Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[Ajay]

I can almost guarantee it is closer to 99.9999% of companies. You do not overclock a production machine. Period. No if's, but's and maybe's. You just don't.

They used to use overclocked systems for Wall St. traders. I remember them wearing a suit coat, shit and tie with shorts!

 

[unseenmorbidity]

Dear Customer,

Your service request : SR #{ticketno:[---------------]} has been reviewed and updated.

Response and Service Request History:

Thank you for your email.

We can advise that 1.35 to 1.425 is practical vcore range for OCed CPUs with appropriate cooler.

Thank you for contacting AMD.



In order to update this service request, please respond, leaving the service request reference intact.

Best regards,

AMD Global Customer Care

 

[Elfear]

Dear Customer,

Your service request : SR #{ticketno:[---------------]} has been reviewed and updated.

Response and Service Request History:

Thank you for your email.

We can advise that 1.35 to 1.425 is practical vcore range for OCed CPUs with appropriate cooler.

Thank you for contacting AMD.



In order to update this service request, please respond, leaving the service request reference intact.

Best regards,

AMD Global Customer Care

Woot! That's a broader range than I expected from AMD as they'd tend to be more conservative.

 

[thepaleobiker]

I think this is the first review of Ryzen 7 (and simulated Ryzen 5s) using mid-range graphics....! Fascinating. Check it out.

http://www.techspot.com/review/1360-amd-ryzen-5-1600x-1500x-gaming/page4.html#commentsOffset

Regards,
Vishnu

 

[imported_jjj]

I think this is the first review of Ryzen 7 (and simulated Ryzen 5s) using mid-range graphics....! Fascinating. Check it out.

http://www.techspot.com/review/1360-amd-ryzen-5-1600x-1500x-gaming/page4.html#commentsOffset

Regards,
Vishnu

Techspot is trolling ,they simulate but they don't as they OC.
BF1 is tested with DX12 but DX12 results in much lower perf than DX11 in BF1 so they use DX12 anyway just cos Ryzen has problems with DX12 on Nvidia for some reason.
Ryzen does well in F1 2016 and should be on top but they clearly mess up the benchmark and Ryzen is seen as a quad core. We all know about F1 and they try to pass those results ...
Ghost Recon is added this time around because that's another game that can't count the number of cores properly for Ryzen.

The tech sites are somehow terrible nowadays.
Has anyone tried to validate AMD's claims about F1?
Has anyone tried to figure out how to force Ghost Recon to see the correct number of cores?
Has anyone tested the Dota 2 update?
God forbid they miss a marketing press release with nothing in it but creating original content with some value is a no go.

 

[unseenmorbidity]

IDK those seem like really good results to me.

 

[thepaleobiker]

Techspot is trolling ,they simulate but they don't as they OC.
BF1 is tested with DX12 but DX12 results in much lower perf than DX11 in BF1 so they use DX12 anyway just cos Ryzen has problems with DX12 on Nvidia for some reason.
Ryzen does well in F1 2016 and should be on top but they clearly mess up the benchmark and Ryzen is seen as a quad core. We all know about F1 and they try to pass those results ...
Ghost Recon is added this time around because that's another game that can't count the number of cores properly for Ryzen.

The tech sites are somehow terrible nowadays.
Has anyone tried to validate AMD's claims about F1?
Has anyone tried to figure out how to force Ghost Recon to see the correct number of cores?
Has anyone tested the Dota 2 update?
God forbid they miss a marketing press release with nothing in it but creating original content with some value is a no go.

All very valid points. Despite those issues, the performance of the 4C & 6C simulated Chips are still amazing. Isnt it? (or do you think these results are suspect)

 

[Puffnstuff]

You're extrapolating a large company decision based on what happens with an OC'd processor compared to a processor at default clocks. This is a fundamental error on your part.

OC performance is irrelevant, here. Companies aren't going to buy boxes of these things and individually OC them on every system (let alone a small production studio). That is ridiculous.

