[Techpowerup] AMD "Zen" CPU Prototypes Tested, "Meet all Expectations"

[Flapdrol1337]

I expect per core performance like sandy bridge, but more cores on the mainstream socket.

 

[Shehriazad]

No skyrim doesn't use only 2 cores,it uses 4 main threads like any other console port and it's locked at 60FPS anyway so even a celeron won't make a difference to an i7.
If you want, search for dolphin benchmark or check out mame performance,those two do use 2 cores, are compiled with 3rd party compilers (not ICC) and will show you just how far behind excavator is.
No,the difference is not 100% but it is huge.

Well my bad if Skyrim really does handle more than 2, it was just an example.


Anyway, everyone just always saying "search" or "it's huge" instead of giving me something to actually look at.

But I still stand by it. Compared to an FX83XX, Zen should have like what...60% to 65% more IPC? We also have no clue how the multithreading improved and how much better (or worse) the L3 cache is going to be.

If the clock remains high enough, we are going to look at a competitive product either way. Last thing I heard (no idea how legit all this is...Zen has a disgusting amount of rumors surrounding it) was that this chip would end up being some 8 core/16 threads monster.

Worst case would probably be that it only get close to Intels' Haswell in terms of raw IPC. But that still doesn't tell us anything about how the cache is going to perform and how strong DDR4 plays into this as well (AMD also historically being a friend of higher clocked ram and Intel preferring a more conservative clocked ram<- at least officially). And the clocks...never understimate the clocks.

Long story short...yes AMD could fail, but it could also be a smashing success. Both can happen...but damn some people are biased on this site.

I'll end up buying whatever is best for my money at the time, and I'd not be against AMD having something good again for once.

 

[LTC8K6]

Dont think so,

That 40% angle is neat, and why are there two slightly different slides?

 

[Dribble]

Even those fully committed to the hype train deep down inside know it's going to be another Bulldozer and it's not going to save AMD. We all want AMD to do well but the reality is they don't have a hope of competing with Intel.

 

[Tecnoworld]

If amd were competitive today as it used to be 15 years ago, we'd probably have cpus twice as fast on the market, than we have, at half the price.

 

[ShintaiDK]

If amd were competitive today as it used to be 15 years ago, we'd probably have cpus twice as fast on the market, than we have, at half the price.

Right....because that's the only reason.

 

[ShintaiDK]

GloFo doesn't have their own 14nm process, their own efforts completely failed. Instead, they licensed the Samsung 14nm. Zen will be manufactured on this process.

And they couldn't even deliver a single A9 still due to messing that up as well.

 

[Arachnotronic]

And they couldn't even deliver a single A9 still due to messing that up as well.

But...but...according to GloFo, 14nm yields are super duper!! https://www.semiwiki.com/forum/content/5119-globalfoundries-14nm-process-update.html

What a joke.

 

[PG]

I believe that the 40% claim is possible, and I'm not quite sure why others can't see this.
It was generally accepted that the CMT penalty was around 20%.
L2 and L3 cache issues also caused problems. There must be other things that can be improved on as well.
Add all of this up and they should have a much better performing cpu.
Will it be on par with Haswell or Skylake? Well, that might be too optimistic, but the final result should be much more competitive than where they are now.

 

[MrTeal]

Well my bad if Skyrim really does handle more than 2, it was just an example.


Anyway, everyone just always saying "search" or "it's huge" instead of giving me something to actually look at.

But I still stand by it. Compared to an FX83XX, Zen should have like what...60% to 65% more IPC? We also have no clue how the multithreading improved and how much better (or worse) the L3 cache is going to be.

If the clock remains high enough, we are going to look at a competitive product either way. Last thing I heard (no idea how legit all this is...Zen has a disgusting amount of rumors surrounding it) was that this chip would end up being some 8 core/16 threads monster.

Worst case would probably be that it only get close to Intels' Haswell in terms of raw IPC. But that still doesn't tell us anything about how the cache is going to perform and how strong DDR4 plays into this as well (AMD also historically being a friend of higher clocked ram and Intel preferring a more conservative clocked ram<- at least officially). And the clocks...never understimate the clocks.

Long story short...yes AMD could fail, but it could also be a smashing success. Both can happen...but damn some people are biased on this site.

