GPU Market Sees Year-over-Year Growth, AMD Grows Significantly

[btarunr]

The latest edition of Jon Peddie Research covering the PC graphics industry indicates that shipments of GPUs in Q1 2010 increased 44.3% over Q1 2009. There are increases in market shares of both AMD and NVIDIA, year over year, this time at the expense of Intel. The highlight for this quarter, however, has been the significant increase in market share of AMD. The company behind ATI Radeon clocked a 96.3% growth year-over-year. Despite taking a cut, Intel continues to to have the highest market share. With the introduction of Intel's Clarkdale and Arrandale processors that have embedded graphics processors, JPR demarcated these into a new entry called integrated processor graphics (IPG), from chipset graphics. With the propagation of the platform, it is predicted that chipset graphics will diminish in market share.



View at TechPowerUp Main Site

 

[Jstn7477]

I'm surprised to see that SiS is still around.

Anyways, AMD is looking good.

 

[Loosenut]

Interesting news, with AMD having a bigger slice of the pie, maybe we'll see more price wars in the near future. Anyways, good for us, the consumers.

 

[FordGT90Concept]

Intel is losing market share to AMD and NVIDIA. A shame NVIDIA is still over AMD though (NVIDIA has done nothing right for two years running).

 

[Loosenut]

Intel is losing market share to AMD and NVIDIA. A shame NVIDIA is still over AMD though (NVIDIA has done nothing right for two years running).

Bad management maybe?

 

[gaximodo]

Lol ironically while I open this page my pc got locked up and keeps getting BSODs cause of graphics card, scaried the crap outta me, Just fixed a min. ago by reinstalling catalyst

 

[holy_]

Nice to see AMD/ATI grows!

Hope their processor market share grow too

 

[HalfAHertz]

I'm surprised to see that SiS is still around.

Anyways, AMD is looking good.

They're still around in the server space

 

[DaedalusHelios]

Intel is losing market share to AMD and NVIDIA. A shame NVIDIA is still over AMD though (NVIDIA has done nothing right for two years running).

Nvidia has done many things right as far as creating a new architecture. It just needs to be refined. Their pricing is bad lately, but other than that they are just 2nd place in gaming. Is it that bad? ATi holds second place from time to time I have heard.....

Seriously lets not turn this into another ATi love fest/Nvidia hate fest. Lets enjoy both companies for what they bring to the table and encourage them to serve their customers to the fullest while innovating where they can. ATi has made multi-display gaming reach new heights with eyefinity, and Nvidia has pushed its scientific Tesla series to new heights in its respective fields. Remember that these companies make more than just GPUs. Nvidia even makes Tegra products which could grab more profit overall than the Tesla lines. Tegra could power phones to have much more powerful games and higher resolution screens down the line. Battery life could improve more as they make progress in reducing power consumption and allowing it to have an idle state later on.

 

[FordGT90Concept]

Bad management maybe?

Very.

Nvidia has done many things right as far as creating a new architecture.

Doesn't look like it to me. It takes twice the power as AMD and barely keeps up despite being on the same fab. That screams inefficient architecture to me. Not to mention all the dying GeForce 8 series cards, their numbering scheme that has been nuked, extremely late implementation of DirectX 10.1 specifications, and all the trash talk their management is spewing (at OEMs, at AMD, at Intel, etc.). NVIDIA should have been made changes after every mistake listed but they did not. AMD will continue to chip away at them until they wise up.


Intel is probably losing market share faster because of moving the integrated GPU to the CPU die. OEMs that want to offer integrated graphics, therefore, are more likely to ship AMD systems with motherboard integrated graphics rather than Intel with CPU integrated graphics--cheaper to troubleshoot/replace bad parts.


NVIDIA basically got shut out of the integrated graphics sector which is why their GPU sales fell some 20% of the past few years. There is little/no appeal in their discreet products either so I'm surprised it isn't lower.

 

[FreedomEclipse]

the Problem with SiS is that the IGP market has taken off significantly & SiS's failure to break out of the OEM Chipset/IGP market spells total fail...back in 2000, Intel & AMD both might have needed SiS to make chipsets with IGP for motherboards but those days are long gone & both AMD & Intel make their own.

SiS still make controller chips for other hardware though, Id be interested to see what sort of profits they pulling in per year. I dont even understand why they still exist. their probably just scraping by

 

[mdm-adph]

Doesn't look like it to me. It takes twice the power as AMD and barely keeps up despite being on the same fab. That screams inefficient architecture to me. Not to mention all the dying GeForce 8 series cards, their numbering scheme that has been nuked, and all the trash talk their management is spewing (at OEMs, at AMD, at Intel, etc.). NVIDIA should have been made changes after every mistake listed but they did not. AMD will continue to chip away at them until they wise up.


Intel is probably losing market share faster because of moving the integrated GPU to the CPU die. OEMs that want to offer integrated graphics, therefore, are more likely to ship AMD systems with motherboard integrated graphics rather than Intel with CPU integrated graphics--cheaper to troubleshoot/replace bad parts.


