Skylake Core Configs and TDPs

[Sweepr]

And ?? that only shows how far behind Intel iGPU architectures are.



There will be no new arch in Desktop/Laptops in 2014, Broadwell Y will only be used in Tablets this year. Laptop Broadwell SKUs will be release in Q1 2015, more than 6 moths from now.
Broadwell Desktop SKUs with new iGPU arch will only be released one year from now, in Q2 2015.
20nm dGPUs will also be release around this time or even earlier. So, i dont see why people believe that Intel iGPUs will progress faster than dGPUs.

Broadwell-Y targets premium convertibles, not exactly tablets, and nothing stops OEMs from releasing more conventional ultra-thin laptops based on it. Also, as far as I know Broadwell-U GT2 could still be available in late 2014 (Broadwell-U GT3 comes in Jan/Feb 2015).

 

[Fjodor2001]

I wonder how Intel intends to spin this if desktop Broadwell and Skylake actually is released at the same time in summer of 2015? Who would want to get Broadwell then?

Unless they artificially cap the Skylake frequency or something like that. But such a setup sounds really weird too.

 

[jpiniero]

I wonder how Intel intends to spin this if desktop Broadwell and Skylake actually is released at the same time in summer of 2015? Who would want to get Broadwell then?

Unless they artificially cap the Skylake frequency or something like that. But such a setup sounds really weird too.

It won't be the same time - Broadwell-K will be released a month or two earlier if the rumors are true. And the only Skylake models being released then are the desktop locked quad core 4+2 i5 and i7s... everything else will be several months later. There won't be much confusion since the models don't overlap.

 

[Sweepr]

Starting with Devil's Canyon I bet future ''K'' unlocked chips will operate at higher clocks than non-K models (>4GHz). They don't have to artificially cap anything. Whoever buys Broadwell-K gets similar (if not slightly better) CPU performance than lower-clocked & locked Skylake, more OC headroom (unlocked multiplier) and a massive GT3e Gen 8 iGPU.

 

[Fjodor2001]

It won't be the same time - Broadwell-K will be released a month or two earlier if the rumors are true. And the only Skylake models being released then are the desktop locked quad core 4+2 i5 and i7s... everything else will be several months later. There won't be much confusion since the models don't overlap.

Well a month or two really doesn't matter I think.

And how come they only release locked 4+2 i5 and i7s? Are they artifically capping the clock frequency? For what reason? Why not skip Broadwell-K completely and just release Skylake models as normal.

 

[Arachnotronic]

Well a month or two really doesn't matter I think.

And how come they only release locked 4+2 i5 and i7s? Are they artifically capping the clock frequency? For what reason? Why not skip Broadwell-K completely and just release Skylake models as normal.

Why would you skip Broadwell-K when the R&D is done and it can be sold?

 

[Fjodor2001]

Why would you skip Broadwell-K when the R&D is done and it can be sold?

Because Skylake-K is already done and they could release it if desired.

If they intentionally want to delay Skylake only to get ROI for Broadwell R&D investments, why not delay the complete Skylake line?

 

[Arachnotronic]

Because Skylake-K is already done and they could release it if desired.

If they intentionally want to delay Skylake only to get ROI for Broadwell R&D investments, why not delay the complete Skylake line?

There are some markets where Intel needs to push as quickly as possible and shorter-term ROI isn't as important as longer-term competitive positioning.

Intel can't give AMD even an inch if it wants to keep its high margins and that means continually beating its mainstream, (relatively) high volume "big core" APUs into a pulp performance wise if it can do so.

 

[erunion]

I wonder how Intel intends to spin this if desktop Broadwell and Skylake actually is released at the same time in summer of 2015? Who would want to get Broadwell then?

Unless they artificially cap the Skylake frequency or something like that. But such a setup sounds really weird too.

They will target different segments. Eg. skylake for mobile, broadwell for desktop.

 

[Fjodor2001]

So on desktop the logic is: The more you pay, the older CPU generation you get.

Pay most (Haswell-E) get Haswell, pay a bit less (Broadwell-K) get Broadwell, pay the least (Skylake-non-K) and get Skylake. All available at the same time.

Enthusiasts sure aren't any priority for Intel these days...

 

[voodoo7817]

Thanks Insert_Nickname, frozentundra123456, and jpiniero. You all shared ideas that reflect my thinking and led to the question.

I agree it's a good idea to at least wait for Haswell-E results to see how much 6 cores makes a difference on recently released games. Also to see if DDR4 pricing is reasonable or not. It would be nice if my new platform lasts another 4+ years, so if 6 cores will be necessary for that, it would be nice to know asap.

