Overclock.net banner

[CPC] First unofficial Ryzen benchmarks

153K views 1K replies 241 participants last post by  cjwilson 
#1 ·




AMD 2D3151A2M88E
Base clock : 3.15GHz
Turbo all core : 3.3GHz
Turbo one core : 3.5GHz

First results from the famous Doc Teraboule, french hardware guru.
Results are really good, with around +35% better performance than FX8370 at the same clock. Power consumption is really good too.
If AMD achieve to reach better clocks, Intel wll be in trouble and that's a really good news, competition is better for everyone !

http://twitter.com/CPCHardware/status/811883423371526144

Thanks to Samuel for this early results.
 
See less See more
2
#3 ·
Wait for mark.

.
.
.
.
.
BENCH MARK
 
  • Rep+
Reactions: clerick
#4 ·
Quote:
Originally Posted by M4c4br3 View Post

Well, looks to be crap for gaming, but can't really see the clocks. If this thing can compete with my 6700k at 4.7ghz, then great, I love competition, but if it's even 10% slower, then I'll stick to Intel.
Clocks are a little bit low in this review, we'll see with final clocks.
But compared to old FX-8370 that's excellent results.
It will be hard to beat 7700K of course but AMD is back with good gaming performance.

edit : Instructions, latency and throughput :

 
#6 ·
Quote:
Originally Posted by M4c4br3 View Post

Well, looks to be crap for gaming, but can't really see the clocks. If this thing can compete with my 6700k at 4.7ghz, then great, I love competition, but if it's even 10% slower, then I'll stick to Intel.
Please keep in mind they threwARMA in there for gaming benchmarks...

The fact that it's still averaging above a skylake i5 with Arma in the mix and the other games listed aside from BF1 not being decently threaded I'd say it's in a pretty good spot - especially if it's clocked lower than what we can expect..

Regardless - as always there will be fake benches out there so take this with a huge pile of salt and wait for official benchmarks
 
#7 ·
Quote:
Originally Posted by SoloCamo View Post

Please keep in mind they through ARMA in there for gaming benchmarks...

The fact that it's still averaging above a skylake i5 with Arma in the mix and the other games listed aside from BF1 not being decently threaded I'd say it's in a pretty good spot - especially if it's clocked lower than what we can expect..

Regardless - as always there will be fake benches out there so take this with a huge pile of salt and wait for official benchmarks
+1 THIS. ARMA as a game benchmark is a ridiculous test for this CPU as this CPU has low clocks but high number of cores. So the fact Ryzen is competing with similarly clocked i5s is actually quite good.
 
#8 ·
Quote:
Originally Posted by M4c4br3 View Post

Well, looks to be crap for gaming, but can't really see the clocks. If this thing can compete with my 6700k at 4.7ghz, then great, I love competition, but if it's even 10% slower, then I'll stick to Intel.
I don't really know why people expected different. Single core performance is still the biggest factor that determines how well a CPU performs in games.
 
#9 ·
Brute performance and power consumption is impressive. Even the gaming numbers are decent given the library of games they used.

Ryzen is looking more and more like AMD promised. Still have to wait for final reviews before I start doing cartwheels however.

Who is this reviewer? I love his comments about Intel at the end. So very true....
 
#11 ·
Quote:
Originally Posted by CriticalOne View Post

Quote:
Originally Posted by M4c4br3 View Post

Well, looks to be crap for gaming, but can't really see the clocks. If this thing can compete with my 6700k at 4.7ghz, then great, I love competition, but if it's even 10% slower, then I'll stick to Intel.
I don't really know why people expected different. Single core performance is still the biggest factor that determines how well a CPU performs in games.
Usually, but not always.
Unlike four years ago multi-threading in games is a thing now, IPC can't go up forever.

All AMD has to do is be "competitive" in single core performance and ship something that has more CPU cores instead of a useless iGPU.

I might even prefer a CPU without any iGPU just for the fact that having two GPU's screws with Windows so much, especially in laptops.
 
#15 ·
Quote:
Originally Posted by ILoveHighDPI View Post

Usually, but not always.
Unlike four years ago multi-threading in games is a thing now, IPC can't go up forever.

All AMD has to do is be "competitive" in single core performance and ship something that has more CPU cores instead of a useless iGPU.
Mutlitheading doesn't change the fact that most games will still have one or two heavy threads that the rest of the threads will have to synchronize to, still creating an IPC bottleneck. Many parts of games can't be parallelized, so single threaded performance will still matter unless there is a very big paradigm shift in computer programming and mathematics.
 
#16 ·
Quote:
Originally Posted by Olivon View Post

For Raven Ridge :

4C/8T
11CU, 704 units, 1.2GHz for ~ 2TFlops, around RX 460 level.
Now, if AMD starts releasing cut down version of this for the mobile market, at 4C/4T, i am going to drop my money for such laptop.
 
#17 ·
Quote:
Originally Posted by CriticalOne View Post

Mutlitheading doesn't change the fact that most games will still have one or two heavy threads that the rest of the threads will have to synchronize to, still creating an IPC bottleneck. Many parts of games can't be parallelized, so single threaded performance will still matter unless there is a very big paradigm shift in computer programming and mathematics.
Multi-threading might really take off this generation, both Xbox One and PS4 have 8 Jaguar cores from AMD, and they are lot weaker than Piledriver, they do fine, almost all the games from big studios will be creating games for them and they'll port it to PC bringing with them their idiosyncrasies.

Last gen, Xbox 360 and PS3 had different CPU architecture, this time both are x86_64.
 
#20 ·
Multithreading is a very complicated and long process that meets the point of rapidly diminishing returns fairly easily. Not everything in a game can be multitheaded, or is even worth multithreading.