He under clocked the Intel and overclocked the Ryzen and its still well behind its Intel counterpart which is the whole point proving that you didn't even bother to look at the material.

 

[imported_jjj]

All very valid points. Despite those issues, the performance of the 4C & 6C simulated Chips are still amazing. Isnt it? (or do you think these results are suspect)

They test in 4 games and at least in 3 of those Ryzen 6 and 8 cores are penalized in some way. The 4 cores Ryzen might suffer in F1 and Ghost Recon if the games see it as 8 cores.

BF1 in DX11 vs DX12. Focus on Ryzen, 6900k and 7700k FPS and how poorly DX12 scales with cores (at 1080p Ryzen goes from 122 to 89 FPS)
https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps
https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx12-multiplayer-fps
and as a bonus https://www.computerbase.de/2017-03...-frametimes-ryzen-7-1800x-gegen-core-i7-7700k

Clean F1 test but likely not ideal as the game likely sees it as 16 cores not 8C/16T as it should https://www.computerbase.de/2017-03/amd-ryzen-1800x-1700x-1700-test/4/#diagramm-f1-2016-fps

As for Ghost Recon, it sees Ryzen as 16 cores but nobody has tried to figure out if there is a negative impact on perf and how substantial it is.

 

[coercitiv]

He under clocked the Intel and overclocked the Ryzen and its still well behind its Intel counterpart which is the whole point proving that you didn't even bother to look at the material.

The 5960X in that video was used at 4Ghz for the first test. The 5960X has a stock turbo speed of 3.5Ghz

Before accusing people of not watching the material you posted, you might want to do your homework on Intel HEDT CPUs. Alternatively you can stop posting FUD about Zen, but let's not get carried away in wishful thinking

 

[IEC]

He under clocked the Intel and overclocked the Ryzen and its still well behind its Intel counterpart which is the whole point proving that you didn't even bother to look at the material.

I watched the whole video. Jay talks about this in his video, and you are incorrect. His maximum stable overclock on the 5960X is 4.5GHz - but for the purposes of the initial comparison he ran it at 4.0GHz to do a "apples for apples" comparison (use of different graphics cards notwithstanding). His maximum stable OC on the Ryzen 1800X chip is 4.0Ghz. He later ran the 5960X at 4.5GHz, at which speed it completed a few minutes faster.

Once again, the vast majority of users outside of us die-hard overclockers will run at stock.

Ryzen 1800X is 3.6GHz base clock, 4.1GHz XFR (up to 2-cores), and 3.7GHz all-core turbo.
Intel i7-5960X is 3.0GHz base clock, 3.5 GHz turbo (1 core), and 3.3GHz all-core turbo.

At stock clocks, the 1800X will beat the 5960X while coming in at half the price.

That's why video editing geeks are excited by Ryzen.

 

[piesquared]

I would have thought something that reviewers would be interesting in testing is a Ryzen+RX480 compared to Ryzen+GTX1060.

 

[IEC]

I would have thought something that reviewers would be interesting in testing is a Ryzen+RX480 compared to Ryzen+GTX1060.

I suspect we'll see more of those when Ryzen 5 launches. That's aimed at the mainstream gamer.

 

[sushukka]

I can almost guarantee it is closer to 99.9999% of companies. You do not overclock a production machine. Period. No if's, but's and maybe's. You just don't.

Exactly. Three words: stability, warranty, support.

 

[Rifter]

He under clocked the Intel and overclocked the Ryzen and its still well behind its Intel counterpart which is the whole point proving that you didn't even bother to look at the material.

What are you smoking, the intel system was overclocked....

I wasnt sure you were trolling originally but its becoming more obvious as time goes on with posts like this.

 

[Shivansps]

They test in 4 games and at least in 3 of those Ryzen 6 and 8 cores are penalized in some way. The 4 cores Ryzen might suffer in F1 and Ghost Recon if the games see it as 8 cores.