I do live near enough to a Microcenter though, so perhaps if/when they have one of their $200 4790k specials, I'll bite, as it doesn't get much better than that. I almost bit when they had the 4770k at that price.

Bringing this back around to Skylake, is there anything particularly amazing coming, platform-wise, on Skylake? For example, a new speedy and quickly adopted SSD interface? That could be another reason to wait, but if nothing special is coming, a Z97 would be a real nice improvement coming from a P55.

 

[mrmt]

The problem is neither of them is having any success in that. Both of them are having MPU divisions in free fall.

AMD had a functional bleeding edge CPU division and Nvidia wanted to create one. Today AMD doesn't have one, and Nvidia gave up that idea, but both players are now focusing on developing an embedded CPU division, using lagging edge CPU cores but bleeding edge graphics IP. If they can breach into this market, it will allow them to fund dGPU R&D for the next generations.

Don't be fooled by the numbers you are seeing now, both players seem to have good prospective paths here, but they certainly won't be competing against Qualcomm, Intel or Samsung.

 

[cytg111]

Because Skylake-K is already done and they could release it if desired.

If they intentionally want to delay Skylake only to get ROI for Broadwell R&D investments, why not delay the complete Skylake line?

- Yea, doesnt add up with skylake being 20-40% faster (clocks and ipc?) than previous gen. Even locked down skylake vs. broadwell-k it would be a weird match... skylake would have to be confined to tablets for an entire year to make sense of that roadmap..

 

[jpiniero]

- Yea, doesnt add up with skylake being 20-40% faster (clocks and ipc?) than previous gen. Even locked down skylake vs. broadwell-k it would be a weird match... skylake would have to be confined to tablets for an entire year to make sense of that roadmap..

Don't really have a good feel in IPC improvement for Skylake versus Broadwell but it probably will be closer to 5-10% as usual. Maybe if apps used AVX3 you might approach that.

Still expect clocks to be worse on Broadwell and Skylake compared to Devils Canyon.

 

[cytg111]

Don't really have a good feel in IPC improvement for Skylake versus Broadwell but it probably will be closer to 5-10% as usual. Maybe if apps used AVX3 you might approach that.

Still expect clocks to be worse on Broadwell and Skylake compared to Devils Canyon.

actually picked up 20-40% here

http://forums.anandtech.com/showpost.php?p=36427620&postcount=164

(which sounds awsome .. with 20-40% i'd be torn between that and haswell-e)

 

[Idontcare]

So on desktop the logic is: The more you pay, the older CPU generation you get.

Pay most (Haswell-E) get Haswell, pay a bit less (Broadwell-K) get Broadwell, pay the least (Skylake-non-K) and get Skylake. All available at the same time.

Enthusiasts sure aren't any priority for Intel these days...

Not enough of us who are willing to buy high ASP SKUs to fully amortize the development costs of the enthusiast products.

We get what we deserve in a truly free-market sense.

Most of us want to read reviews on how awesome $1k SKUs are, only to go build rigs with $200 K-chips in them and hold onto them for 2 or 3 yrs while continuing to live vicariously through reading Anandtech CPU reviews and tales from the one-off forum member here who splashes out for the spendy upgrade.

Look at Skulltrail, Intel lost a bundle developing that platform only to watch just a handful get bought. You can't keep justifying those kinds of internal projects when the sales don't follow.

The enthusiast segment has been dying on the vine for years now, and its only going to get worse as more and more "enthusiasts" opt to spend more per year upgrading their phone and phone plan than they will spend upgrading their "enthusiast" rig.

Intel is right to chase after the money, wherever it may be. If more of us enthusiasts put our money where our mouths are then Intel would be cost/benefit justified to push a more aggressive (read - more expensive to develop) enthusiast roadmap.

 

[frozentundra123456]

actually picked up 20-40% here

http://forums.anandtech.com/showpost.php?p=36427620&postcount=164

(which sounds awsome .. with 20-40% i'd be torn between that and haswell-e)

Think that means 20 - 40 percent better performance per watt. Don't think it means that much higher absolute performance. Kind of like ivy bridge vs sandy. Way lower power consumption, but not that much maximum performance increase.

 

[cytg111]

Think that means 20 - 40 percent better performance per watt. Don't think it means that much higher absolute performance. Kind of like ivy bridge vs sandy. Way lower power consumption, but not that much maximum performance increase.

from the article

Chipzilla is claiming 60% lower TDP for Broadwell with 20-40% better performance, 10-45% better SoC power, and a 50% reduction in package footprint.

If that's "20-40%" perf/watt we're talking about then that is some very creative wording ... and sort of redundant to the other numbers mentioned.

