You must choose right now: Zen or Skylake-E (8 core)

[moonbogg]

What would it be? I'd take the chance on Zen. I'd jump on it right now. I already know it will be fast and more than good enough. Faster than my Sandy-E I expect. AMD made the effort to fix their CPU's and I'd reward them for it by purchasing their platform. If it gets stuck at 3.5 I'd regret it, but I doubt it will. Its either that or a Skylake-E 8 core, but who knows how high that will clock? Gamble either way.
Most of you would just wait and see. Follow your heart instead. Go with your GUT!

 

[imported_bman]

If I had to pay for it and assuming AMD is going to undercut the price significantly: Zen
Otherwise: Skylake-E

 

[moonbogg]

Yep, good answer. Its time to switch camps. Zen will be a great CPU and will almost certainly cost less than the alternatives. Won't be cheap at all, but should cost less. Its going to be a compelling and very exciting chip. I can SMELL IT.

 

[HurleyBird]

Skylake-E is the safer choice, but I'll wait until I see both in action.

 

[escrow4]

Skylake-E without hesitation. You remember how some of those FM2(+) boards were overvolting the CPU? Something to do with the voltages anyway. Intel is a known quantity, AMD hasn't had anything good CPU wise since 2006 and nothing platform wise recently. I buy what I know will work. And thinking that Zen will actually match Skylake - remember Bulldozer?

 

[whm1974]

I don't think that the 8 core Zen and Skylake-E are in the same market segment.

 

[NTMBK]

If you're giving me them for free, Skylake-E. It's a quad (or possibly hex!) channel platform, with AVX-512 on 10 cores. No competition.

If I have to buy them? Zen, probably.

 

[NTMBK]

I don't think that the 8 core Zen and Skylake-E are in the same market segment.

Yup. The competitor for Skylake-E will be 16 core Zen on a server socket (i.e. two Zen dies in one package).

 

[nathanddrews]

I voted for Zen, assuming that:

1. I have to pay for it and I assume they will undercut Intel
2. IGP will be really good and I won't be able to afford a dGPU afterward

 

[guskline]

Skylake E for this reason. AMD will introduce the ZEN which will be well received and probably be in the ballpark of a 5960x for the 8c/16t Z.

However, Intel ("Evil Empire" to AMD afficianados) wants to control the high end so I expect Skylake E to come out blazing.

 

[Abwx]

Skylake-E without hesitation. You remember how some of those FM2(+) boards were overvolting the CPU? Something to do with the voltages anyway. Intel is a known quantity, AMD hasn't had anything good CPU wise since 2006 and nothing platform wise recently. I buy what I know will work. And thinking that Zen will actually match Skylake - remember Bulldozer?

Some of those FM2+ and about all X99 isnt it, this latter plateform is not reliable enouh in respect of mainstream plateforms, so Zen is the CPU of choice given that it will have a large panel of MBs at all prices, and in the waiting of lower CPU price one can plug a 4C or even an APU while with the Intel E series you are stuck with very expensive items only...

 

[mnewsham]

with AVX-512 on 10 cores. No competition.

Is AVX-512 good for anything?

https://indico.cern.ch/event/327306...ttachments/635800/875267/HaswellConundrum.pdf

 

[Arachnotronic]

I have already put aside the money for my Skylake-X build which I plan to do in 2H 2017.

It is unlikely that an 8C/16T Zen will offer more performance than my 10C/20T Broadwell-E @ 4.3GHz. Even if Zen is per-core/per-clock roughly equivalent to Broadwell (and I don't think a single Blender render under undisclosed software/hardware configurations is enough to draw a conclusion either way), I would be very surprised if it is able to hit the kinds of clocks that Broadwell-E can, let alone Skylake-X which should benefit from being built on 14nm+ (on top of the architectural/circuit design enhancements that could help frequency)

Don't take this as thread-crapping, but I think the "over exuberance" exhibited by the OP is premature. Ask yourself this: why did AMD need to down-clock the Broadwell-E part in order to do the clock-for-clock comparison? Why couldn't the Zen chip be clocked at similar levels to the Broadwell-E? Why did they use a multi-threaded benchmark if they were interested in comparing perf/clock?

 

[SlitheryDee]

Right now I go skylake-e. I'm an amd fan at heart, but they're at the point where I need proof in the form of benchmarks before I buy a processor from them. The odds favor amd because I'd go with them even in the event of a virtual tie, but I'm not going to do it without having some numbers at hand.

 

[NTMBK]

Is AVX-512 good for anything?

https://indico.cern.ch/event/327306...ttachments/635800/875267/HaswellConundrum.pdf

It's good for playing around with I've been waiting for AVX-512 for ages... it'll be fun to play around with an instruction set much more suited for vectorization. Double the vector registers, true masking for every operation, extra mask registers so you no longer waste a whole vector register on storing a 0xFFFF "mask"... Fantastic.

 

[TeknoBug]

Still too early to determine.

If Zen turns out well I may build a Zen PC for my racing seat/rig.

 

[moonbogg]

I have already put aside the money for my Skylake-X build which I plan to do in 2H 2017.

