Skylake Core Configs and TDPs

[frozentundra123456]

Lets try some more recent numbers instead of 2010:

A lot changed since 2010 as you use as latest numbers. Your numbers exclude the entire SB series and up. And while somewhat stagnant numbers from 1998 to 2010. Its gone downhill since:
http://jonpeddie.com/press-releases...-market-down-in-q1-nvidia-holds-market-share/

And IC design, R&D and wafer cost is only going up.

dGPU shipments are dropping fast.

If you look at the last graph that you yourself provided, the biggest decline in AIB shipments was in the most recent quarter, and that was only a one or two percent, so I would hardly call that "dropping fast". In fact, with the exception of one quarter, the sales is almost constant from 2012 to the present.

 

[cytg111]

They're not after the 500w monsters with iGP

People don't buy things on pure performance numbers. They buy things in order to do given things. Once they're doing iGPU's which can do 'reasonable' gaming at 1900*1080 (750ti/270 levels for instance), that's a huge threat to the market for dGPUs.

I dont see it. This will only happen if the gaming segment stands still (thus performance requirements), with the occulus and clones coming out, rebirth of VR and augemented reality, holographics, 4K gaming, whatever.
That an IGP should be "good enough" is about analog to the reasoning that 640K should be enough for everyone.. And it would be enough... if innovation stood still. Thank god it does not ..

 

[rtsurfer]

But CPUs are a lot different from GPUs. Increasing GPU performance is easy: just ass more EUs, CUs, SMMs. Adding more cores won't increase single threaded performance.

Intel is doing some things Nvidia and AMD can't to catch up. They're improving their outdated microarchitecture to dGPU levels. Making a bad architecture good is easier and higher rewarding than making a good architecture very good. And they're using their process advantage to brute force their way into performance territories that would otherwise have been impossible.

How many times you think Intel can do thier miraculous performance increase, once or twice. ?

You can't keep on increasing performance by more than 50% every year.
That doesn't happen.

Oh I fully agree. But you could also say the same about nVidia with their Tegra. However neither AMD or nVidia will ever reach that point vs competition. That train simply left long ago for both of them.

Well unless Nvidia Tegra scores a flagship smartphone win, its useless.

Shield isn't popular.
How many cars are they going to sell with K1 in it.?
100000 isn't enough.
Google has a dev Tablet, but unless they include Tegra in the next Nexus 7. It won't sell enough.
They need to score at least 1 flagship smartphone or more.

 

[witeken]

How many times you think Intel can do thier miraculous performance increase, once or twice. ?

You can't keep on increasing performance by more than 50% every year.
That doesn't happen.

The reason they are/were so far behind is because their microarchitecture sucks. If they get that on par with AMD and Nvidia, it just becomes a matter of how much die area they spend and how much they will benefit from smaller nodes (and for the consumer also the price).

 

[blackened23]

I dont see it. This will only happen if the gaming segment stands still (thus performance requirements), with the occulus and clones coming out, rebirth of VR and augemented reality, holographics, 4K gaming, whatever.
That an IGP should be "good enough" is about analog to the reasoning that 640K should be enough for everyone.. And it would be enough... if innovation stood still. Thank god it does not ..

Are you arguing IGP from the perspective of desktop or mobile? Are you yet again arguing solely from the perspective of PC gaming? IGP was created initially for mobile and is thrown onto desktop SKUs as a free bonus. But the design intent was never for desktop. It was not solely for gaming. It was always for superior battery life and performance per watt for mobile products.

And if you look at the vast majority of ultrabooks, most of them are using intel HD graphics. Nearly all of them are. And these products, you gotta be kidding yourself if you think every student and every corporate worker that buys an ultrabooks want to game. Most of them don't game.

You must keep in mind that most people are not gamers. If I had a dollar for every time some ridiculous argument aumomatically gets thrown into a "gaming" mindset, i'd be a millionaire....that isn't the sole design intent of IGP. Intel created the iGPU at Apple's beckoning initially and gaming obviously isn't the strong suit of the macbooks. That said, intel has gained an incredible amount of performance with Iris Pro even WITH gaming. The MOBILE IRIS PRO has great performance in games considering the die size investment and has very very good performance per watt.

That is ultimately what IGP is all about. MOBILE. Stop thinking in terms of desktop. If you want desktop graphics, get real man. You should just get a dGPU and be done with it. If you want 4k gaming, on a desktop, you get a dGPU.

This is not the original design intent of IGP and never was. HOWEVER, with that being said, intel has made incredible strides here even with gaming performance. No, you can't run 1080p ultra. 1366x768 with decent detail gaming? Yes, doable with Iris Pro. 30 fps low detail 1080p with Iris Pro? Yes, possible. The mobile core with Iris Pro is actually trading blows with AMD's desktop LGA in graphics performance. Which is funny when you think about it, because the 48xx MQ with Iris Pro is mobile while AMD's 7850k is a desktop part. A mobile part trading with a desktop part. You tell me if that should happen.

