• We should now be fully online following an overnight outage. Apologies for any inconvenience, we do not expect there to be any further issues.

Intel Iris & Iris Pro Graphics: Haswell GT3/GT3e Gets a Brand

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

sontin

Diamond Member
Sep 12, 2011
3,273
149
106
What's there? 900MM isn't for GPU only as you stated, but GPU, software, VLSI, Content and Process R&D. Of those Tegra gets benefts of Software, VLSI, content and Process R&D, on top of the 300MM exclusively dedicated to Tegra.

Without all of these segments nVidia could not bring a GPU to the market. There is a reason why Kepler is so dominant.

When a product is responsible for less than 20% of your revenues, has negative operating margins but eats 25% of your R&D budget and benefits from a sizable portion of the other 75%, this is where you are betting your farm.
And we ignoring that they have much more new products in the pipeline which are not out right now. Tegra 4i and i500 are two products which have no preprocessor but need R&D. On the other hand: Intel's Silvermont has negative operating margins, too. ;)

But... You have to be very careful when reading Nvidia investor material. Reading Nvidia material is the business equivalent of racing against Dirk Dastardly.
I don't said anything about their marketing material. I only showed what they investing into their GPU business.

have a look at slide 15. Did you notice the nice CAGR number? Nice, isn't it? Did you notice what the higher numbers are? Yes, not GPU.
I guess that the reason why they are investing into Tegra...

But this is just the beginning. Did you notice that they chose FY10 (2009) for baseline? Yes, this is the first shennanigan here, because it was a very bad year for the GPU business, far worse than 2007 or 2008. Put those two and you may have close to 0% CAGR.
On the next slide you see all three years. :rolleyes:
BTW: 10 years ago nVidia's revenue was only around $500 millions per quarter.

But there is more. Still on slide 15, what about that CAGR number for Geforce Gamer and Geforce OEM? Sounds a healthy market, doesn't it? But once you cross reference the numbers with Nvidia overall market share and overall size market, you can see why they can get those nice numbers. Nvidia is eating AMD market share, as you can see in slide 19.
That's a problem for you? Ironic because:
The x86 market crashed? I didn't see that from Intel revenues. I saw the market shrinking, not crashing. The only thing crashing in the x86 market was AMD, but not the overall market.
On the same slide 19 they show us something new. They mention a 12% CAGR for their GPU business and mentions the 3.2 billion in revenues. This 3.2 billion in GPU revenues includes 300 million of Intel settlement, and this revenue is pratically net. Purge this and CAGR becomes far smaller.
Why should nVidia? The reason Intel is allowed to sell Haswell with a iGPU is because of the settlement. Maybe we should "purge" the iGPU from Haswell and discuss the threat of Intel to the low-end without an iGPU.

Once we purge this revenues we can verify that the overall market has slightly shrunk since 2009. Purge the revenues from the PSB business and you can see that Geforce business shrunk even more.
The Geforce business increased the revenue from last year, you can read it here: http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9MTcxNjMzfENoaWxkSUQ9LTF8VHlwZT0z&t=1

As I said, the downward trend in the dGPU market is there for everyone to see. Nvidia management is doing the right thing in running away from GPU. The market is shrinking, and the attack it is going to face from Intel and AMD isn't something they can support.
The market is shrinking, but not their revenue. It's no different to Intel which eating into AMD's share to get higher revenue.
On the other hand nVidia is expanding their GPU IP into more and more markets which will help them to growth further.

Intel CPU business isn't reporting losses because Haswell or Broadwell aren't generating revenues, same with Nvidia GPU business, they aren't reporting losses in their GPU business because Maxwell isn't generating revenues. This is to be expected, no? You fund R&D for future products with current products revenues.

"Current products revenues" without "current products"? How is this working?

And you can clearly see that on Nvidia financial statements. Nvidia didn't report an operating loss of $157MM for every business of them except GPUs, but for the Tegra processor business only, meaning that only what is directly related to Tegra is included as operating expense to get to this number.
They bought Icera for more than $300 millions, investing into LTE and integration of LTE into an application processor. Validate their LTE on different carrier networks. All this cost money. Maybe you should look up what "investment" means.

With the data they provided you can also have a glimpse of their sub-30% gross margins in the Tegra business... but let's leave that for another thread.
Pls, stop it. nVidia disclosed all numbers. Gross margins for Tegra products around 50%. That in the near of Intel's gross margins. Tegra attributes to the overall gross margins because it'S higher than the chipset business and the Geforce OEM market.
 