It is unlikely that an 8C/16T Zen will offer more performance than my 10C/20T Broadwell-E @ 4.3GHz. Even if Zen is per-core/per-clock roughly equivalent to Broadwell (and I don't think a single Blender render under undisclosed software/hardware configurations is enough to draw a conclusion either way), I would be very surprised if it is able to hit the kinds of clocks that Broadwell-E can, let alone Skylake-X which should benefit from being built on 14nm+ (on top of the architectural/circuit design enhancements that could help frequency)

Don't take this as thread-crapping, but I think the "over exuberance" exhibited by the OP is premature. Ask yourself this: why did AMD need to down-clock the Broadwell-E part in order to do the clock-for-clock comparison? Why couldn't the Zen chip be clocked at similar levels to the Broadwell-E? Why did they use a multi-threaded benchmark if they were interested in comparing perf/clock?

Premature? Of course its premature. I've always been told I was a premature kind of guy. I'm just trying to be consistent.

This. Is. Zen.

And you should be excited about it. You should also understand that this thread is not only for discussion, but for fun. I find it fun to take a hypothetical risk and go for Zen right now, even though no one could possibly make that choice yet. We will all see the numbers before buying, like it or not. Also, you should understand that I am moonbogg. No no, I mean,

I. Am. Moonbogg.

 

[Face2Face]

It's all about dem clocks. If ZEN can do 4.2-4.5GHz without tripping a 15amp breaker, then I think we will have a real winner on our hands...assuming CPC it's as fast as BW-E.

 

[moonbogg]

It's all about dem clocks. If ZEN can do 4.2-4.5GHz without tripping a 15amp breaker, then I think we will have a real winner on our hands...assuming CPC it's as fast as BW-E.

Lol nice. Maybe they can sell a special circuit protector thingy along with the CPU. Big box item.

 

[frozentundra123456]

Sorry, but we don't know clocks, overclocking potential, price, or performance in a wide range of benchmarks for Zen. Doesnt make *any* sense to try to make a decision now. Guess I am no fun (my wife has said that a few times too.) I do think Zen though will not bring the huge price/performance increase that some are expecting. If it overall is only competing with hex core intel, there is already a 400 dollar cpu there. If it manages to compete overall with the thousand dollar 8 core, then they could price it at 750ish and offer very good value for those that need the cores. AMD is lucky that BW-E was a very small, if any, increase except for the 10 core, so that gives them an easier target at launch, but I expect (am hopeful anyway) that Skylake E will bring a decent IPC increase and better overclocking.

Edit: one plus for Zen is that the platform may be significantly cheaper, and in the long term, will allow upgrading to more cores from a quad core. But that is really only relative to the eventual quad core Zen or Zen APUs.

 

[zinfamous]

It's all about dem clocks. If ZEN can do 4.2-4.5GHz without tripping a 15amp breaker, then I think we will have a real winner on our hands...assuming CPC it's as fast as BW-E.

Should I go ahead and put in an order for a 220 outlet with a 30a breaker?

I was thinking about installing a -80C freezer in the house, anyway.

:hmm:

 

[moonbogg]

Sorry, but we don't know clocks, overclocking potential, price, or performance in a wide range of benchmarks for Zen. Doesnt make *any* sense to try to make a decision now. Guess I am no fun (my wife has said that a few times too.) I do think Zen though will not bring the huge price/performance increase that some are expecting. If it overall is only competing with hex core intel, there is already a 400 dollar cpu there. If it manages to compete overall with the thousand dollar 8 core, then they could price it at 750ish and offer very good value for those that need the cores. AMD is lucky that BW-E was a very small, if any, increase except for the 10 core, so that gives them an easier target at launch, but I expect (am hopeful anyway) that Skylake E will bring a decent IPC increase and better overclocking.

Edit: one plus for Zen is that the platform may be significantly cheaper, and in the long term, will allow upgrading to more cores from a quad core. But that is really only relative to the eventual quad core Zen or Zen APUs.

Allow me to shock you with my predictive abilities. The following has already happened:

8 core Zen has an average OC of 4.2Ghz using moderate voltage increases.
8 core Zen costs $650.00
8 core Zen is within 10% performance clock for clock of Broadwell.

After Zen is released, you will come back to this very post and your jaw will hit the floor so hard it will register at the USGS.

 

[Arachnotronic]

Allow me to shock you with my predictive abilities. The following has already happened:

8 core Zen has an average OC of 4.2Ghz using moderate voltage increases.
8 core Zen costs $650.00
8 core Zen is within 10% performance clock for clock of Broadwell.

After Zen is released, you will come back to this very post and your jaw will hit the floor so hard it will register at the USGS.

It would be a shocker to me if Zen 8C/16T could hit 4GHz on high end air or even an AIO CLC.

 

[The Stilt]

It would be a shocker to me if Zen 8C/16T could hit 4GHz on high end air or even an AIO CLC.

You're even more optimistic than I am. I'd say reaching 4GHz will require sub-zero temperatures. Polaris 10 maxed out at 1700MHz on LN2 (while using deadly voltage levels). If the same percentage of OC is applied on 3.2GHz Zeppelin, that's about 4.3GHz even on LN2.

 

[Justinbaileyman]

I will be buying Zen after all,as long as AMD doesn't try to over charge for there CPU's to try to compete 1/1 with Intels pricing scheme.I just can't see myself spending $1k or more for 8c/16t CPU. I am hoping for a $600.00 price point or less and then I will be all set. The only way I could even remotely see myself ever buying Skylake-E would be if Intel dropped its pricing and the CPU would have to be close to 2 times the performance of the Zen Cores.