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

[frozentundra123456]

For any widespread consumer/enterprise use, an igp is mandatory. If the price/performance is good enough, there will be a niche high end desktop market, but it will be rather small, and they are fighting directly against Intel there as well. Zen really has to do well in servers from the start to get some return before the APUs are ready.

I dont think it was a mistake really, just that AMD felt they had to bring a new cpu to the market, and the APUs were not ready.

 

[frozentundra123456]

I'm not so sure. OEMs at this point have pretty much ignored Intel's efforts on Iris so far. A large reason for this is because they very much favor nVidia GPUs on machines marketed for gaming. It's gonna be tough for even Intel to fight that.

Yes, but the performance is not really quite there, and availability has been restricted to very high end chips. An APU for mobile has to deliver acceptable performance at the right price. The way Intel has handled iris pro if just extremely frustrating. All quad core mobile processors should just have GT3e, and priced in the 800.00 region. That would be a compelling product. As it is, there are hardly any GT3e laptops even available, except Macs, and if they are available, it is much cheaper to get a quad plus 960M, which gives better performance. I dont blame the OEMs for this, I blame intel for not making GT3e more available in mobile. Can anybody tell me why it made sense to bring out BW-K for the desktop instead of mobile?? Supposedly this will change with Skylake, but I have a feeling that the SKUs with e-dram are going to end up not really being utilized.

 

[jpiniero]

I dont blame the OEMs for this, I blame intel for not making GT3e more available in mobile. Can anybody tell me why it made sense to bring out BW-K for the desktop instead of mobile??

What makes you think they didn't release any on mobile? There's 3 i7 models - 5950HQ, 5850HQ, 5750HQ. There's also an i5 with two cores disabled. OEMs didn't bother.

 

[frozentundra123456]

Well, the 5950HQ is over 600.00, 200 dollars more than the other two. So that pretty much rules that out. The other 2 look reasonable, I dont really know why they arent available in retail. You may be right, OEMs may have ignored them, but we dont know what went on behind the scenes either. We all know broadwell was fraught with supply problems, and who knows, maybe intel didnt push them because non-edram chips were cheaper to produce. I believe that Intel has enough influence with OEMs to have made sure they were available if they were committed to it and the supply was there. But actually this kind of proves my point. Those look like really sweet chips, what should be the standard mobile quad core.


But this is really off topic in this thread.

 

[cbn]

For a laptop, if they could reach something like GT 950/960M performance in a 35 or 45 watt APU and sell it cheaper than intel + discrete, it would be attractive.

960M is 640 Maxwell v1 CUDA cores with 128 bit GDDR5 (so basically GTX 750 Ti).

That is a good amount of GPU.

With that mentioned I wonder how much space having CPU and GPU integrated like that would save under the hood of a laptop.

See below this example of late 2013 15" Apple Mac Book Pro Retina with 47W Haswell Iris Pro:

https://www.ifixit.com/Teardown/MacBook+Pro+15-Inch+Retina+Display+Late+2013+Teardown/18696

Notice it still takes two fans to cool it.

So integrated, but still needs separate fans (just like having separate CPU and GPU would).

 

[coercitiv]

Notice it still takes two fans to cool it.

So integrated, but still needs separate fans (just like having separate CPU and GPU would).

You're using a PCB designed to accommodate both dGPU and iGPU only products in order to describe how a iGPU only product would look like.

If you're so certain there's little to no benefit in integrating the GPU, why is the Macbook Pro 13 iGPU only?

 

[cbn]

If you're so certain there's little to no benefit in integrating the GPU, why is the Macbook Pro 13 iGPU only?

That is only a 28W CPU though, so there only needs to be one fan.