While the Iris Pro mobile also has (obviously) superior performance per watt. That gap will only widen in the future while AMD is making baby steps in graphics APU performance. And that will put AMD further and further behind in the mobile market. AMD's prospects for mobile, basically, suck. Sorry, but that's just the truth. OEMS want balanced performance and intel overall is just better for mobile in terms of balanced products. And no, 100% of buyers are NOT gamers. Despite that, Intel's gaming performance has just gone way up with Iris Pro. I see a day when high end iGPU actually intersect with low end mobile dGPU. 10 years ago every notebook had a mobile dGPU. Now? Nearly all of them don't.

 

[Erenhardt]

iris pro is not efficient... saying it is 100 times won't change the facts...
Iris pro is the biggest waste of transistors of all time...

 

[cytg111]

Are you arguing IGP from the perspective of desktop or mobile? Are you yet again arguing solely from the perspective of PC gaming?.

- Auch, that burns man . Well sort of desktop yes, the angle I am coming from is where Shintai is implying that intels igps is in the process of edging out dgpu's in general, that it is essentially a dead segnment (descrete). That's about it. I sincerly concur with the rest of your analysis.

 

[Sweepr]

iris pro is not efficient... saying it is 100 times won't change the facts...
Iris pro is the biggest waste of transistors of all time...

If they can do that with an outdated Gen 7.5 architecture @ 22nm, let's see what Gen 8/Gen 9 @ 14nm brings next year.

 

[Erenhardt]

Dunno what is says. Here:

Worth nothing: power deltas
iris pro 80-13=67W
a10-7850k 87-33=54W

So I guess a10-7850k could be a mobile part just as well as its desktop

 

[Homeles]

Dunno what is says. Here:

The important bits are all there in English. This isn't rocket surgery.

It's a power consumption graph, comparing the 4570R to the A10-7850K, and the AMD processor loses miserably.

 

[Sweepr]

Dunno what is says. Here:

Let me help you with this, it says the Core i5 4570R system tested draws less power than the A10 7850K system yet it delivers similar performance in most games (better/worse depending on game and resolution tested). The fact that it's an all around faster CPU is a nice bonus. By the way, that's a 65W BGA desktop part. While Intel's Gen 7.5 has a lot to improve, you are blinding yourself if you think their current IGPs are ''worthless'' and they are standing still (despite 14nm and two new architectures already on the way).

Core i5 4570R delta: 67W vs A10 7850K delta: 88W - Very close to their TDPs.

 

[witeken]

The system draws less power, but the processor doesn't.

 

[mikk]

Gen 7.5 is poor with MSAA, bad ROP performance. Gen8 should do much better there, almost twice the sampler per EU according to the Hotchips presentation. Also Gen 7.5 can't do 2xMSAA in hardware whilst Gen8 can do. Intel has lots of potential to improve with Gen8 and Gen9. Gen 7.x is really outdated now.

 

[blackened23]

- Auch, that burns man . Well sort of desktop yes, the angle I am coming from is where Shintai is implying that intels igps is in the process of edging out dgpu's in general, that it is essentially a dead segnment (descrete). That's about it. I sincerly concur with the rest of your analysis.

Yeah, I can't agree with discrete GPU dying for desktop. That will take several years, there WILL be a market for dGPU even if it slowly shrinks. And that it has, but not by much - and profit margins seem to have gone up on the whole for dGPU since prices have inched upwards. So to tag on your point, from a desktop perspective, I would agree. iGPU doesn't matter for desktop, unless you're talking extreme mini SFF or AIOs. There, iGPU is a good fit but for a traditional desktop ? I don't see iGPU or APU replacing desktop dGPU for many years. The performance delta between the two is too large, and iGPU / APU won't catch up anytime in the next few years.

Mobile is quite a different story though - I can see a day where iGPU will intersect with mid range mobile dGPU at some point and time. That time is not here quite yet, but it will get here. That will be the real turning point for mobile dGPUs and will quite limit NV even though they've had great success with selling mobile dGPUs. And AMD of course, but they haven't done much in terms of mobile dGPU. Perhaps their tonga whatchamacallit could be a game changer for them - they need a more efficient uarch for mobile dGPU graphics. Hopefully tonga can do that for them. Despite my skepticism towards AMD APU/CPUs I do like them being around and striving for robust dGPU graphics competition and all of the consumer net benefits associated with that. Lower prices for mobile dGPUs, etc.

But from a desktop perspective, I agree with you completely. We are many years away from iGPU being remotely able to compete with dGPU. That day may arrive, but it will be some years from now.

 

[Qwertilot]

4k adoption/VR? Yes, very good reasons to expect enthusiast dGPU's to be healthy enough going forwards

The thing is though, there's definitely going to be quite a decent window for iGPUs to take large chunks of the 1900*1080 GPU market if they can reach full 'usable' levels of performance quite soon.