 

[jpiniero]

actually picked up 20-40% here

http://forums.anandtech.com/showpost.php?p=36427620&postcount=164

(which sounds awsome .. with 20-40% i'd be torn between that and haswell-e)

That's the U and Y models... most of that is probably clock speed. You won't be seeing that at the top end.

The enthusiast segment has been dying on the vine for years now, and its only going to get worse as more and more "enthusiasts" opt to spend more per year upgrading their phone and phone plan than they will spend upgrading their "enthusiast" rig.

Someone mentioned this awhile back, and it is so true... the 'enthusiast' market is almost entirely just gamers; and there is really not much reason to upgrade the CPU to play the latest titles. Especially with Mantle and DX12 on the horizon, it's not unthinkable that even something like a 6 year old 920@4 Ghz will last another 4-5 years and give reasonable frame rates.

 

[WaitingForNehalem]

Not enough of us who are willing to buy high ASP SKUs to fully amortize the development costs of the enthusiast products.

We get what we deserve in a truly free-market sense.

Most of us want to read reviews on how awesome $1k SKUs are, only to go build rigs with $200 K-chips in them and hold onto them for 2 or 3 yrs while continuing to live vicariously through reading Anandtech CPU reviews and tales from the one-off forum member here who splashes out for the spendy upgrade.

Look at Skulltrail, Intel lost a bundle developing that platform only to watch just a handful get bought. You can't keep justifying those kinds of internal projects when the sales don't follow.

The enthusiast segment has been dying on the vine for years now, and its only going to get worse as more and more "enthusiasts" opt to spend more per year upgrading their phone and phone plan than they will spend upgrading their "enthusiast" rig.

Intel is right to chase after the money, wherever it may be. If more of us enthusiasts put our money where our mouths are then Intel would be cost/benefit justified to push a more aggressive (read - more expensive to develop) enthusiast roadmap.

Actually, the enthusiast market is growing year over year and commands about 10% of total sales. You wouldn't see Devil's Canyon if wasn't growing.

 

[cytg111]

That's the U and Y models... most of that is probably clock speed. You won't be seeing that at the top end.

Mkay, higher clocks while keeping wattage down in mobile segment.. does not translate into topping out with the same momentum. Ok, guess i can see that! (still, im an ever ever optimistic optimist.. so here's to hoping)

 

[Insert_Nickname]

Bringing this back around to Skylake, is there anything particularly amazing coming, platform-wise, on Skylake? For example, a new speedy and quickly adopted SSD interface? That could be another reason to wait, but if nothing special is coming, a Z97 would be a real nice improvement coming from a P55.

Many Z97 boards already have M.2 slots and SATA Express. I doubt if you'll see anything more exotic with Skylake, except perhaps more PCIe lines dedicated to them. There are rumours floating around that Skylake will have a major upgrade to the DMI link (PCIe 3.0 capable?) between the CPU and PCH, and have -more- lines available from the PCH, and PCIe 3.0 support from it. Other then that, there is only possible PCIe 4.0 support from the CPU, but it'll likely be awhile before anything make use of that.

Personally I'm waiting for it, since I'm on a 3770non-K + Z77. So I can afford to wait and see...

 

[Idontcare]

Actually, the enthusiast market is growing year over year and commands about 10% of total sales. You wouldn't see Devil's Canyon if wasn't growing.

I have not seen any numbers to support such claims, where are you getting this info?

 

[AtenRa]

http://files.shareholder.com/downloads/INTC/3231551923x0x744135/c7a67741-b928-4f9c-819d-decb74512d69/Earnings%20Release%20Q1%202014_final.pdf

Last page

PC Client Group Notebook and Desktop Platform Key Drivers
- Notebook platform volumes increased 2% from Q1 2013 to Q1 2014
- Notebook platform average selling prices decreased 8% from Q1 2013 to Q1 2014
- Desktop platform volumes flat from Q1 2013 to Q1 2014
- Desktop platform average selling prices increased 4% from Q1 2013 to Q1 2014

Laptop Volume increased by 2% Q1 13 to Q1 14 but ASP fell by 8%, that means more Lower-End CPUs were sold.

Desktop volume flat Q1 13 to Q1 14 but ASP increased by 4%, that means more higher-End CPUs were sold than lower-End.

I cant find it but i remember Krzanich said they had record sales on High-End SKUs, and that is the same thing mentioned by RR from AMDs Q1 report about their APUs.

In the desktop channel, we continue to build momentum in the upper part of our portfolio as we enrich our product mix. Our high-end A8 and A10 APUs had a second straight quarter of record unit shipments.

 

[ShintaiDK]

A8 and A10 is not highend.

Its mainstream at best.