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

[Tup3x]

You do realize that you can view the usage of each individual core (actually hardware thread) in task manager, right?

You do know how it looks like when you run an application that only uses one thread?

 

[looncraz]

Take a look at how many threads are running on an idle Windows system. It approaches 1,000 (or can even exceed it with certain configurations).

Some of those threads have been dedicated to making single-threaded applications run faster, but a good 20% of them will be triggering or handling interrupts and user and application events while any decently complex program is running.

Super old games whose game logic, audio code, and graphics code are in the same thread likely already run well enough on a Sandy Bridge CPU at stock clocks that they are meaningless for this discussion. They will run even better on Ryzen, and more cores will help offload other tasks which may be running in the background. Disable a few cores and play those games and you'll possibly see it first hand.

One such game, that I play to this day, is of Age of Empires II. It is horribly limited by processing power on a single thread, no doubt. I used to run it on my old Core 2 Duo E7200. Overclocking had no effect. I upgraded to a Phenom II X3 (2.8Ghz) and had a minor improvement. I overclocked to 3.2Ghz... little better.

I upgraded to an FX-8350, it ran just the same. I upgraded to an i5-2500k @ 5Ghz, it ran no better. I moved to an i7-2600k... still no better. I don't think Ryzen, no matter how fast it is, will help with it, either. It's constrained by its own internal design and will not scale.

Even the new 'HD' design has such a behavior. It has now split the game logic and graphics pipelines, which helps smoothness when things are going bad, but the game still becomes unplayable with too much action.

You just can't fix those types of games with more CPU performance. They are constrained by timings related to game logic or accessing and directing audio and graphics pipelines. Those timings are dictated by system buses, storage, and other factors that don't respond to IPC improvements (they have max ILP already) or particularly well to CPU frequency. They just suck and will suck until whatever is bottle-necking it is resolved... then there will be another bottleneck.

 

[bakyt115]

https://news.xfastest.com/amd/31741/amd-ryzen-7-1700x-benchmark/

first review of AMD RYZEN 7 1700X

Reddit discussion thread
https://www.reddit.com/r/Amd/comments/5vap9x/ryzen_7_1700x_benchmark/

oh sorry page deleted now.

 

[HurleyBird]

404

 

[bakyt115]

new link
http://webcache.googleusercontent.c...st.com/amd/31741/amd-ryzen-7-1700x-benchmark/

 

[blublub]

I doubt that a 3.4ghz can hit 1530 MT points in Cinebench, but I'd it's true it's a hell of a CPU.

Voltage seems a little low with 0.6v

 

[OrangeKhrush]

http://semiaccurate.com/forums/showpost.php?p=282977&postcount=7026

Why does this guy continue to be self gratifying, in essence the only thing he has right to date is the clock speed limit.

The Sandybridge ~ Ryzen thing he has on is getting annoying. ~ implies :

Like, similar, having characteristic off, in relation to ...

He posted his comparisons of architectures per CB 15 ST all at 3Ghz from Anandtech reviews.

2600K (Sandy) - 106
3770K (Ivy) - 110
Ryzen - 118.5
4770K/5960X (Haswell/Haswell E) - 120
4790K (devils canyon Haswell) - 122.5
5775C (Broadwell) - 126
7700K (Kabylake) - 128.6
6700K (Skylake) - 129

Ryzen ~ Haswell more like it

 

[krumme]

I doubt that a 3.4ghz can hit 1530 MT points in Cinebench, but I'd it's true it's a hell of a CPU.

Voltage seems a little low with 0.6v

Yeaa but it seems the 1800x will beat the 6900k here.
For firestrike and cb the 1800x will be fastest 8c cpu.
Damn impressive. Not much to say.
Seems very legic to me right?

Can we ask op to change content? Thank you.

 

[lolfail9001]

Looks like we got our XFR in action: the Physics CPU-Z. A whole... 100mhz over supposed boost clock! Btw, adjusting to clock differential, Physics test looks the worst of the bunch, even assuming that the reported 39 is irrelevant and it was running at 3.5Ghz. OTOH in Cinebench it looks fine and dandy. I'll wait for geekbenches, of course, llvm being my primary workload and all.

Voltage seems a little low with 0.6v

It still uses 1.78.1. Or power management is that good.

 

[csbin]

WOW！！！

 

[malitze]

I'd like to see some faster memory in my upcoming serving of leaks please

 

[csbin]

！！！！！！！

 

[inf64]

Looks like we got our XFR in action: the Physics CPU-Z. A whole... 100mhz over supposed boost clock! Btw, adjusting to clock differential, Physics test looks the worst of the bunch, even assuming that the reported 39 is irrelevant and it was running at 3.5Ghz. OTOH in Cinebench it looks fine and dandy. I'll wait for geekbenches, of course, llvm being my primary workload and all.