And that'll be the large majority of the market for maybe 5(+?) years.

If you look at it another way, you might wander why Intel couldn't beat the consoles rather easily given the process advantage. They're on iGPU's of course.

 

[Homeles]

Gen 7.5 is poor with MSAA, bad ROP performance. Gen8 should do much better there, almost twice the sampler per EU according to the Hotchips presentation. Also Gen 7.5 can't do 2xMSAA in hardware whilst Gen8 can do. Intel has lots of potential to improve with Gen8 and Gen9. Gen 7.x is really outdated now.

Gen8 also gets 16x MSAA support.

Edit: looks like I might be remembering things incorrectly.

 

[ShintaiDK]

If you look at the last graph that you yourself provided, the biggest decline in AIB shipments was in the most recent quarter, and that was only a one or two percent, so I would hardly call that "dropping fast". In fact, with the exception of one quarter, the sales is almost constant from 2012 to the present.

Look at the black baseline. And if you move it 1 year forward to the SB release. It only gets worse.

And for each new IGP release it only gets worse for dGPUs. And with eDRAm getting more widespread, it will accelerate it further. Not to mention the entire new GT4 class.

 

[ShintaiDK]

I dont see it. This will only happen if the gaming segment stands still (thus performance requirements), with the occulus and clones coming out, rebirth of VR and augemented reality, holographics, 4K gaming, whatever.
That an IGP should be "good enough" is about analog to the reasoning that 640K should be enough for everyone.. And it would be enough... if innovation stood still. Thank god it does not ..

I think you focus on the wrong thing. Its not about IGPs being as fast as the best dGPU or anything. Its about when there is no money developing new dGPUs due to ROI.

 

[mikk]

Gen8 also gets 16x MSAA support.

Edit: looks like I might be remembering things incorrectly.

You are right with 16xMSAA, although not important.

Broadwell also supports 2x and 16x multisampling, in addition to the 4x and 8x supported by Gen7.

http://lists.freedesktop.org/archives/mesa-commit/2013-November/046098.html

 

[Homeles]

You are right with 16xMSAA, although not important.


http://lists.freedesktop.org/archives/mesa-commit/2013-November/046098.html

It's nice to have the option.

 

[Fjodor2001]

I think you focus on the wrong thing. Its not about IGPs being as fast as the best dGPU or anything. Its about when there is no money developing new dGPUs due to ROI.

Intel will never increase iGPU performance even close to dGPU performance, simply because it would just be wasted silicon for the majority of users that don't require so much from their iGPU. Heck, they are already fast enough for most users anyay.

And the ones that actually do use the dGPU performance today (gamers, video editing, ...) will still want the fastest GPU available in the future. iGPUs will not be good enough (simply because there is no economics in making iGPUs that good as mentioned in the previous paragraph). So they'll keep buying dGPUs and hence will be a market and enough ROI in that business. That is unless the PC gaming segment (and similar) will be shrinking, which I don't think we'll see happening.

 

[NTMBK]

One does not strike where the enemy is strong.

One strikes where the enemy is weak.

That's why I said "cut their legs from under them."

Nvidia has inflexible per unit costs from TSMC.

Intel can sell the cards at cost without getting in trouble for dumping at below cost and it will be far below what Nvidia's cost structure can handle.

Since Intel dgaf about the dedicated gpu market, almost-dumping is the best way to get rid of a competitor.

The problem with almost-dumping is that it trashes Intel's margins- and investors really don't like that.

 

[NTMBK]

Intel will never increase iGPU performance even close to dGPU performance, simply because it would just be wasted silicon for the majority of users that don't require so much from their iGPU. Heck, they are already fast enough for most users anyay.

This is why they make multiple different GPU sizes. On Haswell they already have GT1, 2 and 3, with GT3 being 4 times larger than GT1. And Skylake brings in GT4!

 

[AtenRa]

http://jonpeddie.com/press-releases...-market-down-in-q1-nvidia-holds-market-share/



Intel rapidly increased its share in 2013 for example and dGPU shipments drop fast. Like it or not, dGPUs are dying.

We have already talked about that,

Both AMD and NVIDIA ASPs were higher than before the last Quarter. They loosing entry level sub $100 dGPU shipments but they increase their Higher-End dGPUs shipments.
Yes the new AMD and Intel APUs/iGPUs have a negative effect in sub $100 dGPU shipment volume but both AMD and NVIDIA make enough from the higher-end to sustain their $100+ dGPUs development.

 

[AtenRa]

Just to add another graph,
Total GPUs shipped in 2013 are more than the combined total shipments of Desktop and Laptop PCs.
We can also see that both Desktop and Laptop shipments are decreasing from 2010 onwards and that makes dGPUs shipments also decreasing.