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

[Greyguy1948]

If you look at Vishera results, with 8 threads, you will find similar scores when overclocked. As this is a test that seems to scale quite well with the number of threads/physical cores, it seems quite unlikely that a CPU with double the count of FPU units & threads (not to mention a lot more schedule/execution/retire resources per core) will score the same as an overclocked old generation part. So something is quite off in that test, I don't know if it's fake or simply we are looking to a issue with the test itself.

Yes the total score is close. But if you look at the details:
Bench Zen Vishera
Int 458 407
Long 137 102
Quad 2.7 8.1
Single 348 251
Double 201 139
Quad fp 9.5 4.6

So how are they doing total score?
I have tested geomean, harmean and average but none of them is used.....

 

[Sweepr]

Ryzen Lineup (Rumor)

http://www.coolaler.com/threads/amd-r7、r5、r3-型號出爐-有17款之多.342093/

 

[Magic Hate Ball]

Ryzen Lineup (Rumor)

http://www.coolaler.com/threads/amd-r7、r5、r3-型號出爐-有17款之多.342093/

I say BS simply because Lisa Su herself had a giant slide on the screen that said 3.4ghz+ for "base clocks" specifically already during the Horizon event.

 

[inf64]

That is way too many SKUs IMO. Also positioning 12T Ryzen that has 300Mhz base range differential against 4T i5 KL range, is that a joke?

 

[lolfail9001]

I say BS simply because Lisa Su herself had a giant slide on the screen that said 3.4ghz+ for "base clocks" specifically already during the Horizon event.

We never established whether that meant: "For literally every Ryzen CPU" or just "We will totally have a 3.4Ghz or higher base clock SKU".

Otherwise, i call BS because that naming brings up associations with dumpster tier GPUs.

That is way too many SKUs IMO. Also positioning 12T Ryzen that has 300Mhz base range differential against 4T i5 KL range, is that a joke?

Well, if we assume that all these "PRO" marked SKUs are basically server/workstation chips ala Opterons, then SKU count is not really outrageous. Basically 2 clock bins per tier, looks about right to me.

 

[lolfail9001]

Fake, as pointed out at the HardOCP forums the CPU identifier is all wrong compared to the ones used at CES demos. Most importantly the end of the string has a "Y" to indicate an iGPU, which is completely impossible for a Summit Ridge part.

I was one of those to point that out and i am in doubt myself now because of following:
https://videocardz.com/65654/amd-ryzen-6-core-cpu-exists

Looks like sample with this name does exist, after all.

And god damn, i really do not want this screen to be true.

 

[Atari2600]

I thought naming convention was already known as SR3/SR5/SR7?

 

[Glo.]

That is way too many SKUs IMO. Also positioning 12T Ryzen that has 300Mhz base range differential against 4T i5 KL range, is that a joke?

Pricing range, first. Performance follows.

If this is true, and everything what we know about Ryzen is correct, the pricing is disrupting the market pretty badly for Intel. Because, according to this, for 230$, you will get 6C/12T CPU.

The only way Intel can market their Kaby Lake SKUs is by MASSIVELY lowering the prices.

Also SR7 1700X can be the aforementioned 3.45//3.7 GHz CPU from the New Horizon demo. For 400$...

 

[jpiniero]

If this is true, and everything what we know about Ryzen is correct, the pricing is disrupting the market pretty badly for Intel. Because, according to this, for 230$, you will get 6C/12T CPU.

Translation says the Intel comparison is performance (in what though?), not price. The X models are the unlocked ones it suggests.

 

[Burpo]

i really do not want this screen to be true.

 

[Doom2pro]

Ryzen Lineup (Rumor)

http://www.coolaler.com/threads/amd-r7、r5、r3-型號出爐-有17款之多.342093/

If there really are that many SKUs the reviewers are going to be busy for quite a while... O_O

 

[Glo.]

From SA forum:

Translation says the Intel comparison is performance (in what though?), not price. The X models are the unlocked ones it suggests.

If the translation says that 1800X is competitive with 6900K, how come 8 core suddenly can be competitive with 6 core CPU only, or 4 core, for that matter?

