Intel Skylake Platform

8 sandy-comparable cores at a mainstream price? Color me sceptical, but maybe AMD doesn't want to make any profit off of the chip; it COULD happen... ;)

Frankly I'd expect AMD to torpedo its profit margins before I expect zen's performance to rival Intel's... *shrug* Maybe I'm being too pessimistic, who can say. At this point though I must admit I'm wondering if zen is actually going to get released. AMD has a disturbing tendency to just cancel products these days because of lack of resources/funds. Maybe this project will be prioritized. It seems to be somewhat of a hope-filled symbol which company management clings to, heh.

They've probably canceled a few things just to make sure they get Zen out of the door. I mean, at this point, either they do or it's all over.
 
8 sandy-comparable cores at a mainstream price? Color me sceptical, but maybe AMD doesn't want to make any profit off of the chip; it COULD happen...
Shouldn't be all that hard if they are foregoing integrated graphics, look at a core i3 die shot for example.
 
will be futile tho. We already see intel doesn't mind selling a 6 core haswell at $300 . I don't think it will bother them to drop a 8 core into a lower price point also and then AMD will be back to where they are now , the low end of the market.

I do hope I'm wrong though.
 
Yeah, AMD's big opportunity here is if Intel is slow to respond with making 8 real cores reasonably priced. Given Intel's love of money, I actually expect this to be the case... at least for a short time. But even if Intel adapts, Zen at least represents the potential to reasonably competitive again (with perf/watt) and an even greater long term hope as process migrations begin to slow down. Being a year behind on a 6-10 year cycle is not nearly as bad as a year behind on a 2-3 year cycle. And who knows, other fabs may even catch up with Intel....
 
Intel already makes 8-core CPU, and 10-core, and 12-core, and 14-core, and all the way up to 18 cores per socket. The only thing "stopping" Intel from dropping any number of those into a consumer market is demand.

The overwhelming, super-majority of consumer software applications do not use that kind of core count. Even if you wanted to make that argument, you'd still have to fight with a 6c/12th Haswell for $300 that I wager would still eat an 8-core Zen for breakfast and have room for tea and another biscuit.

I have yet to understand why everyone thinks "real" core count is going to somehow save AMD. They need IPC. They need to fix their memory access (cache) latency issues. They need to get their power budget under control. If they can fix those things, they're competitive at any core count. Leave them all as-is, and 8 "real" cores would still lose.
 
I'd like to add that AMD's problem, IMHO, isn't the top end of the market.
AMD's problem is that people are overwhelmingly choosing Celerons over A4s, because single threaded performance is more valuable than a more powerful but still way too weak GPU. Not to mention the Intel brand.
Celerons are 10-50% faster in singlethreaded software while also costing the same or being $5 cheaper. Not to mention power consumption savings (with CPU+MB costing $70, even a few watts matter).
 
The only thing "stopping" Intel from dropping any number of those into a consumer market is demand.
No. The demand is already there (I know because I am part of it). And $1000 dollars for an 8 core is just greed; they do it because they can.

you'd still have to fight with a 6c/12th Haswell for $300 that I wager would still eat an 8-core Zen for breakfast and have room for tea and another biscuit
There is no evidence to support this claim. I prefer to deal in what is.

Leave them all as-is, and 8 "real" cores would still lose
They aren't leaving them as is, so I don't see the point here.
 
Not to mention the Intel brand.

That's something that there is no need to be mentioned. It is just a name without any real weight on the silicon's quality itself. That is just a 5-letter word with no value.

Zen will be 8 core 16 threads with SMT, how convenient how some mentioned the 6-core, 12 - thread but skipped that detail.

We need competition and I would be happy if you help and not go against it by all means.
 
Zen will be 8 core 16 threads with SMT, how convenient how some mentioned the 6-core, 12 - thread but skipped that detail.
It does not matter. Most of those who buy anything above 4 cores know what they are doing. They'll just choose the core that fits their needs, be it from AMD or Intel. If Intel's hexcore has better value than AMD's octacore, they'll go for the hexcore. If AMD has better value, they'll go for AMD.

That's something that there is no need to be mentioned. It is just a name without any real weight on the silicon's quality itself. That is just a 5-letter word with no value.
No value to us.
It has value to the average consumer who knows that Intel just works™

We need competition and I would be happy if you help and not go against it by all means.
Most people don't go against it. Most people choose whatever has the most value per dollar. Value may include performance, reliability, perception, features...
I have often bought AMD. Once because I just needed the absolutely cheapest option, once because I wanted 8 native SATA ports and did not care about performance, once because I needed more than an Intel iGPU but did not require a discrete one, once because I expected AM3+ would not be abandoned... problem is, right now, most peoples' needs (value determinants) favor Intel, not AMD.

No. The demand is already there (I know because I am part of it). And $1000 dollars for an 8 core is just greed; they do it because they can.
The demand has to be high enough to justify all of the nonrecurring expenses of creating a consumer hex/octa core part.
In other words, the demand in the gap between a quad core and a hex core on a consumerized 1S server platform has to be large. IMHO, it cannot possibly be.