It still uses 1.78.1. Or power management is that good.

It looks good overall. Compared to 6900K the perf./$ is going to be massively better. Still intel platform overall offers a bit more features so it's a trade-off.

Hmm that Physics score when normalized for clock speed looks like 10% slower than previous leak? Memory frequency bound maybe?

Previously leaked by videocardz: ZD3406BAM88F4_38/34_Y scores 20249 @ 4Ghz
Now this leak shows 3.9Ghz scoring 17916. At 4Ghz it should get 10% less than what videocardz leaked.

 

[cytg111]

I'd like to see some faster memory in my upcoming serving of leaks please

Yea, mmm, next time should you so decide to leak something, could you do it with faster memory? Thanks!

 

[lolfail9001]

Hmm that Physics score when normalized for clock speed looks like 10% slower than previous leak? Memory frequency bound maybe?

More likely it picked up the speed during the finishing parts of it that always looked so slow for me, they had to single threaded on buildzoid's streams. I expect physics test itself was at 3.5Ghz... 10% lower clock instead of 10% slower normalized score.

 

[inf64]

More likely it picked up the speed during the finishing parts of it that always looked so slow for me, they had to single threaded on buildzoid's streams. I expect physics test itself was at 3.5Ghz... 10% lower clock instead of 10% slower normalized score.

Yep I think you are correct. If that was the case then it matches up perfectly with previous leak by videocardz.

Also in case XFR boosted the clock to 3.9Ghz(+100 over max turbo) then CB R15 @ 3.7Ghz should match perfectly with previous Chinese leak of 6C ES Ryzen @ 3.7Ghz : 146pts. 154x3.7/3.9=146.

 

[OrangeKhrush]

I doubt that a 3.4ghz can hit 1530 MT points in Cinebench, but I'd it's true it's a hell of a CPU.

Voltage seems a little low with 0.6v

We know the 1600X gets 1136 at 3.4 so

1136 * 8/6 * 3.4/3.4 = 1514 which is less than a percentage to that so it is very real

 

[lolfail9001]

Oh, btw, Khrush, calling you out on your "1T Command rate" BS. At least until i see a leak with it.

 

[coercitiv]

Yeah, dunno what to do with it - there are too much data flying around now and its hard to know what is true and what is a dud.

Let that thread wait for better quality info, not questionable leaks. Will eventually get revived by either some decent leak (like the one from CPC Hardware) or simply when reviews come up.

 

[lolfail9001]

simply when reviews come up.

That judging by upload names of the pictures should be pretty damn soon.

 

[PhonakV30]

https://videocardz.com/66182/amd-radeon-7-1700x-pictured-and-tested

My reaction :

http://imgur.com/gallery/rGo3GFu

 

[bjt2]

It looks good overall. Compared to 6900K the perf./$ is going to be massively better. Still intel platform overall offers a bit more features so it's a trade-off.

Hmm that Physics score when normalized for clock speed looks like 10% slower than previous leak? Memory frequency bound maybe?

Previously leaked by videocardz: ZD3406BAM88F4_38/34_Y scores 20249 @ 4Ghz
Now this leak shows 3.9Ghz scoring 17916. At 4Ghz it should get 10% less than what videocardz leaked.

The test is showing the results, so it's finished. I don't think that an up to 32 thread test has run at 3.9GHz... Maybe there is a light system task in background that not let the clock sitting at idle levels.

 

[CentroX]

Intel is done. Happy days indeed.

That cinebench score at 3.5ghz is insane. An OC 1800X is going to reach the 1800 dollar 6950x score.

 

[OrangeKhrush]

And I bought a 5960X a few months ago for over 900USD

 

[inf64]

I tested CPU mark 99 on my 4690K @ 4.4Ghz and got 758-760pts (with DDR3 @ 2400Mhz). CPU usage of the app itself varies between 32 and 35 % during the test , so I reckon that 5960x score of 561 has been done at around 3.3Ghz (math works out to around 3.25Ghz going back from my own score) so Ryzen has been running the test at MT turbo clock bin of 3.5Ghz (1 above stock, just like 5960x did). Ryzen at 4.4Ghz (matching my clock on Haswell) should be scoring 583 x 4.4/3.5=732, some 3.5% lower than Haswell. I would expect some faster memory to bridge those couple of percents of difference.