AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[USER8000]

I am shocked by the result - its around 8% to 10% slower than a Core i7 6900K,which is probably running at a slightly higher clockspeed.

So,AMD actually managed to get close to Broadwell level IPC. That is much better than any of us thought they would get.

Edit to post.

If the Intel system is running quad channel memory,it does indicate the Ryzen memory controller is also probably up to scratch too.

 

[bjt2]

TBM 3.0 is a feature and with websites testing with that feature turned on we have to wait and see what the actual settings are for 6900k with each site. Similarly for Ryzen final clocks are going to be the real deal. We already know Lisa Su has confirmed base clocks of 3.4 Ghz or higher. The all core turbo and max turbo are the key for Ryzen to make a good impression. No wonder AMD want to keep those details closer to launch.

FX8370E managed a 3.3 base and 4.3 max turbo @ 95W on the 32nm SOI. They declared same energy/cycle than XV, and the 8370E, on the known graph of efficiency and IPC, draws even more than XV. if the base clock is 3.4GHz, i would expect at least 4.4GHz turbo max, near my forecast of 1 month ago (4.5).

 

[mikk]

Power consumption is in line with AMDs test roughly at i7-6900k level. IPC is way below Skylake I would say. Because i5-6500 and i7-6900k run at a similar clock speed and Broadwell is ~12% above Skylake in the gaming tests, which shows that a thread/core advantage makes a decent difference in the gaming tests.

 

[inf64]

Power consumption is in line with AMDs test roughly at i7-6900k level. IPC is way below Skylake I would say. Because i5-6500 and i7-6900k run at a similar clock speed and Broadwell is 12% above Skylake in the gaming tests, which shows that a thread/core advantage makes a decent difference in the gaming tests.

Compensate for clock speed difference between Ryzen and 6900K (ST Turbo 3.7Ghz vs 3.4Ghz as shown in the charts) and voila you are almost at the Broadwell-E level again: 97.3x 3.7/3.4=105.9 Vs 107.4 for 6900K. Oh and the games they tested (apart from BF1) are not well threaded and value the clock speeds and IPC over cores.

 

[cytg111]

There is a pattern in those gaming benchmarks, adjusting + for clockspeed and adjusting - for perf dont scale 1:1 with clock increase .. does Zen come in ~5% behind Broadwell at the same clock for singlethreaded performance?

 

[USER8000]

Power consumption is in line with AMDs test roughly at i7-6900k level. IPC is way below Skylake I would say. Because i5-6500 and i7-6900k run at a similar clock speed and Broadwell is ~12% above Skylake in the gaming tests, which shows that a thread/core advantage makes a decent difference in the gaming tests.

You have some interesting views but maybe you need to look at the numbers and people who apparently read the article.

If you look at the results an 8C/16T Core i7 6900K is tested.

The Ryzen chip was running at between 3.1GHZ to 3.2GHZ(if you look at the Reddit thread) and the Core i7 6900K was probably using quad channel RAM.

The Core i7 6900K runs at between 3.2GHZ to 3.7GHZ,and if we say its 3.5GHZ for the tests,it means Ryzen is very close to Broadwell IPC.

In the gaming tests the Core i7 6900K is 10% ahead,which again means Ryzen is close to Broadwell IPC.

 

[inf64]

As a general rule of thumb, if you OC your chip 10% you should see close to that much performance improvement, maybe a percentage point less.

 

[Vesku]

I can see why AMD paid for more foundry freedom from GF. Zen is looking very good given the R&D disparity between AMD and Intel. They need to fast track the mobile APU variant, imo.

 

[cytg111]

As a general rule of thumb, if you OC your chip 10% you should see close to that much performance improvement, maybe a percentage point less.

I thought that was div 2 .. overclocking your chip doesnt raise memory or the busses... But I wil admit its been some time, so Ill take your word for it.

Are we still in agreement that this is fabbed at glofo ?

 

[USER8000]

As a general rule of thumb, if you OC your chip 10% you should see close to that much performance improvement, maybe a percentage point less.

It looks quite competitive overall. The last time I remember AMD getting this close was with the Phenom II against the 45NM Intel Core2 chips,but Intel had the Core i7 920 above those. This time it looks like AMD is competitive with even Intel CPUs which are a £1000.

This has not happened for nearly a decade.

 

[cytg111]

I still wanna see it on a shelf at my local hardware pusher before I believe it .

 

[mikk]

Compensate for clock speed difference between Ryzen and 6900K (ST Turbo 3.7Ghz vs 3.4Ghz as shown in the charts) and voila you are almost at the Broadwell-E level again: 97.3x 3.7/3.4=105.9 Vs 107.4 for 6900K. Oh and the games they tested (apart from BF1) are not well threaded and value the clock speeds and IPC over cores.

Wishful thinking. I don't think the scaling would increase linearly with the clock speed because of other bottlenecks like GPU limitation or something else. This effect can be seen if you compare Intel CPUs or basically other reviews

i5-6400 2700-3300 Mhz
i5-6500 3500-3900 Mhz


Assuming that the i5-6400 run at 3000 Mhz and i5-6500 at 3700 Mhz gives a 23% clock advantage, in the gaming test the advantage is reduced to less than 10%.

 

[PPB]

I see people making some oddball arguments here.

No, 8c/16t doesnt need to ship with higher clocks than 3.4ghz base to be competitive just because it falls short in gaming scenarios to the likes of 6700k/7700k. 4c/8t needs to ship with higher clocks than 3.4ghz to do that. Because that is the sku that should be competing with it in the first place.

If AMD gets around 14nm LPP voltage curve and gets to ship the 4c/8t one at 4ghz, even at 95w tdp, they have a really good product there.