In contrast, here is close-up of the 47W Iris Pro PCB and cooler:

https://www.ifixit.com/Teardown/MacBook+Pro+15-Inch+Retina+Display+Late+2013+Teardown/18696

 

[cbn]

An earlier comparison I did:

http://forums.anandtech.com/showpost.php?p=37734179&postcount=2521

As an example of different cooling gear for different TDPs I thought these teardowns of the 13.3" and 15" Apple Mac Pro Retina (late 2013) models were interesting:

https://www.ifixit.com/Teardown/MacBook+Pro+13-Inch+Retina+Display+Late+2013+Teardown/18695

https://www.ifixit.com/Teardown/MacBook+Pro+15-Inch+Retina+Display+Late+2013+Teardown/18696

The first laptop is the 13.3" model with a 28W Haswell dual core.

The second laptop is the 15" model with a 47W Haswell quad core with Iris Pro graphics.

Copper heatsink size is about the same for both, but the 47W model uses two fans (each with blowing air into a very small aluminum finned area). This compared to the 28W model which only uses one fan, but notice the aluminum finned area is much larger.

So overall, I would say both 28W and 47W set-ups have the following amount of material:

Copper heatsink material: About the same for both the 47W and 28W
heatpipe: twice as much for the 47W
fans: twice as much for the 47W
aluminum finned material (this is the part next to the fan): About the ~same total weight for both, the 47W has less of it for each fan.

So it appears the amount of metal used for cooling isn't really that different between the two TDPS, but difference in ventilation is much in favor of the 47W.

 

[coercitiv]

I ask again: if you're so certain there's little to no benefit in integrating the GPU, why is the Macbook Pro 13 iGPU only?

 

[Abwx]

I ask again: if you're so certain there's little to no benefit in integrating the GPU, why is the Macbook Pro 13 iGPU only?

I suggest that you find another thread for your thread crapping.

And the endless amount of other rubbish cases. They just copy/paste randomly from forums and make up as they go. Using wccftech as a source means you already lost. Why people ever visit that site is beyond understanding.

They said that they have some exclusive source of theirs...

Today’s update comes after we have finally managed to confirm through sources exclusive to WCCFTech when AMD plans to release its latest inception to the market.

I repost the relevant infos to get the thread back on track after all the thread crapping above this post.

we’ve learned that AMD’s planning to debut Zen based desktop FX CPUs by Q4 2016, just in time for the holiday season. While Zen based APUs are expected to debut at a later date in 2017.

.......so far it’s been made very clear to us that late 2016 and beyond is AMD’s current target.

Fortunately that’s not all we have for you today, as we’ve managed to learn about a number of additional interesting tidbits as well. There are several SKUs planned based on the Zen CPU die, with 8, 6 and 4 cores with the possibility of even higher core counts later on. The Zen core itself is surprisingly compact and power efficient. The CPU die itself isn’t particularly large and the Zen FX CPUs will target mainstream price points in which they will compete very aggressively with their upcoming Intel counterparts.

Zen FX CPUs will debut with a brand new chipset on socket AM4 that’s planned to push AMD’s connectivity, storage and entire feature set forward and much closer to parity with Intel. The upcoming FX processors will also include enhanced platform security features built directly into the hardware.

Moving over to the APU and server side of the Zen equation. APUs and enterprise products based on the new Zen microarchitecture are planned for release sometime in 2017. We’ve also learned of a particularly exciting piece of information about AMD’s next generation APUs and that they will boast performance levels that are comparable with the Playstation 4 and XBOX ONE, even inside frugal notebook chips.

http://wccftech.com/amd-zen-launch-q4-2016/

 

[cbn]

I ask again: if you're so certain there's little to no benefit in integrating the GPU, why is the Macbook Pro 13 iGPU only?

1.) 13" Mac Book Pro with 28W CPU is not a good comparison when the original topic in post #755 was something much more powerful.

2.) I have nothing against iGPU in a laptop.

But I question why AMD needs to make their iGPU so large when they also sell dGPUs.

