Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[lolfail9001]

Anyways, if i can push 1700 to ~3.5Ghz all core within ~95W thermal output (so my itx rig will not melt), i am gettin it... Well, when itx boards are out, anyways.

 

[CentroX]

Oh

Not good hopefully 1700 gets to 4ghz.

This is what the guy said:

A few pages later he also mentions that Ryzen doesn't have any headroom for overclocking when under water or air cooled, so most likely nothing beyond stock clocks for the average user

http://www.neogaf.com/forum/showpost.php?p=230799701&postcount=1463

 

[unseenmorbidity]

The price difference between the 7 1700 and 7 1700X is $70, but the 7 1700X comes with a better heatsink/fan.

So if you have to spend another $40 for a better heatsink/fan, the 7 1700X might not seem so bad.

Better, but is it really going to end result of your rig? For some people, sure, but for people wanting to OC, probably not.

I got a Noctua D15 on jet for $70.

 

[lolfail9001]

A few pages later he also mentions that Ryzen doesn't have any headroom for overclocking when under water or air cooled, so most likely nothing beyond stock clocks for the average user

Do they at least clock to single core xfr boost on all cores? They should be, right?

Also, did AMD really manage to push out SKUs that are literally clocked to the very limit?

 

[CentroX]

Do they at least clock to single core xfr boost on all cores? They should be, right?

Also, did AMD really manage to push out SKUs that are literally clocked to the very limit?

well probably because they haven't mention OC at all. I am getting the 1800X so I atleast have a cpu with 4GHZ.

Very disappointing news though.

check that vcore:

 

[ultima_trev]

Asides from XFR allowing for an extra 100MHz over the peak turbo speed, was there any mention about turbo steppings, e.g. all core turbo, all core turbo+XFR. Thermals and power delivery willing, I'm hoping XFR will allow at least a 3.7 GHz all core boost on the 1700X.

 

[unseenmorbidity]

Do they at least clock to single core xfr boost on all cores? They should be, right?

Also, did AMD really manage to push out SKUs that are literally clocked to the very limit?

The stock coolers are mediocre and bad. The Wraith isn't even as good as a 212, so I don't see how this could be true.

 

[Magic Hate Ball]

AMD Ryzen 7 1800X did topple the world record of Cinebench R15, achieved with Intel Core i7-5960X but Intel Core i7-6950X gets 2862: http://hwbot.org/submission/3450072_8_pack_cinebench___r15_core_i7_6950x_2862_cb

Yeah that's true, but if you divide the 1800X's 2449 by 8 then multiply by 10 you get 3061... so the Intel part is under performing per core noticeably.

And that 6950X is at a slightly higher core clock, and still loses per thread. That's damned impressive.

 

[Mopetar]

This is what the guy said:

A few pages later he also mentions that Ryzen doesn't have any headroom for overclocking when under water or air cooled, so most likely nothing beyond stock clocks for the average user

http://www.neogaf.com/forum/showpost.php?p=230799701&postcount=1463

Don't know how much faith I'd put in that considering someone had it over 5 GHz, albeit on LN2. I don't think it will be a great OCer, at least not the earliest chips, but I suspect at least 4.2 GHz on water, possibly up to 4.5 GHz for lottery winners.

 

[lolfail9001]

The stock coolers are mediocre and bad. The Wraith isn't even as good as a 212, so I don't see how this could be true.

Wraith is as good as 212, basically.
Also, guy i am quoting quotes the guy who actually overclocked these, on water too. Not just stock cooler.

Yeah that's true, but if you divide the 1800X's 2449 by 8 then multiply by 10 you get 3061... so the Intel part is under performing per core noticeably.

Yes, yes, yes, Ryzen has better throughput per core in Cinebench. But there are no 10 core Ryzens so that's sort of irrelevant.

Don't know how much faith I'd put in that considering someone had it over 5 GHz, albeit on LN2.

5960X hits 6Ghz on 8/16 cores/threads. Broadwell-E (latest result) tops out at ~5.3Ghz. Both do that on lower voltage than Zen had in this run. Matter of fact, this is the voltage people pump into 7Ghz Kaby Lake runs.

 

[StinkyPinky]

It's possible the "X" is a factory overclocked unit like the 4790K and therefore little to no room in overclocking.