 

[inf64]

I thought that was div 2 .. overclocking your chip doesnt raise memory or the busses... But I wil admit its been some time, so Ill take your word for it.

Are we still in agreement that this is fabbed at glofo ?

Manually OCing via multiplier is the same thing manufacturer does, only they provide 100% guarantee and do it within the certain TDP spec.
That is why 7700K is ~6-10% faster than 6700K, it clocks higher out of the box and is architecturally almost the same chip(IPC is exactly the same, there are some slight differences but they do not affect the IPC).
So if AMD was to provide us 3.7Ghz Turbo boost on retail Ryzen then that chip would perform more or less 10% better in games than the ES in this preview.

It looks quite competitive overall. The last time I remember AMD getting this close was with the Phenom II against the 45NM Intel Core2 chips,but Intel had the Core i7 920 above those. This time it looks like AMD is competitive with even Intel CPUs which are a £1000.

This has not happened for nearly a decade.

Yeah, amazing given the R&D budget differences and process node differences. Well done I must say.

 

[USER8000]

I see people making some oddball arguments here.

No, 8c/16t doesnt need to ship with higher clocks than 3.4ghz base to be competitive just because it falls short in gaming scenarios to the likes of 6700k/7700k. 4c/8t needs to ship with higher clocks than 3.4ghz to do that. Because that is the sku that should be competing with it in the first place.

If AMD gets around 14nm LPP voltage curve and gets to ship the 4c/8t one at 4ghz, even at 95w tdp, they have a really good product there.

I love how some are spinning the results whilst ignoring the Core i7 6900K which is running quad channel memory and appears to have higher clockspeeds,on motherboards with production BIOSes.

Once AMD gets these released,I expect we will see Turbo working better,which should make it very close to the Core i7 6900K. Using their weird logic,they probably think the Core i7 6900K is a failure too.

Quad channel parts with hyperthreading,should be able to make use of the TDP headroom better,so they should be closer to 4GHZ,and this will be very competitive for gaming.

I have not had an AMD CPU for a very long time and this looks like one I will be seriously considering.

 

[PPB]

Suddenly my gift card'ed cash in amazon has a purpose this summer/winter (depending where you live). Now I just have to wait for the usual Hw.Fr to give me Vray results

 

[Glo.]

There is a little patter apparent. This sample has 3.15 GHz base and 3.5 GHz boost. AMD in their New Horizon presentation showed sample running on 3.45 GHz. So if the scaling stays, BASE 8C/16T CPU will be 3.45 GHz/3.8 GHz.

If this also falls in line with what was leaked already, that base SR7 chip will cost 350$ - AMD has killer platform.

 

[inf64]

There is a little patter apparent. This sample has 3.15 GHz base and 3.5 GHz boost. AMD in their New Horizon presentation showed sample running on 3.45 GHz. So if the scaling stays, BASE 8C/16T CPU will be 3.45 GHz/3.8 GHz.

If this also falls in line with what was leaked already, that base SR7 chip will cost 350$ - AMD has killer platform.

CanardPC's quote on the front page of the magazine tells it all:

"Preview: AMD (Ry)Zen, un retour fracassant". Un retour Fracassant= shattering/devastating return .

 

[PPB]

Dont see the 8c one shipping for less than 400 bucks honestly.

You have to factor some things:

- AM4 has already a platform value advantage over 2011-3 (So, even if Zen was to be shipped at the same price for same ammount of cores/clocks than BW-E, the total cost will be cheaper).
- AMD has to make room for their whole APU line that with Vega+Zen, is a lot more competitive now, so it cant and wont ship at current APU prices.

Only caveat is that CPU 4c/8t may be a little cheaper or same price as 4c/8t APU. Depending how marketing goes (if all products are multi unlocked, if CPU only will ship with higher clocks, etc).

 

[USER8000]

I would be surprised if it costs $300 but the Core i7 6900K is so expensive,AMD can still price it at $500 to $600 and still look better value. AMD also has a great stock cooler now,and the motherboard costs should be lower.

 

[PPB]

Yeah, my bet is 8c/16t in the landmark of 500 to 600 bucks.

I would love lower prices, but that is how it is from a bussiness perspective.

 

[sirmo]

That French leak looks really really good. If that's really the performance on those clocks, it could even have a higher IPC than Skylake. It's looking faster than the 3.6Ghz turbo i5 6500. Impressive stuff. The only question left is how high can it clock.

Also if this ends up being true, than AMD under-hyped and over delivered. Because that's quite a bit more than 40% IPC gain on Excavator.

 

[PPB]

So when the thread title change to "Summit Ridge (Zen) Benchmarks CHECK PAGE 153" ?

 

[bjt2]

And those ES did not even have the auto OC feature. Before new horizon event, i would not even imagine such a feature. They hid it very well and probabily is absent in most of the ES (in those given to journals for sure), so also in this. If this feature manage to give +200:+300MHz average, other than the promised base frequency, we can have even 4GHz mean frequency in most workloads...

 

[lolfail9001]

Wishful thinking. I don't think the scaling would increase linearly with the clock speed because of other bottlenecks like GPU limitation or something else. This effect can be seen if you compare Intel CPUs or basically other reviews

i5-6400 2700-3300 Mhz
i5-6500 3500-3900 Mhz


Assuming that the i5-6400 run at 3000 Mhz and i5-6500 at 3700 Mhz gives a 23% clock advantage, in the gaming test the advantage is reduced to less than 10%.

6400 runs at 3.3 to 3.1Ghz and 6500 at 3.6 to 3.3Ghz, come on now, that's the less than 10% difference you see.