I'll end up buying whatever is best for my money at the time, and I'd not be against AMD having something good again for once.

http://www.anandtech.com/bench/product/1197?vs=1200
(Keep in mind the 7850k will turbo to 4GHz while the 4360 doesn't turbo, so exact clock speeds are unknown)

Depends on the benchmark, but most of the single threaded CPU benches have Haswell significantly more than 40% ahead of Steamroller.

 

[TheELF]

Anyway, everyone just always saying "search" or "it's huge" instead of giving me something to actually look at.

Dolphin benchmark running on haswell g1820@2.7Ghz
(and not even in a clean environment)

Compared to A10-7850 ( @4Ghz? where does it auto turbo up to with only 2 cores under load? )
http://www.anandtech.com/show/7933/the-desktop-kabini-review-part-1-athlon-5350-am1/4
That's a ~33% clock speed difference and the celeron is still ~11% faster, clock to clock that is a difference of ~44% !!!
That's almost double.
So yeah IF zen has 40% more IPC and IF they manage to run them at at least 4Ghz and IF software will take advantage of the wider core* and IF ad nauseum well then...maybe.

*Just look at how crappy 7zip or handbrake handles big cores

 

[cytg111]

It's on wikipedia, therfor it must be true.

https://en.wikipedia.org/wiki/Zen_(microarchitecture)

It's going to suck megawatts and kick all out ass, intel dummies will scramble to regain som level of formation in dispair and horror - an utterly hopeless maneuveur before getting squashed under the mighty force that is the Zen.

.
.
.
.
.
.
.
.
.
.
.
[giving you a chance to type alot of stuff, then erase it. your welcome.]

On another note, suppose everything is fine and good and even glofo get its act together, I am reading target TDP's for zen as low as 5 watts for dual cores and upper 95 for performance.. thats like an exact copy of what Intel is doing, stretching one uarch over the entire field. That cant be a good sign in cotext of "competing with Intel", I dont see the angle? How can they hope to copy, step by step, and still end up better?

 

[looncraz]

If they try to sell these at Intel prices they better perform like intel chips, other that that sub $150 for a Octocore would be nice. Im happy with my 8320E overclocked. But if they overprice these like they did in the beginning with Bulldozer well I will personally wait to see what Intel does with their prices, hopefully this is what we need to see some price wars, and faster tech.

AMD can't afford to be the cheaper option, exactly. If they start a price war, Intel will demolish them and AMD will collapse.

Already they are pricing themselves effectively equally.

AMD's trick here, though, is that they will be able to carve out a market that doesn't exist due to a gaping hole in Intel's lineup. i7s have a hefty premium over i5s. 2011 chips have a hefty premium over mainstream chips.

If AMD prices the chips more or less equally with the competing Intel chips, but still uses the AM4 platform throughout, it will be cheaper, and easier, to adopt an AM4 platform, with a Zen-based 6/8-core CPU, with or without SMT enabled, than Intel's often neglected 2011 platforms.

I am hoping for the following (highest clocked parts):

FX-8490 BE - w/ SMT - $650
FX-8490    - no SMT - $580
FX-6490 BE - w/ SMT - $450
FX-6490    - no SMT - $375
FX-4490 BE - w/ SMT - $300
FX-4490    - no SMT - $240

All CPUs with Haswell IPC, all unlocked multipliers, all full featureset - except for SMT. I'm in for the big dog, which should compete with the i7 5960X - a $1000 eight core Haswell-E CPU. Of course, Zen will have fewer PCI-e channels, and only dual channel memory support, so it should be expected to come in at a lower price point, as I have here.

All of this assumes Zen can clock equally to Intel. There's a chance it will be much worse, and a slight chance it will be better.

 

[looncraz]

On a low power optimized 14nm Samsung process? I disagree.

In theory, 14nm LPP has higher clocking capabilities than 28nm SOI.

Lower leakage at lower power, true, but the transistors switch much faster, which allows more complex stages to complete that much sooner, which enables higher clock speed designs.

https://www.semiwiki.com/forum/content/5119-globalfoundries-14nm-process-update.html#comments

So long as Zen can keep the power where it is needed, keep the signals integrity high, and dissipate the heat from the hot spots effectively, all of which is handled by modeling tools, then it should be able to clock quite well.

It will have one heck of a power curve, though. Might be 95W at 3.5Ghz, and 150W at 3.8...

I hope AMD can release stock clocks in the ~4GHz range, with some decent turbo room and overclocking room. Probably not on the eight-core models, though :'(

 

[BurnItDwn]

I hope they are better than top end i7, in real life, I expect somewhere around i3. I havent dumped my AMD shares yet, but every day I am tempted.