Besides, $1000 for an octacore is not greed. It's a way of earning high margins from those consumers who can afford to pay for top performance. If one needs an octacore or RAS, one is a pro, and should pay for it. Be happy, this subsidizes R&D for the dual and quad cores too.

What I do expect is that at some point the server platform might get at least a GT1 GPU on the die.
What I expect from Zen showing up is prices compressing. I do not expect Intel to release a consumer octacore. However, the 4790k equivalent might drop to $200-$250.
 
Server boards do benefit from the barebones, decrepit VGA built-in (Matrox, ASPEED, Volari) in some way : the server OS is likely to support it and the driver must be equally barebones. Contrast with Broadwell, presumably Skylake needing bleeding edge Linux and full of complicated features. If you're running something like RHEL (conversative long time support Red Hat Linux) or FreeBSD the latest Intel GPU will not be supported particularly well.


As you're talking of high end, RAS and GPU : there was some PR about server and high end Skylake and here's what I refer to :
http://www.hardware.fr/news/14219/xeon-skylake-2017-28-c-urs-6-canaux.html

Gone the 2011-3 socket that did 2-socket servers and 2011-1 (Haswell-EX and Broadwell-EX, up to 8 sockets) for the very high end, instead come Socket R which is 1S only! and Socket P for 2S, 4S, 8S.
1S R is the natural socket for Skylake equivalent of i7 5930K, 5960X etc. and probably an acknlowedgement that run of the mill workhorse servers and high end workstations only need one socket, afterall.

Socket P is for mid/high end, up to definitively replacing Itanium I think. This is where you find RAS, not the 1S socket.
Lots of goodies there, such what as I interpret to be a supercomputer interconnect, putting a Xeon Phi in a CPU socket, and there's talk of accelerators : among which FPGA attached to a Skylake CPU, and "Cannonlake graphics". It's anyone's guess what that means exactly but it seems like something big (or just Quicksync stuff + some GPGPU) and funnily for computers much higher end than a one-socket Core i7 with 8 cores.
 
Which is a fancy way of saying greed.
No. It's one of the ways of supporting their huge R&D expenses.
In other words, I'm happy that octacores cost $999 instead of $500, because it probably means that the quadcore I want costs $200 instead of $250.
 
No value to us.
It has value to the average consumer who knows that Intel just works™

Do you realise that this can not be serious and is a LIE and not fair.
AMD just works perfectly fine as well™

Otherwise you need to post links with sources and proves that AMD's products don't.
And if you don't, they can sue you.
 
AMD just works perfectly fine as well™
Yes, it does. And I know that.
Problem is, perception by the ordinary consumer matters. In forums in my country, I read a lot of unsubstantiated "AMD to me is an unknown", "I've always relied on Intel, never have been wrong", "AMD is crap" comments. Not to mention people who haven't been following computing buying Intel because "AMD may be faster, but Intel will not throw any surprises at me", i.e. when the price is the same, they don't bother doing research and just buy Intel.

Otherwise you need to post links with sources and proves that AMD's products don't.
That's easy. One can always go back to this.
http://techreport.com/review/13832/an-update-on-amd-790fx-motherboards/5
http://techreport.com/review/14261/amd-780g-chipset/10
http://techreport.com/review/21019/bulldozer-mobos-from-asus-and-msi-sabertooth-990fx-990fxa-gd80/8
Yes, one can see how it went from "almost horror" to "slightly worse performance", but who is to say this will not happen again?
And I realize it can happen to Intel too.
 
Problem is, perception by the ordinary consumer matters. In forums in my country, I read a lot of unsubstantiated "AMD to me is an unknown", "I've always relied on Intel, never have been wrong", "AMD is crap" comments. Not to mention people who haven't been following computing buying Intel because "AMD may be faster, but Intel will not throw any surprises at me", i.e. when the price is the same, they don't bother doing research and just buy Intel

This is because people have no clue. Their brains are washed with some predetermined "truth" and it is almost never allowed to jump in (the correct) another direction.

I agree and admit that for this equal responsibility has AMD itself too. It is such a pity when a company gives up and agrees not to compete.
 
Do you realise that this can not be serious and is a LIE and not fair.
AMD just works perfectly fine as well™

Otherwise you need to post links with sources and proves that AMD's products don't.
And if you don't, they can sue you.

Well I did find this ;)
3584527m.png


A short errata description for the various BD re-spins.
 
@Davros There's tons of errata listed for Intel processors as well, they have a database somewhere listing all of it. Some can be worked around using firmware microcode fixes, but not everything. The transactional memory handling for example was unfixably broken in both Haswell and Haswell-E series processors and was disabled entirely with a firmware fix for example. There's lots more if one just bothers to check...

Not that any of this is anything strange really, today's CPUs are so complicated that edge-case errors will crop up now and again, despite tons of testing and debugging.
 
The irony is that with the help of a bulldozer, piledriver and excavator AMD constructs very beautiful grave, but the wrong one - it turns out to be their own.

:(
 
Back
Top