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

[Vesku]

I don't think a 95W Zen chip is going to have 20MB of L3, so there might be room for at least 6 cores 12 threads if not 8/16. Time will tell.

 

[AtenRa]

I don't think a 95W Zen chip is going to have 20MB of L3, so there might be room for at least 6 cores 12 threads if not 8/16. Time will tell.

I dont believe AMD will make a different die for the desktop segment. They most probable will use the same die for server, workstation and desktop.

What people dont take seriously in to consideration is the 14nm FinFet process and how good it can perform at low power.

 

[boozzer]

I 100% wish for a zen that is comparable to current gen intel cpus. PC hardware really need a damn performance jump it has been stale for the last 5+ years. the last major jump was 2500k, that is Q1 2011.

 

[TechGod123]

I 100% wish for a zen that is comparable to current gen intel cpus. PC hardware really need a damn performance jump it has been stale for the last 5+ years. the last major jump was 2500k, that is Q1 2011.

AMD will come back swinging and the ferocious CPU wars will begin once again...

 

[guskline]

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.

For a Junior member with only 1 post this poster hit the nail directly on the head.

I've owned so many AMD and Intel chips it's hard to remember how many. I started with an Intel 386DX-33 and replaced it with a faster AMD 386-40 (the days when they were neck and neck), I've lived through the infamous Intel P60 bug and soldiered on through the hey day of the faster AMD chips. They were, for awhile less expensive and a bit faster than Intel. However, I noticed a big jump in performance from Intel with the I5-2500k. At the time I stayed away from the Intel 900 series because it was too expensive for my taste but it was fast.

AMD released the 965 and I loved that chip but I noticed Intel was pulling ahead. Bulldozer and later Piledriver were way over hyped but priced OK.

When the I5-2500k arrived on my desk I made the switch to Intel. At the time I had a Phenom 1100T(if I remember correctly) and it was a solid chip with more cores than Intel but that 4 core/8 thread 2500k was magic and OCed like nothing I've ever seen. At the same time I was running a heavily OC'd AMD 8350 which was fast but very hot and power hungry. In gaming the 2500k was the clear Monster and whipped the 8350 (at4.6Ghz) in almost every synthetic etc.

I then moved into the I7-3930k realm to truly experience multi-core beasts and finally arrived at the 5960X BEAST.

What I noticed along the way was AMD falling farther and farther away from Intel on the high end and even mid level.

If Zen can put them back in the fight like I think the Fiji has in the gpu realm, they have a chance to be viable. However, the poster above is very savy and realistic about the likely results. Zen is not likely to be an Intel "killer".

Finally, though the 5960X is stock clocked at 3 Ghz with 140W TDP, those of us who own it know that it is very easy to OC to at least 4Ghz and with some tweaks upwards to 4.5Ghz. Hard to describe what a monster it it is. Yes the price is very salty, but it's a beast.

Like all of you I'm following ZEN closely but have realistic expectations of it.

 

[Jimzz]

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.

AMD and Intel measure TDP differently. The i7 5960X uses around 93-106watts under load. Also remember the intel chip has the VRM on the chip vs off for AMD so that has to be accounted for in the TDP as well.

 

[CHADBOGA]

Exactly.

Of course -- Zen could retail for $75, offer quadruple the performance of Haswell, benchmark Cinebench with 1 billion points and operate with a TDP of 5 watts and Shintai will still call it a flop.

Gotta love it.

Are you Shintai's psychological prisoner? D:

Whether Zen is a flop or not will come down to whether it allows AMD to increase marketshare and/or profitability.

We should have a good idea after a few reviews by respectable sites how it is likely to fare though.

 

[Magic Carpet]

What people dont take seriously in to consideration is the 14nm FinFet process and how good it can perform at low power.

Ah really, Skylake surely didn't look to me much more power-efficient/faster than Haswell. But if AMD will have the superior tech, why not indeed. I could use Zen to replace my Thuban rig with.

 

[MiddleOfTheRoad]

Are you Shintai's psychological prisoner? D:

Whether Zen is a flop or not will come down to whether it allows AMD to increase marketshare and/or profitability.

We should have a good idea after a few reviews by respectable sites how it is likely to fare though.

Actually, even market share isn't as important as maximizing profitability. General Motors chased marketshare at the expense of profitability for years -- which is why they collapsed into backruptcy. Better for AMD to maximize the margins at what the market will bear for Zen to try to heal their balance sheet.

 

[mysticjbyrd]

Actually, even market share isn't as important as maximizing profitability. General Motors chased marketshare at the expense of profitability for years -- which is why they collapsed into backruptcy. Better for AMD to maximize the margins at what the market will bear for Zen to try to heal their balance sheet.

I thought it was because the people at the top of the pyramid used it like their personal piggy bank, and took billions from the company in pay, benefits, perks, homes, cars, jets, vacations, retreats etc.. But it was the marketshares, hmm.

 

[AtenRa]

Ah really, Skylake surely didn't look to me much more power-efficient/faster than Haswell. But if AMD will have the superior tech, why not indeed. I could use Zen to replace my Thuban rig with.

Thats because the huge increase in efficiency is not at 4GHz but at 2.0GHz to 3.0GHz range (lower voltage).