NVIDIA basically got shut out of the integrated graphics sector which is why their GPU sales fell some 20% of the past few years. There is little/no appeal in their discreet products either so I'm surprised it isn't lower.

As much as I, a long time ATI fan, love seeing Nvidia release crappy space heaters once again, I really, really hope their discrete GPU business stays solvent and they don't become some sort of ARM Tegra-only business. That would leave just ATI around, and since ATI is attached to AMD now, that would leave the only remaining great GPU company also competing with Intel and, though Intel doesn't have discrete GPU's, Intel does have the power through its monopoly to push AMD out of markets.

 

[DaedalusHelios]

Doesn't look like it to me. It takes twice the power as AMD and barely keeps up despite being on the same fab. That screams inefficient architecture to me. Not to mention all the dying GeForce 8 series cards, their numbering scheme that has been nuked, and all the trash talk their management is spewing (at OEMs, at AMD, at Intel, etc.). NVIDIA should have been made changes after every mistake listed but they did not. AMD will continue to chip away at them until they wise up.


4. Intel is probably losing market share faster because of moving the integrated GPU to the CPU die. OEMs that want to offer integrated graphics, therefore, are more likely to ship AMD systems with motherboard integrated graphics rather than Intel with CPU integrated graphics--cheaper to troubleshoot/replace bad parts.


5. NVIDIA basically got shut out of the integrated graphics sector which is why their GPU sales fell some 20% of the past few years. There is little/no appeal in their discreet products either so I'm surprised it isn't lower.

1. Nvidia's management was trading insults with other companies due to inflamatory websites. Reacting to insults isn't much of a surprise. Of course they could have ignored it a refuted it with facts but that is not like Nvidia or many other powerful people for that matter.

2. All new architecture needs a die shrink or other changes since creating a new architecture is a rebuilding phase. Think NFL, when you build a new team from scratch there is a rebuilding phase. In this case, the team is the GPU.

3. Dying 8 series cards? 4 series ATi had more DOA issues than 8 series Nvidia as far as I have seen. If we are talking G92 it ran cooler than 4 series products. If you are comparing G80 to ATi's offerings then you would be speaking of the failure that was the 3-series.

4. The durability of a CPU is greater than an AMD integrated graphics chipset. Less about troubleshooting and more about failure rate.

5. Nvidia stopped making chipsets thanks to getting blackballed by Intel due to a lack of info sharing related to new sockets. AMD and Intel both made bundle deals where you buy a CPU and get the motherboard for free to OEMs like Dell and HP. Also made sure to fight Ion development by withholding info followed by bundling CPU with 945 chipsets where the chipset came free to board partners. That is just the start of why they had to pull out.

 

[FordGT90Concept]

As much as I, a long time ATI fan, love seeing Nvidia release crappy space heaters once again, I really, really hope their discrete GPU business stays solvent and they don't become some sort of ARM Tegra-only business. That would leave just ATI around, and since ATI is attached to AMD now, that would leave the only remaining great GPU company also competing with Intel and, though Intel doesn't have discrete GPU's, Intel does have the power through its monopoly to push AMD out of markets.

I agree. I don't want NVIDIA gone but I want them to fix the massive mess they made (generally, that means new leadership).


1. Answering an insult with an insult is childish.

2. NVIDIA, as a corporation, their highest priority is profits. Releasing a new architecture before the process is ready is foolish as it costs a lot of money, produces excessive heat, consumes excessive power, and has little to show for it. Moreover, it was a very nearsighted, horrible decision to focus a new GPU design on GPGPU work. GeForce products are the bulk of NVIDIA's profits, not Tesla.

3. It's the G92 and G86 cards that are failing left and right. G92b cards have a lower rate of failure.

4. It's more difficult to troubleshoot and more expensive to repair (if the IGP died).

5. NVIDIA refused to update their license for the QuickPath/DMI architecture. NVIDIA has no one to kick but themselves.

 

[btarunr]

2. NVIDIA, as a corporation, their highest priority is profits. Releasing a new architecture before the process is ready is foolish as it costs a lot of money, produces excessive heat, consumes excessive power, and has little to show for it. Moreover, it was a very nearsighted, horrible decision to focus a new GPU design on GPGPU work. GeForce products are the bulk of NVIDIA's profits, not Tesla.

GPGPU is what helps sell GeForce.

3. It's the G92 and G86 cards that are failing left and right. G92b cards have a lower rate of failure.

Not G92, only G86.

NVIDIA refused to update their license for the QuickPath/DMI architecture. NVIDIA has no one to kick but themselves.

Intel refused to give NVIDIA QPI tech, saying their chipset agreements don't cover processors with IMCs. (by logic processors with IMCs have QPI). NVIDIA never had a "DMI" license. In nForce 700i and below, the SPP (northbridge) connected to the MCP (southbridge) using an 8-bit HyperTransport link (with 8 GB/s bandwidth, four times that of DMI).