 

[Face2Face]

I have no plans to replace my 5820K in my main build, but I'd love a $200-ish 6/12 RYZEN for my server. I'm hoping they're also released soon.

 

[unseenmorbidity]

This is what the guy said:

A few pages later he also mentions that Ryzen doesn't have any headroom for overclocking when under water or air cooled, so most likely nothing beyond stock clocks for the average user

http://www.neogaf.com/forum/showpost.php?p=230799701&postcount=1463

Maybe he just meant the 1800x?! Surely there couldn't be that much of a difference in binning!

 

[lolfail9001]

Surely there couldn't be that much of a difference in binning!

Did you forget Polaris?

 

[GroundZero7]

well probably because they haven't mention OC at all. I am getting the 1800X so I atleast have a cpu with 4GHZ.

Very disappointing news though.

You have to look at it from a different point of view. Both consoles are 8 core, almost all new games are Vulkan/DX12/Multithreaded.

Almost all new games will scream on Ryzen chips due to MT. Almost all single threaded games will still perform very well. Ryzen is the most future proof architecture right now with at least 3 upgrade cycles ahead of it.

I doubt you will be missing anything by having Haswell IPC, double the cores, and insane multi threaded performance. There really isn't going to be an IPC deficit that will affect you. Especially if you have Freesync.

 

[Atari2600]

Hmmm.... I suppose its worth bearing in mind at this point that AMD launched BD on B2 stepping with at best 3.6/3.9 GHz clocks.

12 months later, on the same process and same power PD took that to 4.0/4.2 GHz clocks.

A further 12 months later, and ~ doubling power, PD increased clocks up to 5.0 GHz at a high voltage - on water.


So, I guess its not reasonable to expect any significant increase in the life of Zen, but a reasonable bump with Zen2 is not unreasonable.

 

[lolfail9001]

Wait a minute, guys, i think i just noticed something really weird.

Why the hell does Zen have almost 5 BILLION (4.8 to be exact) transistors?
For reference, it is almost the same transistor density as Polaris, and much much larger than Broadwell-E's.

 

[OrangeKhrush]

I just watched linus new vid and all i can hear is Journey, don't stop believing.

 

[unseenmorbidity]

Did you forget Polaris?

That was like a 5% difference.

 

[french toast]

This is what the guy said:

A few pages later he also mentions that Ryzen doesn't have any headroom for overclocking when under water or air cooled, so most likely nothing beyond stock clocks for the average user

http://www.neogaf.com/forum/showpost.php?p=230799701&postcount=1463

Oh well, you can't have your cake and eat it seemingly, still we have outrageous performance per £ at stock clocks so we should be grateful

 

[jpiniero]

Why the hell does Zen have almost 5 BILLION (4.8 to be exact) transistors?
For reference, it is almost the same transistor density as Polaris, and much much larger than Broadwell-E's.

Partial PCH possibly.

 

[french toast]

Partial PCH possibly.

A fair few transistors on sensors and duplicate internal testing logic, wouldn't explain it all though.

 

[lolfail9001]

That was like a 5% difference.

Not really, initially they varied from undervoltable samples to samples that leaked like all hell even on stock and throttled as hard as reference 290.

Partial PCH possibly.

Or that circuit doubling for these SenseMI schemes, didn't bjt mention something like that? Anyways, it is actually way more transistors than BDW-E 10-core with more cache, more memory channels and more PCI-E.

 

[PhonakV30]

Maybe SMT costs transistors more than Intel's Hyper-threading ?

 

[SAAA]

Wait a minute, guys, i think i just noticed something really weird.

Why the hell does Zen have almost 5 BILLION (4.8 to be exact) transistors?
For reference, it is almost the same transistor density as Polaris, and much much larger than Broadwell-E's.

Because that's the final nail in the coffin that mad me understand Intel and every other count different "transistors".
I had the suspect when comparing 44mm^2 of Zen quad cluster with 1.4B transistors vs Broadwell 82mm^2 at 1.3B (the first 14nm dual core die) already, but now I'm 100% sure: either one is counting individual "fins" as transistors and the other multiple finned transistors or numbers don't add up.
Indeed if you double Broadwell's count (assuming 2 fin each transistor) it comes at 2.6B, and densities almost match (but we are comparing a whole soc against the denser logic cores/caches).