Why not compete against Intel by offering more CPU instead? (ie, dedicate more silicon die area to CPU cores rather than iGPU). An example would have been to offer Carrizo as hexcore (3M) with 256sp or 384sp instead of 2M with 512sp.

In this way they wouldn't compete with their own dGPU division and their 28nm CPU would compare better to Intel 14nm 2C/4T.

 

[DrMrLordX]

Plan A;
Summit Ridge FX14 -> 8C/16T "Zen"
Raven Rdige FX14 -> 4C/8T "Zen" + 1 Shader Engine "GCN4" [16 CUs]

Plan B;
Viper Gecko 22FDX -> 20C/20T "Tunnelborer"
Basilisk Gecko 22FDX -> 10C/10T "Tunnelborer" + 1 Shader Engine "GCN4" [16 CUs]

What? 20c/20t? 10c/10t? How does that even make any sense? And 22FDX was announced as being used for a number of low-power/embedded applications. Plan B sounds like complete fiction.

When your posts don't line up with what the Internet STRONGMAN Juanrga has been saying, who are we to believe?

I fear it is only a matter of time before Dresdenboy becomes his next victim.

Do not mention the name of Juangra here! He will not stop with Dresdenboy. Nobody is safe from his pwnage!

 

[dark zero]

What makes you think they didn't release any on mobile? There's 3 i7 models - 5950HQ, 5850HQ, 5750HQ. There's also an i5 with two cores disabled. OEMs didn't bother.

OEM retarded mode in the purest status... I only see U tier laptops and low tier Atom craptops.... What the hell..

 

[mikk]

I still think it's a 2016 product, a paper launch though.

Given that delays are business as usual, I don't expect Zen to be available next year.

Paper launch is a different matter, they could paper launch whenever they want.

 

[Phynaz]

I suggest that you find another thread for your thread crapping.

Ahhhhahaha. He's not thread crapping, he's supporting your posistion you idiot.


Insults and name calling are not allowed here.
Markfw900

 

[Shehriazad]

Oh here's a question; AMD hasn't really done what Intel did in the past aka 4 cores or 4cores + hyperthreading (i5 vs i7). Instead they just had their weird modules that sold as X core amount.

So if AMD says 8 actual cores + hyperthreading(or even without hyperthreading)...would that not (partially) make up for the performance difference between Intel i7 and AMDs Zen that Intel will definitely have when just looking at a single core.

Like...not sure if any math can be done here, but if we had Haswell level performance but with 8 actual cores vs Skylake 4 cores + hyperthread...how would that fare?

 

[flopper]

Oh here's a question; AMD hasn't really done what Intel did in the past aka 4 cores or 4cores + hyperthreading (i5 vs i7). Instead they just had their weird modules that sold as X core amount.

So if AMD says 8 actual cores + hyperthreading(or even without hyperthreading)...would that not (partially) make up for the performance difference between Intel i7 and AMDs Zen that Intel will definitely have when just looking at a single core.

Like...not sure if any math can be done here, but if we had Haswell level performance but with 8 actual cores vs Skylake 4 cores + hyperthread...how would that fare?

depends on what they designed the cpu for.
single core performance or ipc is important still for many games and engines.
Thats the main issue for amd atm.
so adressing that, would you really make a difference with a haswell vs a skylake or 2 generations back in actual desktop use?
Most wouldnt.
so let say they are able to meet and match a skylake from today then the next generation will still be a bit dependent on die shrinks. arcitechnical changes wont do much as we seen with intel for every generation the die shrink is what allows better performance mainly.

a 14/16nm cpu with a new design from amd will rock any way due to what they have now isnt.

 

[Abwx]

So if AMD says 8 actual cores + hyperthreading(or even without hyperthreading)...would that not (partially) make up for the performance difference between Intel i7 and AMDs Zen that Intel will definitely have when just looking at a single core.

Like...not sure if any math can be done here, but if we had Haswell level performance but with 8 actual cores vs Skylake 4 cores + hyperthread...how would that fare?