BF1 in DX11 vs DX12. Focus on Ryzen, 6900k and 7700k FPS and how poorly DX12 scales with cores (at 1080p Ryzen goes from 122 to 89 FPS)
https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps
https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx12-multiplayer-fps
and as a bonus https://www.computerbase.de/2017-03...-frametimes-ryzen-7-1800x-gegen-core-i7-7700k

Clean F1 test but likely not ideal as the game likely sees it as 16 cores not 8C/16T as it should https://www.computerbase.de/2017-03/amd-ryzen-1800x-1700x-1700-test/4/#diagramm-f1-2016-fps

As for Ghost Recon, it sees Ryzen as 16 cores but nobody has tried to figure out if there is a negative impact on perf and how substantial it is.

It scales poorly because it is already a problem having game threads and driver API threads landing on different CCXs, DX12 and Vulkan starts more threads than DX11 and OpenGL drivers, and each of those have a chance of running on a diferent CCX.

Dont get me wrong, in the future Ryzen gonna pull ahead of quad cores, but even a 6800K looks to me more future proof than 8C Ryzens, and i dont think they have a chance vs future true 6C.

 

[IEC]

It scales poorly because it is already a problem having game threads and driver API threads landing on different CCXs, DX12 and Vulkan starts more threads than DX11 and OpenGL drivers, and each of those have a chance of running on a diferent CCX.

Dont get me wrong, in the future Ryzen gonna pull ahead of quad cores, but even a 6800K looks to me more future proof than 8C Ryzens, and i dont think they have a chance vs future true 6C.

Intel HEDT chips use a ring bus... Is that a "true" 10 core? AMD uses Infinity Fabric to connect its CCXes. Different solutions, with different pros/cons.

i7-6950X die shot:

AMD's memory controller isn't anywhere near its limits. Once they update microcode and BIOS to support the higher speed (3600+) DDR4 I expect you'll see the cross-CCX communication penalty become much less noticeable.

 

[imported_jjj]

It scales poorly because it is already a problem having game threads and driver API threads landing on different CCXs, DX12 and Vulkan starts more threads than DX11 and OpenGL drivers, and each of those have a chance of running on a diferent CCX.

Dont get me wrong, in the future Ryzen gonna pull ahead of quad cores, but even a 6800K looks to me more future proof than 8C Ryzens, and i dont think they have a chance vs future true 6C.

You didn't actually looked at DX11 vs DX12 in BF1. DX12 scales poorly and it's not Ryzen specific, look at the 6900k. You don't want to use DX12 in BF1 if you have more than 4 cores.
Ryzen has an additional problem on top of that but under DX11 it is still way better than the 7700k with both higher FPS and less variance in BF1.
If the DX12 "problems"are related to Nvidia, hard to say since nobody is looking at it.
The CCX theory is just a theory until there is solid evidence and the impact can be measured. As of right now it is unwise to take it as fact.

 

[Shivansps]

Intel HEDT chips use a ring bus... Is that a "true" 10 core? AMD uses Infinity Fabric to connect its CCXes. Different solutions, with different pros/cons.

i7-6950X die shot:

AMD's memory controller isn't anywhere near its limits. Once they update microcode and BIOS to support the higher speed (3600+) DDR4 I expect you'll see the cross-CCX communication penalty become much less noticeable.

Is more "true" than AMD design for sure. Thats more similar to Intel Xeon HCC or MCC

and more fabric speed with faster ram does not mean less latency.

 

[Shivansps]

You didn't actually looked at DX11 vs DX12 in BF1. DX12 scales poorly and it's not Ryzen specific, look at the 6900k. You don't want to use DX12 in BF1 if you have more than 4 cores.
Ryzen has an additional problem on top of that but under DX11 it is still way better than the 7700k with both higher FPS and less variance in BF1.
If the DX12 "problems"are related to Nvidia, hard to say since nobody is looking at it.
The CCX theory is just a theory until there is solid evidence and the impact can be measured. As of right now it is unwise to take it as fact.

