*** AMD "Zen" thread (inc AM4/APU discussion) ***

[pastymuncher]

Humbug, on your game screenshots what are you using to give the onscreen stats in the top left corner or is that actually part of your game?

 

[Mr.Fantastic]

It's called MSI Afterburner

http://www.guru3d.com/files-details/msi-afterburner-beta-download.html

 

[pastymuncher]

Didn't know Afterburner gave you cpu load and temps as well. Been using EVGA Precision and that only gives you gpu data.

 

[theforce]

I definitely want AMD to offer a more competitive product and Zen should hopefully present a nice upgrade from my current i5. It's the same feeling l had before purchasing my current cpu after reading the disappointing reviews for Bulldozer. Same before it with Q6600 vs Phenom.

I wish Zen will offer me a logical upgrade path but in all honesty l am not in a hurry to upgrade in terms of CPU performance. I am more interested in the Zen APU with HBM coming in 2017. While it might target HPC for 2017, the 2018 version of the chip might find it's way to the desktop. I am wondering if they will use or if it's possible to use the HMB as an L4 cache for the CPU.
Let's say your game utilizes 7 out of 8GB of memory the CPU could have a nice amount of L4. It should be quicker than having to go to RAM.

 

[kaku]

Didn't know Afterburner gave you cpu load and temps as well. Been using EVGA Precision and that only gives you gpu data.

I think they both leverage RivaTuner for the OSD which can be configured separately. By default it only shows GPU-related stuff, but in the options you can add other things to the OSD.

 

[opethdisciple]

I'm not keen on CPU's with integrated graphics cores. They should have two types one with integrated graphics and maybe locked and at a cheap price and another without the graphics core, fully unlocked at a higher price point.

I'm really holding out for a decent upgrade from my 3570k. But I want something with real physical cores rather than HT.

However my sister wants a new PC in Dec, so I might flog her my existing gaming rig and get whats good at the time. Which will probably be a 6700k. Unless we are just months off Zen then I might do with out gaming for a little while and wait.

 

[vakuza]

The real question with Zen is will we see apt motherboards? There's not much point going with Zen if the motherboard isn't good for what you want, whether it's 24/7 operation, looks or overclocking ability.

 

[Martini1991]

The real question with Zen is will we see apt motherboards? There's not much point going with Zen if the motherboard isn't good for what you want, whether it's 24/7 operation, looks or overclocking ability.

I don't think this will be a concern, my only possible concern one would be ITX capability.

When AM3 first launched, the motherboards were fine.

 

[CAT-THE-FIFTH]

Someone mentioned this on SA forums:

http://dl.acm.org/citation.cfm?id=2934586

Does that mean Bristol Ridge and Stoney Ridge use FIVR??

 

[piehole]

I'm hoping Zen will be comparable to at least Haswell. I won't say I'm an AMD fanboy, but I do try and support them when possible (one of my machine is an 8320 and I have a Fury X and R9 390).

 

[Emlyn_Dewar]

If there isn't an ITX board available from early on in Zen's life, I'll be very sad.

 

[mmj_uk]

Shipping with limited quantities indicates that it's a very good chip, as good or better than they expected. Otherwise, they would build an inventory before selling to capitalise on hype. I'm also betting that this is a performance at a price that customers will hold off upgrading for, even if they can't get one for a month or so.

I'm expecting this to be good.

If it was that a good a chip then why would they restrict quantities? makes no sense. If quantities are low it'll be difficult to manufacture and have low yields but that is still no gauge of performance. I suppose you can argue that manufacturing difficulty means a big powerful chip but again AMD have a history of making bigger chips with fairly ordinary performance.

 

[Scougar]

Shipping with limited quantities indicates that it's a very good chip, as good or better than they expected. Otherwise, they would build an inventory before selling to capitalise on hype. I'm also betting that this is a performance at a price that customers will hold off upgrading for, even if they can't get one for a month or so.

I'm expecting this to be good.

Don't do it to yourself.

 

[Panos]

Don't do it to yourself.

I agree.

It wont be a good overclocker at 14nm, see Intel's predicaments at that node size. Broadwell-E running hot, burning gazillion watt and struggling to overclock....

Who wants to put a bet that AMD going to have exactly the same issues, and tbh all CPUs from now on going to have that issue, from both companies?

 

[AthlonXP1800]

http://wccftech.com/amd-zen-es-benchmarks/

First Zen ES Ashes of the Singularity benchmark leaked.



Zen ES has 2.8GHz base clock and 3.2GHz turbo clock but the performance is slower than Haswell 4790 CPU, it not at skylake level AMD claimed months ago.

Zen ES are seemed underwhelmed for 8C/16T, if the final retail are similar or slight improved peformance at higher base and turbo clock then guess I will wait for Kaby Lake.

 

[drunkenmaster]

It's not at the Skylake level which AMD claimed months ago.

Now first off AMD haven't claimed it was at Skylake level, second, a gpu benchmark and a single benchmark at that isn't a way to claim where performance is at and lastly this isn't shipping clocks. What AMD claimed was 40% faster than.. something, people argue over Bulldozer, Excavator, etc. This was 38% faster and if the retail chips came at say 3.6Ghz boost speed it would be somewhere around 55% faster than Bulldozer, which would probably be around 40% faster than Excavator... which is actually what AMD always claimed.

Second, http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/15

Oh wait, look how much massively faster the 6700k is in gaming over a 4770k.... in some games it's slower than an Ivy bridge let alone the lower clocked Haswell than the one from the Aots benchmarks.