Last edited:

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
Talking about knowledge. Why dont you get your facts straight and stop doing selective reading to try and argue for your misguided visions?

Let me ask you again, what costed more. The 3820 or the 2700? What is the diesize of the 2?

Can we then put the moronic BS about diesizes and direct consumer cost to rest?

If diesize cost was so essential. Then that turd you keep promoting should cost a fortune.

I have never said that Intel could not price them lower, what i have said is that due to a bigger die and lower volume, selling them lower than GT2 Quad Core CPUs will get them lower margins and profits, what’s so hard to understand ??
 
Feb 19, 2009
10,457
10
76
Yes, and this can be entirely attributed to the improved iGPUs in Sandy Bridge, Ivy Bridge, Llano and Trinity. So extrapolate from there, what happens to the next segment up (lets say $100 to $150) when Intel and AMD bring out their even more powerful next gen APUs?

It's flat out wrong to assume its going to happen. Intel has killed $50 discrete, because these were rubbish to start with, the dies are tiny and the TDP is low. It can be matched by a iGPU easily because iGPU have similar specs.

If Intel wants to compete with mainstream cards with 125W TDP, or approaching mid-range with 200W TDP, they are going to have to start releasing monster APU dies >400mm2 and a TDP of 200W+ combined.

Sure, it can happen, but then it won't be CHEAP.. thus, it wont have the volumes to threaten anything.

tldr: Intel doesn't do magic, perf/mm2 and perf/w still is the defining factor to determine whether Intel can threaten a certain market of discrete GPU or not.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
I have never said that Intel could not price them lower, what i have said is that due to a bigger die and lower volume, selling them lower than GT2 Quad Core CPUs will get them lower margins and profits, what’s so hard to understand ??

It's a lot more complicated than that. Because Intel's actual ASPs(thus margins and profits) are on the low-$100 range, there's lot of flexibility and options for pricing on the higher end chips. There would be no point of pricing the GT3e $150 higher because than the volumes would be low and the impact of ASPs minimal. By having it only $50 higher, lot more buyers can be reached and the impact of ASPs greater so costs incurred by greater die will be easily recouped.

The extra cost may not be that much. There's a $20-30 lower price for the S-series 65W chips compared to the regular TDP chips. If rumored addition of $50 on GT3e is true, that'll make it only $20-30 higher.
 

inf64

Diamond Member
Mar 11, 2011
3,884
4,692
136
GT3e is supposedly ~50% faster than GT3 so how can they only ask 20-30$ more for chips that have GT3e? Intel always asks money for extra features and extra performance, this time they have valid reason to ask a lot more for GT3e parts as they will outperform, by a big margin, regular GT3 parts.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
GT3e is supposedly ~50% faster than GT3 so how can they only ask 20-30$ more for chips that have GT3e? Intel always asks money for extra features and extra performance, this time they have valid reason to ask a lot more for GT3e parts as they will outperform, by a big margin, regular GT3 parts.

There are no "regular" GT3 parts except for the U-series chips. Maybe you meant GT2?

GT3e needs to be competitively priced with similarly performing discrete chips, it doesn't matter how much it performs compared to GT2 if OEMs ignore it because of the price compared to using discrete.

Like I said, they put GT3e on the lower-clocked CPUs, which are normally quite a bit cheaper, and add graphics on that. If you ignore the graphics the R-series are basically S chips and H-series mobiles are basically successors to the OEM-only 36xxQM chips.
 

inf64

Diamond Member
Mar 11, 2011
3,884
4,692
136
Yeah I meant the GT2 on U series. I guess market is the best judge for any product so we can see how intel's pricing affects the adoption rate.
 

Piroko

Senior member
Jan 10, 2013
905
79
91
GT3e needs to be competitively priced with similarly performing discrete chips, it doesn't matter how much it performs compared to GT2 if OEMs ignore it because of the price compared to using discrete.
Yea, that is probably the plan they're pursuing. I'm just not that sure if its perceived value is as high as that of a 50$ cheapo-discrete. The advertising flyers of our local stores have cards like the GT630 printed in the same font size as the CPUs, anything beyond that and you can barely see the price compared to all the Nvidia logos (a bit dramatized).
 

SiliconWars

Platinum Member
Dec 29, 2012
2,346
0
0
The problem is surely that the R series are i7's. Who is going to pay $350 or more for a lower clocked, locked desktop i7 with 640M level graphics?
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
The problem is surely that the R series are i7's. Who is going to pay $350 or more for a lower clocked, locked desktop i7 with 640M level graphics?

