also, if DX 11 is fluff why is nVidia working very hard to deliver it? And as a bonus question, will you still think it's not worth the $$ drop after Fermi comes out?

When Fermi comes out DX11 will not be a reason to upgrade. PhysX, 3D and the other features it's bringing will be more compelling. DX11 will not take off for awhile (or maybe not at all thanks to the 360).

Neither PhysX nor DX11 are relevant for the sub-$100 value cards which make up the bulk of sales. IMO the $70-120 market was all NV in Q4 09. Nicely priced G92 cards as well as the heavily marketed if underperforming 2XX 40nm cards. The 4850 and 4870 supplies were scarce, pushing prices well over $100.

I agree with this. The lower end DX11 cards don't have the power to use DX11 (just as you won't be running Physx on a single GT240).

 

[digitaldurandal]

While AMD may not have really changed how market share is divided with the 57xx/58xx cards, I have to think the average selling price has gone up for them. If you can keep idential market share but introduce parts that sell for more money and presumably more profit, that'd still be a very good thing I'd think.

This.

Unfortunately, since I bought a 5870. That's what impatience gets me I suppose. I believe they will lower costs to be under Fermi when it is released and once again be bargain winners.

I think their current prices are justified because they lack real competition for this gen, theyre competing with nvidia cards that simply have less features.

 

[evolucion8]

A 5770 is generally slower than a GTX260 and the 5750 is slower than a GTS250
http://www.techreport.com/articles.x/17747/1

http://www.anandtech.com/video/showdoc.aspx?i=3658&p=5 <Look at there, does 2fps really make the difference?

http://www.anandtech.com/video/showdoc.aspx?i=3658&p=6 Or here, a heavily nVidia favored game, and yet the HD 5770 trade blows with the GTX 260, proves that the HD 4870 in the end was the better buy when it was launched since it is faster than both.

I agree with this. The lower end DX11 cards don't have the power to use DX11 (just as you won't be running Physx on a single GT240).

The same with the 9800GT which cannot run Dark Void with PhysX on without a huge impact in performance, GTX 260+ fares better, yet who will buy a low end DX11 card to max out games? Who will buy a GTS 240 for games and physx at the same time? No matter what you think, AMD simply has the better lineup with much more advanced features, better performance and is not in love with propietary technology who won't take off because after all, PC gaming is a microsoft platform, not an nVidia platform.

How many GPU accelerated PhysX games exist? Less than 10 games, how many DirectX 11 games are here already and in development? More than 20, not bad considering that PhysX has been in the market for more than 3 years and DirectX11 less than 4 months.

 

[Wreckage]

How many GPU accelerated PhysX games exist? Less than 10 games, how many DirectX 11 games are here already and in development? More than 20, not bad considering that PhysX has been in the market for more than 3 years and DirectX11 less than 4 months.

LOL so you use "in development" for DX11 to boost the numbers.

Only 2 available, one a patch.

 

[Aaluran]

LOL so you use "in development" for DX11 to boost the numbers.

Only 2 available, one a patch.

For 4 months. That's the point. In just a 9th of the time span (if PhysX is really 3 years old) it has more than a 5th of PhysX's working titles. With more to come.

 

[crisium]

GTS 250 beats a 5750? What? Maybe if cherry picking and looking at launch drivers.

Anything new shows the 5750 is clearly ahead with AA. And let's not even consider how much more of a &#37; gain the 5750 can get from overclocking than a GTS 250 can. Oh I'll cede that the GTX 260 is generally a tiny bit faster than a 5770, but the 250 cannot match a 5750.

Newer drivers. Just compare the 5750 to 250. Generally the 5750 is slightly faster, besides a few exceptions of course.

http://techpowerup.com/reviews/MSI/HD_5670_1_GB/5.html
http://techpowerup.com/reviews/MSI/HD_5670_1_GB/30.html

Mainly at higher res, the 5750 starts to walk away. I hope I don't have to prove to you that a 5750 will overclock higher.