People have been saying to expect games to become a lot more parallel since the Phenom II X6 CPUs launched. Since then,what has happened is that 6 and 8 core Intel processors have at most a marginal lead over the 4 core i7 processors with there still being new games out there where the 4 core i7 is still the fastest. These 4 core i7s are still very competitive CPUs.

Would I like to see mulitthreading in video games take off? Yes. However, the chances of us getting much more performance out of games from improved multithreading is low.
 
#22 ·
Quote:
Originally Posted by M4c4br3 View Post

Well, looks to be crap for gaming, but can't really see the clocks. If this thing can compete with my 6700k at 4.7ghz, then great, I love competition, but if it's even 10% slower, then I'll stick to Intel.
The gaming benchmarks are fine for 3.3-3.5GHz.

I wouldn't expect it to match your 4.7GHz 6700K if gaming is your only concern, but we'll have to see how it OCs.
Quote:
Originally Posted by Gilles3000 View Post

Must be an engineering sample, as AMD stated the lowest base clock for any Ryzen CPU would be 3.4Ghz. And engineering samples aren't always very representative of performance in every application.
This close to launch most any ES part in circulation should be almost perfectly reflective of final performance. Indeed, if it even made it beyond AMD's hands, it's probably a nearly final stepping...otherwise it wouldn't be very useful for validating the hardware/firmware ES samples are sent out for.

I'm not convinced there will be no Ryzen parts below 3.4GHz base clock either. My interpretation of Lisa Su's statement and the AMD slides was that there would be at least one 8-core Ryzen SKU with a 3.4GHz+ base clock, but not that every Ryzen SKU would be that fast.
 
  • Rep+
Reactions: mohiuddin
#23 ·
I look forward to a mobile raven ridge at a 35-45 watt tdp but unless AMD makes some serious inroads to get better specced devices to market, I will be dissapoint.
There isn't even a single 35 watt Bristol ridge laptop on the market. And by the benches I have seen that line of CPU's is actually competitive even at 15 watt tdp
 
#24 ·
Salut Doc Teraboule! They are not bad at all, if true. Huge grain of salt though, as always.

Keep in mind that the 6900K should be our point of reference, since it is also an 8 core CPU. The other is that the 6900K costs approximately $1050 USD on Newegg right now. https://www.neweggbusiness.com/product/product.aspx?item=9b-19-117-645

The rumors are that Ryzen is going to be $350 USD for the 8 core and apparently $500 USD for a binned 8 core variant. Keep in mind that the motherboards for X370 on AMD are likely to be cheaper as well than X99. Zen isn't going to be the "absolute best" for those trashing on this, but it's going to be a "good enough" solution at a competitive price. That's the whole idea - price to performance and marketshare.

From a price to performance POV, that's not bad at all considering the IPC. They made their 40% IPC targets and actually exceeded them. It's close to what I expected Zen to be at.

  • Computer performance: 193.4 / 168.7 = 14.64 % for 6900K advantage
  • Gaming performance: 107.4 / 997.3 = 10.38 % for 6900K advantage
  • Power consumption: As in the demo, the Ryzen uses a bit less energy, 3.22% to be exact versus the 6900K.
Edit: Of course this doesn't take into account clockspeed differences (we need to know what the release clockspeeds will be). New information was released after this post in regards to clockspeed; will update in a post below.

The power consumption is impressive considering the Intel CPU is at 14nm and the AMD CPU is using a 14/20nm hybrid processes from Samsung made by Global Foundries. They are measuring from the ATX 12V connector as well. I would like measurements from the way,, but power consumption looks good.

It's possible that the compute performance is being weighed down by the a slow AVX implementation. I'd be very interested to see the AVX and non-AVX split, then the AVX split between AVX128 (like on Sandy Bridge) and AVX256 (introduced with Haswell).

As Blameless noted, this is pretty close to launch. We are either at release stepping or near that. Maybe a respin might give slightly higher clocks but that is it. The IPC is acceptable, and there is the possibility that this is the release clock or that it is slightly faster (not much faster this close to release silicon).

The only question is the OC headroom. It's not as good as Haswell, but if these benchmarks are true, we're at the point where it is "good enough". If it does go over 4 GHz, then it is actually a pretty good deal, if they price it they way we expect them to.

I'm actually optimistic too for the Zen Opterons. If they price it a bit aggressively (which they will), the Naples platform with its 8 channel memory controller may steal some marketshare from Intel and give AMD some much needed cash. I suspect that it is just 4 dual channel memory controllers together connected via a high speed MCM link. It's kind of like what Intel did for their HCC dies, only on separate chips. I wonder though how fast the MCM links are for Naples. That's a really interesting question.

Merci beaucoup for posting this.

Quote:
Originally Posted by CriticalOne View Post

I don't really know why people expected different. Single core performance is still the biggest factor that determines how well a CPU performs in games.
With DX11, that is the case, because DX11 can only access one core from the GPU.

DX12 does change that, so it is no longer an assurance. So too do the games that use Vulkan.
 
#26 ·
Quote:
Originally Posted by Blameless View Post

This close to launch most any ES part in circulation should be almost perfectly reflective of final performance. Indeed, if it even made it beyond AMD's hands, it's probably a nearly final stepping...otherwise it wouldn't be very useful for validating the hardware/firmware ES samples are sent out for.

I'm not convinced there will be no Ryzen parts below 3.4GHz base clock either. My interpretation of Lisa Su's statement and the AMD slides was that there would be at least one 8-core Ryzen SKU with a 3.4GHz+ base clock, but not that every Ryzen SKU would be that fast.
Hmmm, you're probably correct.

I don't think a 3.4Ghz minimum base clock would be unreasonable for the Ryzen desktop sku's. Some of the low power and mobile sku's will probably have lower clocks though.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top