Third

http://semiaccurate.com/2016/03/01/investigating-directx-12-cpu-scaling/aots-dx12-cpu-scaling-2/

As you can see, over 6 threads reduced performance(well cores, they didn't do 3 cores + HT) so can't see if that would be better or worse than 6 real cores. Assuming that because it's a real 8 core THIS benchmark should be much faster if it has more cores is simply illogical, it's not a CPU benchmark it's a game and wasn't designed with a purely 8 core 16 thread CPU in mind.

The ES chip has below shipping clock speeds, so using the current performance as an idea of final performance couldn't be more flawed. That it's not faster than a i7 Haswell you're using as proof it's no where near Skylake, despite the reality being that Skylake is often no faster at all in gaming benchmarks than Haswell. Games aren't perfect CPU benchmarks. It's a monumental leap from Bulldozer and that is what Zen claimed to be. With higher clock speeds the gap to Bulldozer only increases which means AMD is dramatically more competitive than they were before. Going from lets say 40fps to 58fps compared to a more expensive Intel chip getting 65fps, is a massive improvement whether you choose to ignore it or not.

 

[AllBodies]

It's not at the Skylake level which AMD claimed months ago.

Now first off AMD haven't claimed it was at Skylake level, second, a gpu benchmark and a single benchmark at that isn't a way to claim where performance is at and lastly this isn't shipping clocks. What AMD claimed was 40% faster than.. something, people argue over Bulldozer, Excavator, etc. This was 38% faster and if the retail chips came at say 3.6Ghz boost speed it would be somewhere around 55% faster than Bulldozer, which would probably be around 40% faster than Excavator... which is actually what AMD always claimed.

Second, http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/15

Oh wait, look how much massively faster the 6700k is in gaming over a 4770k.... in some games it's slower than an Ivy bridge let alone the lower clocked Haswell than the one from the Aots benchmarks.

Third

http://semiaccurate.com/2016/03/01/investigating-directx-12-cpu-scaling/aots-dx12-cpu-scaling-2/

As you can see, over 6 threads reduced performance(well cores, they didn't do 3 cores + HT) so can't see if that would be better or worse than 6 real cores. Assuming that because it's a real 8 core THIS benchmark should be much faster if it has more cores is simply illogical, it's not a CPU benchmark it's a game and wasn't designed with a purely 8 core 16 thread CPU in mind.

The ES chip has below shipping clock speeds, so using the current performance as an idea of final performance couldn't be more flawed. That it's not faster than a i7 Haswell you're using as proof it's no where near Skylake, despite the reality being that Skylake is often no faster at all in gaming benchmarks than Haswell. Games aren't perfect CPU benchmarks. It's a monumental leap from Bulldozer and that is what Zen claimed to be. With higher clock speeds the gap to Bulldozer only increases which means AMD is dramatically more competitive than they were before. Going from lets say 40fps to 58fps compared to a more expensive Intel chip getting 65fps, is a massive improvement whether you choose to ignore it or not.

Good summary.

And also, in general, if it did turn out Zen gave ~10% less FPS in games than Intel, but offered 2 or 4 cores extra for the same money, I bet many people would take that trade. I certainly would.

 

[AthlonXP1800]

After searched for my CPU frame performance on ATOS standard 1080p benchmark database, surprised to see Zen ES is about 20% SLOWER than my 4 years old 3770K CPU.

Here is list of CPU framerate how powerful Intel CPU performance over the last 5 years compared against Zen ES:

Broadwell-E 6950X 10C/20T: 127.9
Skylake 6770K 4C/8T: 98.8
Devil's Canyon 4790K 4C/8T: 87.7
Haswell 4770K 4C/8T: 69.8
Ivy Bridge 3770K 4C/8T: 69.7
Sandy Bridge 2600K 4C/8T: 70.3
Zen ES 8C/16T: 58
Sandy Bridge 2500K 4C/4T: 47.9

Zen ES running at 3.2GHz turbo CPU performance sit on Sandy Bridge level between 2500K and 2600K, really very poor for 8C/16T. AMD probably will set 3GHz as final base clock in in the next few months that would bring CPU performance closer to 2600K. Look like I definitely will get Kaby Lake 7700K after saw poor Zen ES CPU score in first benchmark, it would be a downgrade from my 3770K CPU if I go for Zen CPU.

 

[Martini1991]

Good summary.

And also, in general, if it did turn out Zen gave ~10% less FPS in games than Intel, but offered 2 or 4 cores extra for the same money, I bet many people would take that trade. I certainly would.

Depends if it performed better when all cores are in use, if not, why would you want more cores but lesser performance overall? If it's 10% slower per core, then that's fine, but if it's 10% slower 8 core versus a 4 core i7 when they're going full whack, then that's not good enough.

Either way, I don't think a single Singularity result is enough to start worrying over.

 

[DragonQ]

http://wccftech.com/amd-zen-es-benchmarks/

First Zen ES Ashes of the Singularity benchmark leaked.



Zen ES has 2.8GHz base clock and 3.2GHz turbo clock but the performance is slower than Haswell 4790 CPU, it not at skylake level AMD claimed months ago.

Zen ES are seemed underwhelmed for 8C/16T, if the final retail are similar or slight improved peformance at higher base and turbo clock then guess I will wait for Kaby Lake.

If you compensate for the clock speed difference, Zen is faster than the rest in that benchmark figure you showed. It'll be a while before we know how well Zen clocks though. I have no idea what the hell it's testing though, does it include the IGPs or are they using a separate GPU?