 

[inf64]

That chart is from 3dcenter website and they just used CPC HW preview numbers and extrapolated from there. Nothing new, we have been doing that on this forum since CPC leak came out.

 

[PhonakV30]

Nahhhh.

It will be 4.5GHz 95W, consuming 70W actually, with 5GHz OC and BDe level IPC at 60-70% it's price.

Intel is doomed.

Sent from HTC 10
(Opinions are own)

Yep as you can see here!

 

[Vesku]

Translation says the Intel comparison is performance (in what though?), not price. The X models are the unlocked ones it suggests.

AMD has publicly stated all Zen are unlocked. OC governed by chipset.

If there are "Pro" SKUs hopefully that is ECC capability for low mark up.

 

[jpiniero]

AMD has publicly stated all Zen are unlocked. OC governed by chipset.

There's no guarantee that all Summit Ridge models will be branded as Ryzen. The other possibility that you only get XHR (or whatever it's called) on non-X models.

If the translation says that 1800X is competitive with 6900K, how come 8 core suddenly can be competitive with 6 core CPU only, or 4 core, for that matter?

Clock speed plus IPC difference.

 

[Glo.]

Maybe it is not exactly fake, but simply: WRONG chart?

 

[jpiniero]

Maybe it is not exactly fake, but simply: WRONG chart?

Yeah I wouldn't put too much stock in it.

 

[bjt2]

We never established whether that meant: "For literally every Ryzen CPU" or just "We will totally have a 3.4Ghz or higher base clock SKU".

Otherwise, i call BS because that naming brings up associations with dumpster tier GPUs.

Well, if we assume that all these "PRO" marked SKUs are basically server/workstation chips ala Opterons, then SKU count is not really outrageous. Basically 2 clock bins per tier, looks about right to me.

Maybe PRO are with full cache and non PRO are with half cache. I would guess that yields are not great now and that CPU with damaged cache would not be thrown away.
Also I agree that there can be sku with 3 up to 3.6GHz of base clocks. But i hope that these are early rumors and final clocks will be a little bit higher...
EDIT: i didn't see that there are three designations: maybe PRO without L3, plain with half L3 and X with full L3.

 

[lolfail9001]

Maybe PRO are with full cache and non PRO are with half cache. I would guess that yields are not great now and that CPU with damaged cache would not be thrown away.

Hm, that is interesting proposal, but i think it works better with those "X" SKUs, because otherwise this letter is utterly meaningless.

 

[bjt2]

Translation says the Intel comparison is performance (in what though?), not price. The X models are the unlocked ones it suggests.

They should be all unlocked, unless you mean XFR full unlocked (in TDP)

 

[bjt2]

Hm, that is interesting proposal, but i think it works better with those "X" SKUs, because otherwise this letter is utterly meaningless.

I edited. The cache can be disabled, half enabled and full enabled..

 

[PhonakV30]

Here Full chart for Kaby Lake , Skylake
https://www.bit-tech.net/hardware/2017/01/03/intel-core-i7-7700k-i5-7600k-review/1

according to coolaler , 6C/12T at 3.2/3.6 ~ 4C/4T at 3.8/4.2 , Fair or Unfair ?

 

[zinfamous]

That is way too many SKUs IMO. Also positioning 12T Ryzen that has 300Mhz base range differential against 4T i5 KL range, is that a joke?

yes, that's a joke. But tbh, no one really knows the actual performance.

Still, the fact that the base clocks on this chart aren't in line with what has been announced and actually tested, requires the consumption of much salt in order to digest this image.

 

[Abwx]

If you look at Intel 16 T for newer systems you are in range 700-800.

When overclocked to the roof, a 3.5GHz 6950X is at 501, dunno where your 700-800 come from, besides Intel s scores are highly inflated by AVX2 and FMA..

If you look at Vishera results, with 8 threads, you will find similar scores when overclocked.

FTR it score about 180 at 4.2GHz, so to get to 318 it would need to be clocked at 7.42GHz, i guess that like the previous member you are confusing scores made with different versions of Sandra, the score of Zen was made with the 2015 version.