Have a look and compare the mobile Core i5 4200U (22nm Haswell) vs Core i5 6200U (14nm Skylake).

and

 

[erunion]

I thought it was because the people at the top of the pyramid used it like ou personal piggy bank, and took billions from the company in pay, benefits, perks, homes, cars, jets, vacations, retreats etc.. But it was the marketshares, hmm.

GMs primary problem was market share loss. But only because It had massive fixed legacy costs, but faced declining market share. It was forced to close dealers and discontinue brands as it simply no longer had the volume to support them.( This was true of ford and Chrysler as well)

A main fixed cost for GM was its pension.(When GM was larger it had more employees, those employees retired and were being supported by today's smaller work force) As their volume declined their pension liability equaled thousands of dollars per car sold. Which led to rampant cost cutting which led to bad cars, and the cycle continued.

Most large companies, some successful and some not, spend extravagantly on their execs. So it definitely can not be isolated as a cause of failure.

 

[jpiniero]

Better for AMD to maximize the margins at what the market will bear for Zen to try to heal their balance sheet.

They need both high volume and high margins really because of the costs involved. But their attempts to keep prices high (the FX 9000 series initially and now the 300/Fury GPUs) have backfired on sales. If AMD were to sell the Zen 8 core for $500 they won't sell that many.

 

[frozentundra123456]

Well, that depends on the performance. If 8 core zen = 8 core intel HEDT then it would be a steal at 500.00. But I seriously doubt that it will. Not sure the latest Intel roadmap, but shouldn't Skylake E be coming out around the time Zen appears?

If it can slot in between 6 and 8 core intel performance, then 500.00 is realistic. However, I expect performance at best to equal six core intel, which you can get under 400.00 now, so, yes, 500.00 is too much.

In any case, they either have to do well in servers or get Zen apus out quickly to get any kind of volume. Despite the enthusiasm for high end desktop chips in these forums, the consumer/enterprise market will be very limited without an igp.

 

[jihe]

LOL meeting all expectations. We all know what our expectations of AMD are...

 

[TechGod123]

LOL meeting all expectations. We all know what our expectations of AMD are...

That Zen will do fine and get Haswell or maybe slightly more IPC level performance? I thought that too!

 

[AtenRa]

Not sure the latest Intel roadmap, but shouldn't Skylake E be coming out around the time Zen appears?

Nope, Broadwell-E is coming Q2-Q3 2016 and ZEN probably Q4 2016 or Q1 2017.

 

[sm625]

Whatever AMD does, I hope they do not offer any Zen SKU that has less than 4 real cores. If they want to get back into the game they need to set a floor, with zero SKUs that could tarnish their name, like "E1-1200". Ugh. Ideally the absolute minimum die will be 8 cores and harvested dies will go down to 6 core and no lower. For APUs the minimum can be 4 cores but that puts them in a pickle when it comes to die harvesting. They could in theory use the harvested parts for mobile chips, given how few they sell lol.

 

[AtenRa]

Whatever AMD does, I hope they do not offer any Zen SKU that has less than 4 real cores. If they want to get back into the game they need to set a floor, with zero SKUs that could tarnish their name, like "E1-1200". Ugh. Ideally the absolute minimum die will be 8 cores and harvested dies will go down to 6 core and no lower. For APUs the minimum can be 4 cores but that puts them in a pickle when it comes to die harvesting. They could in theory use the harvested parts for mobile chips, given how few they sell lol.

Im thinking they will follow Intel with 6-8 cores for the Workstation and High-End Deskstop SKUs and 2-4 Cores for the APUs (Mobile and Desktop).

 

[Sweepr]

Well, that depends on the performance. If 8 core zen = 8 core intel HEDT then it would be a steal at 500.00. But I seriously doubt that it will. Not sure the latest Intel roadmap, but shouldn't Skylake E be coming out around the time Zen appears?

Yes. Skylake-EP is H1-2017. Expect more cores for the desktops.

 

[Magic Carpet]

Thats because the huge increase in efficiency is not at 4GHz but at 2.0GHz to 3.0GHz range (lower voltage).

What kind of efficiency are you talking about in a desktop chip? If that's not raw performance, I am not that much interested, to be honest.

 

[AtenRa]

Yes. Skylake-EP is H1-2017. Expect more cores for the desktops.

Broadwell-E was pushed to Q2-Q3 2016. Making a Skylake-E H1 2017 release highly questionable.

 

[AtenRa]

What kind of efficiency are you talking about in a desktop chip? If that's not raw performance, I am not that much interested, to be honest.

Im just saying that the efficiency curve will not be in high clocks but at lower frequencies due to the FF process.
Im waiting to see the 8-core Broadwell-E and compare it to 22nm 8-core Haswell to draw any conclusions.

 

[Magic Carpet]

Im just saying that the efficiency curve will not be in high clocks but at lower frequencies due to the FF process.
Im waiting to see the 8-core Broadwell-E and compare it to 22nm 8-core Haswell to draw any conclusions.

And has efficiency ever been in the highest clocks, regardless of the manufacturing process used?

If you go back, you will find a similar curve.

 

[Sweepr]

Broadwell-E was pushed to Q2-Q3 2016.

No it wasn't. Q2 maybe, Q3 comes from your imagination.

Making a Skylake-E H1 2017 release highly questionable.

I bet it will, and it will put AMD in its place again once it launches.