 

[looncraz]

I'm sick of forum drama. I'd rather talk about the facts when we have them, not fight endlessly over nothing.

Where's the fun in that

 

[dark zero]

It's on wikipedia, therfor it must be true.

https://en.wikipedia.org/wiki/Zen_(microarchitecture)

It's going to suck megawatts and kick all out ass, intel dummies will scramble to regain som level of formation in dispair and horror - an utterly hopeless maneuveur before getting squashed under the mighty force that is the Zen.

.
.
.
.
.
.
.
.
.
.
.
[giving you a chance to type alot of stuff, then erase it. your welcome.]

On another note, suppose everything is fine and good and even glofo get its act together, I am reading target TDP's for zen as low as 5 watts for dual cores and upper 95 for performance.. thats like an exact copy of what Intel is doing, stretching one uarch over the entire field. That cant be a good sign in cotext of "competing with Intel", I dont see the angle? How can they hope to copy, step by step, and still end up better?

INBF is Intel on the shadows giving AMD part of their tech to not to see AMD die for now to avoid European and Chinese courts.

 

[thesmokingman]

This is so pathetic, tech news sites are worth squat. They're either repeating fabrications or selling bias. Real journalism seems to be dead in pc enthusiast news.

 

[looncraz]

The only thing we know as a fact is AMD is betting the farm on Zen.
Therefore, it will be their fastest part they ever made.
Anything else is speculation.

Even if they do have a killer chip, I have high doubts on if GloFlo can deliver the quantity & cost needed to be competitive--even if they can handle sammy's tech (which is a big if).

I think Global will be able to do just fine. If they can't, AMD always has Samsung as a second foundry if demand is high enough :thumbsup:

Nothing quite like being able to take the final taped out version of a chip and send it to another foundry and say, make me some, all the work is already done, we just need more capacity

As it stands, Global Foundries has 14nm LPP up and running with exceptional yields:

https://www.semiwiki.com/forum/content/5119-globalfoundries-14nm-process-update.html

There are some hints here that 14nm LPP may be better clocking than 28nm SOI as well - making it more suited for high performance chips than most any layman thought - even though it is even better for low-power chips (i.e. low leakage). Of course, I argued that point a long time ago :ninja:

 

[looncraz]

40% would not be anywhere near enough though. They would have to more than double the IPC (100% more) to remain competitive with Intel in the mid-range, because currently their cores have about half the IPC of Intel.

40% is almost exactly head-on with Haswell.

Seems weird, I know, but you have to take into consideration that this is 40% on top of Excavator. FX is two full generations behind Excavator.

Zen, then, should have an 80% higher IPC than Bulldozer, which puts it neck and neck with Haswell, provided the IPC improvements are distributed as one would expect (i.e. as it was with phenom II).

Everything then comes to clock rates. Identical IPC means almost nothing if you have 25% lower clockrates.

 

[Ferzerp]

I am pretty sure that AMDs new architecture will meet the expectation that any knowledgeable, unbiased, rational, reasonable person would have.

 

[ThatBuzzkiller]

Whatever Zen will be, it'll be a fun day on it's launch ...

I'm leaning more on the side of wanting Zen to beat Skylake just to really spite the Intel camp ... (Even though I don't expect much from AMD.)

 

[TheELF]

I am pretty sure that AMDs new architecture will meet the expectation that any knowledgeable, unbiased, rational, reasonable person would have.

Well duh,such a person would have no expectations since such a person would not have enough information, so of course it would meet them.

 

[myocardia]

If amd were competitive today as it used to be 15 years ago, we'd probably have cpus twice as fast on the market, than we have, at half the price.

Would that be the same AMD that sold the first ever $1,000+ personal-usage CPUs, and had median retail price of their CPUs that was well over $300, until Intel Conroed them, and forced them to lower their prices?

edit: But yes, I do agree that without competition, we would still likely be using 486s. I do honestly think that without AMD's Athlon 64, Intel would likely never have developed the Conroe core, which was as big of a jump over the P4 as the first Pentium was over the 486.

 

[prtskg]

i7 5960X runs @3ghz and comes with tdp of 140W. Why are people expecting Zen's 8 core to be better than this when it is said to be 95W? Also Intel's foundry is definitely more experienced than that at GF. So no help to AMD there either.
I hope Zen, Artic Islands gpus and k12 are good enough to help AMD survive and keep competition in X86 and gpu market.