The L3 latency @ >6MB is not solid evidence? Yes, Ryzen gona work best with ST graphics API, under Linux for example, Dota 2, works excelent with OpenGL(Pure ST API), but is crap with Vulkan (MT API).
http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-cores&num=2
I cant think of another explanation. Vulkan works best in a X+0 config.

 

[imported_jjj]

and more fabric speed with faster ram does not mean less latency.

You are in over your head.
Ryzen with the DRAM at 3200MHz CL14 can get as low as 70ns memory latency right now when the secondary memory settings are locked. Push those a bit and it gets to Broadwell-E levels.
For CCX to CCX ,it goes pretty much the same way.

The L3 latency @ >6MB is not solid evidence? Yes, Ryzen gona work best with ST graphics API, under Linux for example, Dota 2, works excelent with OpenGL(Pure ST API), but is crap with Vulkan (MT API).
http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-cores&num=2
I cant think of another explanation.

Yeah i am done with you (as in ignore list for a week so don't bother), you pivot to L3 since you are running out of things to say on the topic you brought up.

 

[Shivansps]

You are in over your head.
Ryzen with the DRAM at 3200MHz CL14 can get as low as 70ns memory latency right now when the secondary memory settings are locked. Push those a bit and it gets to Broadwell-E levels.
For CCX to CCX ,it goes pretty much the same way.




Yeah i am done with you (as in ignore list for a week so don't bother), you pivot to L3 since you are running out of things to say on the topic you brought up.

Are you telling me the CCX and L3 Latency are not related? i think im the one im done with you. You are the one that come up with things that has nothing to do with anything as an excuse. Here is a tip, I was never talking about main memory latency.

BTW, you completely ignored the fact that Dota 2 @ Vulkan works best on X+0 configuration because it kills your argument.

 

[IEC]

Are you telling me the CCX and L3 Latency are not related? i think im the one im done with you. You are the one that come up with things that has nothing to do with anything as an excuse. Here is a tip, I was never talking about main memory latency.

BTW, you completely ignored the fact that Dota 2 @ Vulkan works best on X+0 configuration because it kills your argument.

... except the Dota 2 update two days ago changed that. It now works properly without having to set X+0 configuration. And gained about 25% performance...

 

[raghu78]

AMD's design with CCX, Infinity fabric has lots of potential. AMD needs to improve the memory support of higher speeds like DDR4 3200 and even DDR4 4000 as the fabric is tied to memory controller speed. In future generations AMD could work to improve fabric speeds while keeping power consumption under control . When AMD moves to GF 7m I think we could see a 8 core CCX as 7nm brings huge transistor density increase and power efficiency gains over 14nm. With such a different design and certain tradeoffs there is lots of potential to improve performance in future generations. I am curious to see the architecture evolve over the next 2-3 years.

 

[Magic Hate Ball]

AMD's design with CCX, Infinity fabric has lots of potential. AMD needs to improve the memory support of higher speeds like DDR4 3200 and even DDR4 4000 as the fabric is tied to memory controller speed. In future generations AMD could work to improve fabric speeds while keeping power consumption under control . When AMD moves to GF 7m I think we could see a 8 core CCX as 7nm brings huge transistor density increase and power efficiency gains over 14nm. With such a different design and certain tradeoffs there is lots of potential to improve performance in future generations. I am curious to see the architecture evolve over the next 2-3 years.

Yeah, I'm hoping my AM4 motherboard today will work sufficiently with a Zen++ so I can upgrade at that point to a 7nm chip (hopefully) without exchanging my entire rig.

I have to say, for a first gen, total re-arch, this is working out really well. I remember reading about various growing pains on previous Intel and AMD systems over the last two decades, and for a "rushed" feeling launch this is god damned solid material.