Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[formulav8]

But 4.0 in this cheap mobo is pretty awesome ! And at least 2933 on the memory as well.

Yes sir, thanks for the details.

 

[formulav8]

From business perspective it's always hard to break in to a monopoly market. Intel is no exception, they have the resources and I'm pretty sure they have used their leverage whenever they have been able to do that.

They also have the lower morals to do shady things when they feel like it. Not much of a secret though.

 

[richierich1212]

http://www.mersenneforum.org/showthread.php?t=22141

New version of Prime95 released.
"Who might consider upgrading? Users that do trial factoring with prime95. Users that have had difficulty getting prime95 to assign affinities properly. AMD Ryzen users."

 

[krumme]

Frametimes bf1 MP

FPS:

Certainly there is a huge difference for frametimes that is not reflected in fps

 

[imported_jjj]

There does seem to be less variance between scenes with Ryzen and that's why it might appear smoother.

Reviewers that test 99% or 90%, eliminate the bottom 1-10% but maybe looking at the mid 80-90% frames would be a more accurate way to look at it. If you eliminate just dips, you favor spikes and allow them to drag average up.
Or they could look at consistent lows so dips that last for more than let's say 0.3s.

Still, reviewers do need to understand that the load profile changes from low res to normal res (or high FPS vs normal) and that they can't test for the future with current games.Their methodology is flawed to the bone.


Crysis 3 Ryzen (red) vs 7700k, the FPS is similar but the 7700k has a wider variance between scenes.

Source at time https://youtu.be/TId-OrXWuOE?t=338

 

[krumme]

There does seem to be less variance between scenes with Ryzen and that's why it might appear smoother.

Reviewers that test 99% or 90%, eliminate the bottom 1-10% but maybe looking at the mid 80-90% frames would be a more accurate way to look at it. If you eliminate just dips, you favor spikes and allow them to drag average up.
Or they could look at consistent lows so dips that last for more than let's say 0.3s.

Still, reviewers do need to understand that the load profile changes from low res to normal res and that they can't test for the future with current games.Their methodology is flawed to the bone.


Crysis 3 Ryzen (red) vs 7700k, the FPS is similar but the 7700k has a wider variance between scenes.

Source at time https://youtu.be/TId-OrXWuOE?t=338

I think that midrange test would be interesting. But in this case wouldnt a 5% min give more or less the same results?

In the greater picture we have to keep in mind that straight out the gate Zen have an aprox 12% freq deficit and 7% ipc deficit vs KBL. So unless your game is using more than at least 4 threads it will just be that slower.
BF1 issues 10 threads, and does use them better than other games we have at present. Thats the reason Computerbase shows 6900 beeing slightly faster than 6850.
It might be that zen end up beeing memory limited in 4 years and not eg physics limited for heavy gaming. Memory havnt been scaling with cpu perf the last years. Latency for the mem chip is basically the same.
Aside from that argument i cant tell how this situation and debate is any different from the discussion eg 5 years ago. Its the same discussion each time.
The difference is perhaps this time we have bf1 frametime and fps data that shows today the benefit of 8 cores.

 

[tamz_msc]

Frametimes bf1 MP

FPS:

Certainly there is a huge difference for frametimes that is not reflected in fps

Wow, the 7700K is dropping below 40fps; 1800X is above 60 all the time.

 

[imported_jjj]

I think that midrange test would be interesting. But in this case wouldnt a 5% min give more or less the same results?

.

There are differences between games and SKUs, in some games the dips help AVG ,in others the spikes help AVG but we have so few bits of relevant data.
I appears that in BF1 DX11(note that DX12 is slower and scales less well with cores) and in Crysis 3, the 7700k benefits from spikes so it's favored when looking at 99% or 90%. It's also favored in AVG by the higher variance.

Would be nice if some review sites would make the raw data available as opposed to just presenting it the way they deem the most appropriate.

 

[Agent-47]

MSVC is kind of the whole compiler world in gamedev, ain't it.

.

Haha. Y know nothing. Its Intel compiler

 

[rancherlee]

Hopefully BF1 is a sign of games to come, the way I look at it is Ryzen "should" be able to beat a 7700k in any game that effectively uses more than 6 threads. Ryzen running a game on 6 main cores should be on par with a 7700k running 4 main + 2 HT, in theory of course. On BF1 is looks like a core or HT is getting maxed out and causes a hiccup on the 7700k, the 1800x doesn't suffer the same fate.

 

[Dygaza]

Hopefully BF1 is a sign of games to come, the way I look at it is Ryzen "should" be able to beat a 7700k in any game that effectively uses more than 6 threads. Ryzen running a game on 6 main cores should be on par with a 7700k running 4 main + 2 HT, in theory of course. On BF1 is looks like a core or HT is getting maxed out and causes a hiccup on the 7700k, the 1800x doesn't suffer the same fate.

This ofc would mean that all those 6 threads should be heavy. Witcher 3 for example runs 16 threads, but most of them are rather light.

Spoiler

For example audio threads are always very light in games.

 

[guachi]

Frametimes bf1 MP

Certainly there is a huge difference for frame times that is not reflected in fps

I was reading some article where the author binned his FPS into 4 FPS lots and graphed it to show some odd behavior in Ryzen. I took those numbers, converted the FPS to frame times, and then figured out the .1%, 1%, 10%, 25%, 50%, 75%, 90%, 99% and 99.9% FPS. Probably overkill, but once you have the raw numbers in a spreadsheet it's trivial to figure out the rest.

Personally, I think average FPS should be done away with and replaced with median. And absolute minimum should be replaced with .1% and/or 1%. In addition, I'd like a frame time graph and a chart showing binned fps/frame times.

