facebook rss twitter

French print magazine publishes AMD Ryzen ES benchmarks

by Mark Tyson on 27 December 2016, 10:11

Tags: AMD (NYSE:AMD)

Quick Link: HEXUS.net/qadclt

Add to My Vault: x

Over the weekend a French PC hardware print magazine published a review of an AMD Ryzen ES (Engineering Sample) tested upon a beta motherboard (via Guru 3D). The article author is said to be the same person behind the famous CPU-Z utility and warns that the bench test results from the octacore AMD Ryzen ES might not be very similar to those achievable from the final shipping processor.

Canard PC tested an AMD '2D3151A2M88E4' processor. This is but one of four ES processors distributed by AMD to its hardware partners (there were 2x quad core and 2x octacore chips as detailed above). These ES processors seem to run at conservative clock speeds. For example the Canard PC sample was clocked at base/boost of 3.15/3.4GHz. Note that AMD announced that top-end shipping Ryzen chips would have a 3.4GHz base clock.

A trio of benchmark tables was assembled by Canard PC encompassing; Utility benchmarks, gaming benchmarks, and a power consumption comparison chart. You can see that despite the ES constrained clock speeds and beta motherboard the AMD processor does pretty well overall. In the utility benchmarks it fares best for now.

In gaming it seems to lose some ground, for now. The AMD engineering sample is only actually 10 per cent behind the similar configured (8C/16T) Intel Core i7 6900K. A final retail chip with higher clocks plus an updated motherboard could easily catch up.

The Ryzen ES chip tested by Canard PC consumed about 93W. AMD rates its 14nm 8C/16T Ryzen chip at 95W. The above benchmarks are interesting indicators but we will still have final shipping hardware to test, and pricing to consider.



HEXUS Forums :: 18 Comments

Login with Forum Account

Don't have an account? Register today!
Some more details - it is an A0 sample,and the motherboard has an old AGESA firmware version,meaning it can't run later ES and Turbo is not fully functional(apparently). The CPU basically did not actually Turbo past 3.3GHZ at maximum.

Regarding the games,4 to 5 of them don't really scale beyond 4 cores(apart from BF4 OFC). 4 of the games(maybe 5) show no gains with SMT.

The Core i7 6900K has a base clockspeed of 3.2GHZ but will tend to maintain Turbo of between 3.5GHZ to 3.7GHZ looking at reviews.
I guess these are sort-of interesting, but given the caveats ….

- ES not final shipping processor

- “conservatively” clocked

- no pricing

…. I'm not sure what practical, real-world value they have.
It also seems the motherboard is an earlier BIOS and firmware version that the one AMD used with a Titan X. The motherboard would not boot with a GTX1080 so they used a Fury X and AMD drivers are more single core limited,which also hints at decent IPC.

Saracen
I guess these are sort-of interesting, but given the caveats ….

- ES not final shipping processor

- “conservatively” clocked

- no pricing

…. I'm not sure what practical, real-world value they have.

They have a use,since we can estimate Ryzen IPC,which in gaming seems relatively close to BW-E level,and between Haswell and BW-E level for non-gaming stuff and these are the worst case scenarios too.

The games are easier to estimate as 3 of them don't scale beyond 4 cores,and seem to be more dependent on single core performance,one seems to show 90% of of its performance on the first 4 cores,and uses a 5th core slightly but is also dependent on high single core performance - the AMD FX8350 is beaten by a Core i3 4330 in those 4 games.

None of them scale with SMT,which means AMD improvements in performance are mostly down to core IPC improvements,not a better form of SMT over Intel.

One game is really more GPU limited,once you hit a certain performance level of a CPU and the other scales to around 8 cores.

The other factor is the A12 9800 quad core is only slightly slower than an FX8350. Both have similar core performance,but the A12 9800 can suffer under more multi-threaded situations due to lack of L2 cache. The fact that they are so close together also hints at the gaming suite being more favourable of single core performance than a large amount of cores or threads.
But “estimate”, when not knowing how final production chips will compare, or how price-competitive they'll be.

Academically interesting, perhaps, but real-world useful? I certainly wouldn't base commercial decisions on it.
Saracen
But “estimate”, when not knowing how final production chips will compare, or how price-competitive they'll be.

Academically interesting, perhaps, but real-world useful? I certainly wouldn't base commercial decisions on it.

Look at my updated previous post - IPC performance is generally almost BW-E level and AMD CEO Lisa Su is on record saying the retail CPUs will be having a 3.4GHZ base clockspeed.

At that level of clockspeed,it is incredibly close to a Ā£1000 Core i7 6900K over both gaming and non-gaming benchmarks. Like I said in 4 of those games,a Haswell dual core beats an eight core FX8350.

The sample tested had a 3.1GHZ base clockspeed - the AMD retail samples will starting at nearly a 10% higher clockspeed.

The Intel CEO is on record as saying Ryzen will be disruptive during 2017,and from that article I read it is strongly hinting Intel is betting on 10NM to fight Ryzen.

Edit!!

Also the editor behind CanardPC who leaked these results,leaked the first 1.4GHZ Athlon 64 results months before launch.

We can kind of see where that one went! :)

What is the bigger unknown is whether AMD is launching 4C/8T and 6C/12T models and how high they will be clocked at,or whether AMD is skipping the 4C/8T models and moving to the Zen APU.

Every leak so far has only shown 8C/16T IIRC and no lower core count versions. AMD still needs a counter to the 4C,4C/8T and 6C/12T Intel CPUs which are probably their biggest selling desktop performance CPUs.