Here's the R-series chips coming: http://www.xbitlabs.com/images/news/2013-04/intel_haswell_core_r_series_specs.png

The advertising flyers of our local stores have cards like the GT630 printed in the same font size as the CPUs, anything beyond that and you can barely see the price compared to all the Nvidia logos (a bit dramatized).
It's not even direct competition with the socketed chips as the R-series are BGA and will be going into AIOs anyway.
 

SiliconWars

Platinum Member
Dec 29, 2012
2,346
0
0
Thanks I didn't realise there were also i5's coming.

I'm not too impressed by the look of those (that all depends on the price obviously). Graphically I would guess Richland-level performance but at near double the cost, and a 2.7 GHz i5 is not going to be much faster than the 6800K even in CPU benchmarks.
 

Piroko

Senior member
Jan 10, 2013
905
79
91
It's not even direct competition with the socketed chips as the R-series are BGA and will be going into AIOs anyway.
I only know few AIOs, but those do come with low end Nvidia stuff. Or in the case of the 27" iMacs, with GTX660m chips.

Intel might hit a jackpot with those chips after all, dunno. It does make for some enticing products. I'm just not sure if the market digs them or not.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
I only know few AIOs, but those do come with low end Nvidia stuff. Or in the case of the 27" iMacs, with GTX660m chips.

Intel might hit a jackpot with those chips after all, dunno. It does make for some enticing products. I'm just not sure if the market digs them or not.

There's a few listings if you go to google and search for Core i7 3770S all in one. Lots of graphics options are there. HD 4000, Geforce GT 610(why??), Geforce GT 640, Geforce GT 650.

Personally I wouldn't pin my hopes on an AIO either, but I don't know what others think.
 

Fjodor2001

Diamond Member
Feb 6, 2010
4,225
590
126
Somehow I get the impression that the GT3e is just a marketing gimmick. Now when you search on Haswell you'll see a lot of articles claiming "Haswell delivers 3x GFX performance improvement". So then that's what sticks in the mind of the general public. But in reality there will be very few Haswell devices with GT3e sold. And the performance improvement of the GT2 which is what actually is what will be sold in bulk is not stellar.
 

Blandge

Member
Jul 10, 2012
172
0
0
Somehow I get the impression that the GT3e is just a marketing gimmick. Now when you search on Haswell you'll see a lot of articles claiming "Haswell delivers 3x GFX performance improvement". So then that's what sticks in the mind of the general public. But in reality there will be very few Haswell devices with GT3e sold. And the performance improvement of the GT2 which is what actually is what will be sold in bulk is not stellar.

... This is a very intelligent statement.

The only thing I'd add is that most mobile parts will have GT3 which is much better performing than GT2, so GT3 will probably see significant volume.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
The only thing I'd add is that most mobile parts will have GT3 which is much better performing than GT2, so GT3 will probably see significant volume.

Not really. The only really high performing GT3 is on the 28W SKUs for the Ultrabooks, and that'll only go in the bigger 14-15 inches. GT3e on the regular Notebooks is only for quad core, so that's low volume too.
 

Blandge

Member
Jul 10, 2012
172
0
0
Not really. The only really high performing GT3 is on the 28W SKUs for the Ultrabooks, and that'll only go in the bigger 14-15 inches. GT3e on the regular Notebooks is only for quad core, so that's low volume too.

What about standard notebooks that don't have GT3e?
 

jpiniero

Lifer
Oct 1, 2010
16,841
7,285
136
Their cheapest iMac is already equipped with a GT640m GDDR5, everything above it is GT650m or higher. I don't see them move their whole iMac line to GT3s, honestly. The Mac Minis might profit, but aren't those like a niche in a niche?

So it'd be marginally better at most... but that means Apple can make it thinner. Apple was instrumental in getting GT3e created, you better believe you will see it on the iMac.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
So it'd be marginally better at most... but that means Apple can make it thinner. Apple was instrumental in getting GT3e created, you better believe you will see it on the iMac.

Pretty much. There are numerous articles discussing intel's motivation to focus on graphics technology so much, including ones on AT - Apple's insistence has been a key motivator. As stated, you better believe that Apple will ditch discrete at the first possible chance. There really is no reason to get a discrete chip when it is only marginally better or the same in terms of performance.