DMI is just a PCI-Express 1.1 x4 link with some low-level paraphernalia (data transfer code).

 

[DaedalusHelios]

Intel refused to give NVIDIA QPI tech, saying their chipset agreements don't cover processors with IMCs. (by logic processors with IMCs have QPI). NVIDIA never had a "DMI" license. In nForce 700i and below, the SPP (northbridge) connected to the MCP (southbridge) using an 8-bit HyperTransport link (with 8 GB/s bandwidth, four times that of DMI).

DMI is just a PCI-Express 1.1 x4 link with some low-level paraphernalia (data transfer code).

Thank you for that information. I didn't know about the QPI tech, but it doesn't surprise me as Nvidia and Intel have been not on the best of terms for quite some time.

 

[FordGT90Concept]

GPGPU is what helps sell GeForce.

I don't know about that. Ask some random person on the street if they have a GeForce video card and if they answer, yes, ask them why they got it. 9/10 will say because I game. The only ones that care about GPGPU are those that have folding farms (few and far between). Everyone knows PhysX is useless because there's not even two dozen games that support hardware PhysX acceleration--most of which are C-list titles. There's no money in programming for proprietary platforms.

Not G92, only G86.

I had a 8800 GT throwing random BSODs. I'm not alone. There's half as many hits with 9800 GT (90,000), half that for 7800 GT (45,000), and half that for 7950 GT (25,000).

I also had a 8500 GT blow 3 of 4 capacitors. There's actually more hits for 8800 GT (190,000) failure than 8500 GT (130,000).

Intel refused to give NVIDIA QPI tech, saying their chipset agreements don't cover processors with IMCs.

NVIDIA picked a fight over the licencing:
http://www.electronista.com/articles/09/02/18/nvidia.intel.license.fight/

And later decided their chipset business isn't worth the fight:
http://www.tomshardware.com/news/nvidia-geforce-intel-core-i7,7624.html

Was that a bad decision? NVIDIA isn't going to get much chipset business from AMD processors because of ATI branded chipsets. NVIDIA never got much chipset business from Intel processors because of Intel branded chipsets. NVIDIA won't get any chipset business from AMD or Intel soon as the chipsets move to the CPU so, I redact my previous statement: giving up on chipsets was probably the one smart thing NVIDIA did last year.

 

[btarunr]

I don't know about that. Ask some random person on the street if they have a GeForce video card and if they answer, yes, ask them why they got it. 9/10 will say because I game. The only ones that care about GPGPU are those that have folding farms (few and far between). Everyone knows PhysX is useless because there's not even two dozen games that support hardware PhysX acceleration--most of which are C-list titles. There's no money in programming for proprietary platforms.

GPGPU is turning out to be a significant feature-set in the GPU competition. People buy GeForce cards over Radeon because they perform exponentially better at Folding@Home, have a broader range of applications that use CUDA than OpenCL/Stream, and PhysX (for whoever takes it).

I had a 8800 GT throwing random BSODs. I'm not alone. There's half as many hits with 9800 GT (90,000), half that for 7800 GT (45,000), and half that for 7950 GT (25,000).

Many people had problems with Radeon HD 4800 series, too.

The foundry fiasco which made NV spend over $200m in recalls and refits revolved solely around G86/G84 GPUs, not G92. G86's hits are lower because failures are dealt with between the customer and companies like Dell, HP, etc., people don't usually make threads / blogs about them. G92 sold a ton, and there were no incidents where NV had to recall a batch of them from OEMs.

NVIDIA picked a fight over the licencing:
http://www.electronista.com/articles/09/02/18/nvidia.intel.license.fight/

And later decided their chipset business isn't worth the fight:
http://www.tomshardware.com/news/nvidia-geforce-intel-core-i7,7624.html

Right, and that doesn't disprove what I said. The reason NV has no QPI is because it can't make chipsets for processors that have IMCs (because the cross-licensing agreement doesn't cover CPUs with IMCs (Intel's contention)).

 

[FordGT90Concept]

GPGPU is turning out to be a significant feature-set in the GPU competition. People buy GeForce cards over Radeon because they perform exponentially better at Folding@Home, have a broader range of applications that use CUDA than OpenCL/Stream, and PhysX (for whoever takes it).

Same thing: walk up to someone and ask if they have a GeForce and if they bought it for CUDA. Most won't even know what CUDA is. NVIDIA wants people to think CUDA/PhysX is important because they sunk their whole business in to it when the exact opposite is true. This is one reason why NVIDIA's market share has been steadily shrinking while AMD's grows.

Many people had problems with Radeon HD 4800 series, too.

About equal to 8800 GT. So, 8800 GT and HD 4800 had about the same amount of trouble, fair enough; still, NVIDIA had problems with more than just one GPU (GeForce Go chips, G92, G86, and G84). AMD had substantially less overall in the past few years.


I wasn't trying to disprove what you said on the last bit, just explaining how that was really their only coarse of action (get out of the chipset business) and another reason why their GPU market share is shrinking.