AMD Zen Preview

[MJVR1]

http://www.amd.com/en-us/innovations/new-horizon

They're going to be showing off the capability of Zen CPU's today. Seems like its mostly going to be gaming oriented. The live stream is embedded below if you want to check it out. FIngers crossed they show off a 65w TDP part. I really want to go full AMD for my next system.

 

[GuilleAcoustic]

Pretty late here, but from what I saw it is very promising !

 

[MJVR1]

Here is a TL/DW summary of the event.
AMD has released more details on their upcoming ZEN CPU's. The processors with the Zen architecture will now fall under the brand of Ryzen. Although AMD did not dive into specifics about their upcoming CPU's, they did reveal some technologies that helped they achieve and surpass their performance goals.
The Ryzen CPU's have 5 new technologies that helps boost their performance and efficiency goals. These are branded as SenseMI and use machine learning to optimize the CPU workload.
AMD did show off the performance of a ZEN CPU with the following specs:

1. 8 Core/16 Threads
2. Base clock of 3.4Ghz.
3. 20MB of L3 Cache
4. 95W TD

Benchmarks: In the VERY brief Benchmarks, they compared the aforementioned CPU with Intel 6900K, an 8C/16T chip. With both chips are stock, both CPU'S were neck in neck with the AMD CPU coming out on top in a couple of Benchmarks.
Most notably, the Zen CPU beat the Intel in a game stream environment and in handbrake. Since we don't know the settings at which the Benchmarks were run at, take the results with a grain of salt. Furthermore, we did not get exact data for some of the benchmarks. We will have to wait for review samples to get the information we want.
At the end of the event we did get to see a new Vega GPU in action. It ran the upcoming Battlefront Rogue One DLC at 4k. Again, the only information we got was "It's running at higher then 60FPS." We don't have any data you compare against. So yeah...
¯\_(ツ)_/ ¯

 

[Therandomness]

If they can remake the Nano and have it not draw 400W when spiking, I'd happily get one. Would make a perfect replacement for my half-dead 290 xd

 

[craigbru]

I really like the lower TDP. I'm not sure I could justify a new build built around this guy, but it's sure fun to think about...

 

[Phuncz]

So this happened: https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/
Apparently it's a printed review (French, Canard PC magazine) of an engineering sample.

The application benchmarks seems to show the multi-core performance is very, very real. Yet the game benchmarks show that it is seriously lacking, maybe the single-core performance ? Time will tell if they'll uprate the clocks before going on sale, but I for one am impressed. If the clocks will be higher, like 3.8GHz, I'm sure it will put Sky Lake and Kaby Lake in its place. If it is kept at the 3.15GHz the enfineering sample is at, I hope there is plenty of headroom to overclock.

But this is looking mighty interesting. It's almost set in stone that my next system will be AMD. ASAP.

 

[FCase]

Good to hear @Phuncz. I was thunking about waiting for ZEN for an all AMD system. This pretty much seals it. I hope I do not have to wait too long for an mITX board.

FCase

 

[IntoxicatedPuma]

I think AMD is still fighting an uphill battle. Per core, the Intel's still appear to have a pretty strong lead, and it seems instead of going for multi-threaded performance, programs and games are putting computing onto the GPU I was thinking if I built a new PC, what would I switch to. Right now I think for gaming I'd go with the upcoming i3 K processor that is rumored. Maybe if I needed a budget workstation I'd go for the Zen. I hope this brings Intel prices down at least, or maybe AMD can unleash a 5ghz -capable quad core CPU to go against the i3 K.

 

[IntoxicatedPuma]

Techpowerup is saying a French review of Zen was leaked.

Performance seems good but I think we really need to see single threaded performance to tell how good. If single threaded performance is at least on par with Sandy Bridge, then maybe we got a good race going - especially if AMD is serious about the Zen chip having a lot of room for improvement.

 

[3lfk1ng]

They ZEN "leak" is using a 3.1GHz Engineering Sample. The final ZEN product has a base clock of 3.4GHz.
It's a good taste for what's to come but it's not entirely accurate.

 

[Phuncz]

I posted that link 3 posts up
I'm very excited, considering the workstation performance and potential for gaming performance if it is clocked higher. I can't wait for mid January, when it is expected to be fully detailed and with a rumoured release of February.

 

[IntoxicatedPuma]

I posted that link 3 posts up
I'm very excited, considering the workstation performance and potential for gaming performance if it is clocked higher. I can't wait for mid January, when it is expected to be fully detailed and with a rumoured release of February.

Ah sorry, I didn't read very carefully.

I'm excited for some of the stuff at CES. Maybe we need a thread for that!

 

[Phuncz]

Oh I'm pretty sure there will be a CES thread soon, SFF.net is going to be there according to the podcast

 

[Kmpkt]

Pretty sure @[URL='https://smallformfactor.net/forum/members/3lfk1ng.473/']3lfk1ng [/URL]is going.

 

[jeshikat]

Pretty sure @3lfk1ng is going.

Correct.

 

[eutholevonik]

Did anyone else notice during AMDs ryzen test against a 6900k, that ryzen was only using 3-4 watts less then the 6900k under load? Plus you have to remember that during the test, turbo boost was disabled on ryzen and enabled on the 6900k.

Seems a little high to me since AMD is saying ryzen is 95w TDP compared to Intel's 140w TDP.

 

[|||]

Oh I'm pretty sure there will be a CES thread soon, SFF.net is going to be there according to the podcast

I will be going on business, but not sure what parts I will be in, as such (more IoT focused).

They ZEN "leak" is using a 3.1GHz Engineering Sample. The final ZEN product has a base clock of 3.4GHz.
It's a good taste for what's to come but it's not entirely accurate.

So long as I/O isn't bottlenecking the processor (it most likely isn't the case) the processor should scale linearly with clock frequency. Turbo throws a little wrench in there, but we should be seeing an even greater boost to the final production turbo frequency over the engineering sample, there should be a minimum of 8% performance increase over this sample.

 

[3lfk1ng]

Ya and even then AMD's Neural Net Prediction technology throws another wrench into the equation. Rumors are abound that the Neural Net Prediction is responsible for that extra ~10% of unexpected IPC gains.

Either way, we're just a few days away from the floodgates opening.

 

[Therandomness]

Did anyone else notice during AMDs ryzen test against a 6900k, that ryzen was only using 3-4 watts less then the 6900k under load? Plus you have to remember that during the test, turbo boost was disabled on ryzen and enabled on the 6900k.

Seems a little high to me since AMD is saying ryzen is 95w TDP compared to Intel's 140w TDP.

Well, afaik the method to measure TDP was never standardised so AMD and Intel can get vastly different results. Also, TDP isn't a direct measurement of power consumption, and is meant to measure the heat produced..

 

[eutholevonik]

Well, afaik the method to measure TDP was never standardised so AMD and Intel can get vastly different results. Also, TDP isn't a direct measurement of power consumption, and is meant to measure the heat produced..

You are absolutely correct. I just figured overall system draw would have been better then a 4 watt difference.