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

[coercitiv]

Can someone with i7 or i3 run the test with HT disabled?

The following SS contains 2 data sets, both from a Haswell i7 @ 3.4Ghz with DDR3 1600 CL11. The first set contains default values, the second set was run with the benchmark program set with restricted affinity on logical cores 0-2-4-6.

Anybody who can actually disable HT is welcome to come with more appropriate results.

 

[Agent-47]

i5 480M 2c4t... my old laptop. Prime number scores

2c4t => 8
1c1t => 4
2c2t => 8

HT disabled from BIOS. but maybe someone with a newer CPU should run it.

 

[Face2Face]

i7 5820K @ 4.25GHz - No Turbo
DDR4 @ 3GHz Quad Channel

 

[Agent-47]

i7 5820K @ 4.25GHz
DDR4 @ 3GHz

are you on quad channel?

Can you please move to a dual channel config and check any changes in result please

 

[Gideon]

i7 5820K @ 4.25GHz
DDR4 @ 3GHz

Looking at these, it becomes more and more obvious, how insanely good the Integer and Floating Point Math scores of Zen are @3.4/3.8

Even If we normalize them to per Ghz and core count,Ryzen is still:
Zen: (39672 / 3.4 / 8 = 1458.53; 14807 / 3.4 / 8 = 544.37)
HW: (30931 / 4.25 / 6 = 1212.98; 12787 / 4.25 / 6 = 501.45)
~= 16.8% faster in Integer performance per core vs Haswell
~= 7.9% faster in Floating Point performance vs Haswell

in Highly Multithreaded scenarios.

Yet it still loses out in Single Threaded performance:
Zen: 2046/3.8 ~= 538.42 (turbo working, as per the video)
HW: 2542/4.25 ~= 598.12 (no turbo presume?)

So it seems it's per-core scaling, at least in some applications, is better than any Intel product. I wonder what causes that?

Disclaimer: I know, it's only a single (and not that good to begin with) benchmark, but still
... CHOO CHOO!

 

[Face2Face]

i7 5820K @ 4.25GHz - No Turbo
DDR4 @ 3GHz Dual Channel

 

[Magic Hate Ball]

i7 5820K @ 4.25GHz - No boost
DDR4 @ 3GHz Dual Channel

Based on earlier posts, I would be curious what it looks like with HT disabled.

 

[OrangeKhrush]

I am afraid not. If it was quad channel 7700k would have at least half the score of ryzen. Also is evident by looking at the normalized score I posted which shows 7700k has the same IPC as the quad channel intel

IOC and IMC are not the same AIDA64 shows my 5960X max memory performance twice that of a SKL/KBL. Max bandwidth is like 80k vs 35k. Clockspeed helps memory benchmarks as well as timings and latency ram clocks etc.

Memory benches are super synthetic, may as well call them McDonald's

 

[Face2Face]

Based on earlier posts, I would be curious what it looks like with HT disabled.

Sorry, I put my two stick of memory back in, so this run is with Quad Channel memory.

i7 5820K @ 4.25GHz - No Turbo + HT Disabled
DDR4 @ 3GHz Quad Channel

 

[lolfail9001]

Wait a minute...

 

[Magic Hate Ball]

Sorry, I put my two stick of memory back in, so this run is with Quad Channel memory.

i7 5820K @ 4.25GHz - No Turbo + HT Disabled
DDR4 @ 3GHz Quad Channel

Wow, so HT disabled gets you 80 primes vs 57 with it on?

 

[OrangeKhrush]

Looking at these, it becomes more and more obvious, how insanely good the Integer and Floating Point Math scores of Zen are @3.4/3.8

Even If we normalize them to per Ghz and core count,Ryzen is still:
Zen: (39672 / 3.4 / 8 = 1458.53; 14807 / 3.4 / 8 = 544.37)
HW: (30931 / 4.25 / 6 = 1212.98; 12787 / 4.25 / 6 = 501.45)
~= 16.8% faster in Integer performance per core vs Haswell
~= 7.9% faster in Floating Point performance vs Haswell

in Highly Multithreaded scenarios.

Yet it still loses out in Single Threaded performance:
Zen: 2046/3.8 ~= 538.42 (turbo working, as per the video)
HW: 2542/4.25 ~= 598.12 (no turbo presume?)

So it seems it's per-core scaling, at least in some applications, is better than any Intel product. I wonder what causes that?

Disclaimer: I know, it's only a single (and not that good to begin with) benchmark, but still
... CHOO CHOO!

The pass mark leak was a flat 3.4ghz a 1800 between 3.6 to 4 would score 2150-2400 odd, thats sick

 

[Face2Face]

Wow, so HT disabled gets you 80 primes vs 57 with it on?

It appears so... Odd. Let me run it again.

EDIT:

Ran the test again and got the same result, slightly higher actually.

 

[lolfail9001]

That's cache related.

 

[OrangeKhrush]

It is not out the question, HT has latency and some people prefer gaming with it off. I had a i3 2100 at one point,. Terrible for gaming when the two physical cores were saturated

 

[Greyguy1948]

HT on or off should be important in some benchmarks even if this effect is not like in Pentium 4.
Most games we have today should run better with HT off.

 

[malitze]

It appears so... Odd. Let me run it again.

EDIT:

Ran the test again and got the same result, slightly higher actually.

A reason for this might be that since Hyperthreading helps with utilization of functional units but the workload is identical for each thread, it should be harder to parallelize when the needed units aren't available. Just a guess though

 

[thigobr]

Yes, SMT (HT) will only help if the machine is not at peak utilization so splitting resources won't hurt the other thread too much.

 

[Agent-47]

That's cache related.

you mean less thread competing for the same L3? I think I can agree with that.

6900k and 6800k does have more cache per thread compared to ryzen.

 

[dfk7677]

Both Prime and Physics have better results with SMT off and bad results for Ryzen

 

[Greyguy1948]

Isn't this strange?
All memory info is regarding core and not thread.
https://www.cpubenchmark.net/cpu_test_info.html
Most programs have no info about real cores but only logical cores.

 

[Dresdenboy]

Interesting results here! The HT effect might be related to cache thrashing, partitioned core resources, fighting for sparse resources (which incl. cache and mem channels of course). Just fighting for an IDIV unit alone (if used at all) wouldn't cause such a significant drop.

Someone wondered about the string sorting. That test's lower mem dependency could come from cache blocking or easier to detect access patterns for prefetching, or both.

OTOH Passmark has a history of fiddling with IDIV. It's past CPU benchmark result measured the runtimes of loops with special instructions, one being IDIV. But it ran the same amount of instructions (IIRC) and created the score from the whole runtime. Now imagine 1 IDIV loop contributing maybe 80% of a benchmarks runtime if there is no HW divider. This way the per patch deactivated Llano HW divider got some attention.

http://www.passmark.com/forum/performancetest/3705-amd-llano-a-series-benchmark-and-cpu-bug

 

[CatMerc]

This could tell us that Ryzen's SMT doesn't do as good a job at managing the resources between the two threads.
Makes sense, Intel's years and years of work on SMT basically amounted to minimizing the performance regressions in certain cases.

 

[Shivansps]

Hang on, you dont need a Windows update to handle the AMD SMT on Windows 10? Windows may think those are real cores.

 

[Agent-47]

I got tired of scrolling up and down so I took the liberty to make a table off all the results.