Dont count on such a scenario, the infos we have point to Zen being a huge core, so it s likely that they are targeting parity with Intel line.

I even suspect that they could outdo them for FP performance on a thread per thread basis given that EXV FPU is on par with Intel s, and that Zen will have a more powerfull one.

If Intel doesnt improve their process they will be in big trouble in the perf/Watt metric given that GF s second ranked 14nm LPP variant is 20% more efficient than Intel s, and that the first ranked variant will likely add 10% on top of those 20%.

For info 20% better efficency means that GF transistors have barely 10% better conduction (or said otherwise about 10% less resistance when switched on), that s a first possibility.

It could also mean that GF s transistors have the same conductance as the Intel s but that those latter have globaly 20% higher parasictic switching capacitance, my opinion for several reasons is that the first possibility is the right one.

 

[cbn]

For Broadwell I measured each core (with 64KB L1 cache and 256KB L2 cache) as 6.8mm2 to 6.9mm2:

http://forums.anandtech.com/showthread.php?t=2433600

What do folks think Zen (with 32KB L1 cache and 512KB L2 cache) will measured out as?

 

[MrTeal]

I'm not sure why a 1024 shader 4C/8T APU shouldn't be possible from AMD within a reasonable die size. Kaveri is 245mm^2 with two Steamroller modules and 512 GCN1.1 shaders on a 28nm process. The move to GloFo's 14nm process probably won't give the expected 4x density, but a doubling of cores and shaders in a 200mm^2 isn't unreasonable.
The bigger question is how well a dual channel DDR4 interface could keep that kind of APU fed. Even the 512 shader HD7750 showed a massive performance difference between the GDDR5 version and the DDR3-1600 version, and Kaveri keeps on increasing gaming performance with higher frequency RAM up past 2866. Assuming DDR4-3200 is widely available and cheap by then that will help, but four fast Zen cores and a 270X class GPU could be very bandwidth constrained even if thermals and die size look good.

 

[jpiniero]

Even dual channel DDR4 3200 is only ~50 GB/sec compared to the 270's ~180. They would really need to do something additional to make it work... there's been talk about Intel moving to dual EDRAM for instance.

 

[Shehriazad]

Even dual channel DDR4 3200 is only ~50 GB/sec compared to the 270's ~180. They would really need to do something additional to make it work... there's been talk about Intel moving to dual EDRAM for instance.

If AMD makes some monster APU that would be SEVERELY bandwidth starved by any type of dual channel ram...wouldn't it be "obvious" that they'd use something like an L4 cache consisten of HBM or something like EDRAM?

 

[Fjodor2001]

the infos we have point to Zen being a huge core

Perhaps huge in terms of that it has the performance and design properties of a typical high performance "big core" CPU, but the die itself should be quite small according to WccfTech:

"The Zen core itself is surprisingly compact and power efficient. The CPU die itself isn’t particularly large and the Zen FX CPUs will target mainstream price points in which they will compete very aggressively with their upcoming Intel counterparts."

 

[Arachnotronic]

Perhaps huge in terms of that it has the performance and design properties of a typical high performance "big core" CPU, but the die itself should be quite small according to WccfTech:

"The Zen core itself is surprisingly compact and power efficient. The CPU die itself isn’t particularly large and the Zen FX CPUs will target mainstream price points in which they will compete very aggressively with their upcoming Intel counterparts."

Recall that WCCFTech also said that Skylake would have MorphCore...

Would be more credible if they gave specifics that we could then compare to the actual dies when they are released.

 

[Fjodor2001]

Recall that WCCFTech also said that Skylake would have MorphCore...

They never said it would have MorphCore for certain, it was just speculation from their side IIRC.

Anyway, you could use your argument for everything, which would mean that once a site has published a single piece of information that later proves to be incorrect, no info on that site can ever be trusted anymore. If you go by that standard, I don't think there are many sites on the Internet that can be trusted, if any.