EDIT: I'll go so far as to say that a graph like the above actually makes me want to play BF1 because I know that when I eventually get my R7 I'll actually be able to run the game and not have to ever worry about slow downs. Just play and enjoy.

 

[AtenRa]

MSI releases new BIOS that support Haswell-E on AMD X370 mobos

https://www.msi.com/Motherboard/support/X370-GAMING-PRO-CARBON.html#down-bios

 

[IEC]

"- When Haswell-E CPU Raito over 37 and then system may become instable."

:facepalm;

 

[RiverRicer]

This may be an indication as to why the bios' aren't playing nice with Ryzen?

 

[lolfail9001]

Haha. Y know nothing. Its Intel compiler

Oh, do you? Source it, then.

 

[Conroe]

Someone has not gotten enough sleep. Too much work to do.

 

[hotstocks]

WOW, I am so glad Amazon couldn't fill my order. This is a complete mess. I have been building computers (for myself and customers) for 25 years. I am NOT here to FUD. But after reading 200 pages here and over 1000 pages at overclock.net Ryzen forums, there is only one conclusion which is WAIT BECAUSE WE ARE ALL ALPHA TESTERS.
I KNOW how to build a system and I don't cheap out on parts, but what is scary is that even if you buy all the best parts and use the same bios settings, you get 3/10 people who can't post, 3/10 who get bsods, bootloops, crashes, and 4/10 that kinda work. So if 10 people buy an Asus VI Hero, G.skill 3200 ddr4 c14 2x8, Ryzen cpu with good cooler, gtx 1080, and a good 1000-1300 w PSU (Corsair, Seasonic, Silverstone), how the F@CK is everyone getting different (and non stable) results with the same components and bios settings?
Bottom line is I was EXCITED to build a Ryzen system, but there are only two reasons to build a system:
1)You don't have a system and need a system NOW for work, school, business (or games)
2)Your current system is dead and you need a system NOW for work, school, business (or games)
In either scenario you can't buy or a build a Ryzen system because YOU NEED IT TO WORK NOW 100% STABLE. Not, YOU NEED IT TO WORK MAYBE STABLE IN 1-2 MONTHS.
Believe me, I have 8 computers in my home all built by me, and I am SCARED to build a Ryzen system right now. Why pull your hair out for months being an alpha tester (YEAH I CONSIDER MOBOS AND THEIR BIOSES ALPHA NOW, NOT EVEN BETA)? This was THEIR JOB (AMD and Mobo manufacturers) NOT OURS. So if you really need to build a computer today, then just go with Intel or another AMD cpu (if that weak sauce will work for you), just don't build a Ryzen system. If you really want a Ryzen system, then be like me and just wait and use your current system if it works. Best case scenario is in 1-2 months you can build a Ryzens system and everything will just work without voodoo bios settings that change every day and instability. Worst case Ryzen never fixes all these stability issues and Intel releases their 6 and 8 core faster and stable cpus at a slightly higher price than Ryzen, but they just work. Hey, I'm all for saving $150 on a processor, but not if it is going to give me countless hours of aggravation and then still may need to be returned or sold. And before anyone yells FANBOI, I repeat I have 3 AMD CPU systems and 5 Intel CPU systems in my house that I have built. This post is not about Intel vs. AMD, it is about releasing a product that CLEARLY is NOT READY to be released and almost everyone, including professional system builders (like me) are pissed off because pros don't rush out alpha sh!t when they know there are major issues, then say, "Well in May we will probably have memory issues solved.", LOL

 

[inf64]

hotstock

I'm yet to see ONE post from you that is not flame baiting or just pure trolling. I cannot decide if you are unable to understand new technology or you are just mad at the world so you are venting on this forum. It is a mystery.

PS Wall of text with zero paragraphs makes it even harder to read.

 

[hotstocks]

inf64, I am not trolling. And I certainly understand new technology and have overclocked and built systems both AMD and INTEL for 25 years. Yeah, I guess I am venting though, but that still does not change what I have posted is absolute fact. I'm well aware of past new platforms where it took a few minutes or at worst hours in the bios to get your new build 100% stable or memory working at advertised speeds. But this is a WHOLE different ball game, and if you can't see that for yourself, or from the 1000 posts of others with major problems and no stability, then I can't help you. It just wasn't ready to be released. There is no shame or harm in admitting that.

 

[inf64]

So you missed the whole BDW-E "platform maturing" that had lasted for MONTHS?

Ryzen is not without problems but the platform works and if you are not into tweaking the hell out of it, it's 100% stable. The instabilities come from desire of people to run higher specced RAM on boards that have not had enough time to get proper BIOS support (plus Ryzen itself only officially supports 2666Mhz tops, anything above is out of spec and courtesy of mobo makers and AMD).

 

[hotstocks]

Ok, I'll give you that. But with AMD's decision to basically make their 8 core cpu from two 4 core cpus slapped together with slow infinity fabric communication, well if you aren't going to run your memory over 2666mhz, you are seriously gimping the build. All mobo manufacturers are claiming 3200mhz support for a reason, and even some and AMD are claiming above 3600mhz support in May. Higher mem frequency is the only way to get infinity fabric/L3 cache latency lower. It would be FOOLISH to buy a 2666mhz memory kit now, when you will then be buying a 3600mhz or thereabouts kit in a month or two.

 

[IEC]

Running the CPUs at stock gives zero issues, even on the ASUS boards.

It's once you start pushing the limits combined with the immaturity of the BIOS that they decide to fry a microcontroller on board (unless you have fixed BIOS versions).

 

[inf64]

Here is how "gimped" design like Ryzen performs. Poor Ryzen users

https://www.overclockers.ua/cpu/amd-ryzen-7-1800x/2/

If AMD is going to keep giving us such "gimped" products in the future I'm ecstatic for the PC market!

PS Look at those poor gaming scores, even on 4+0 config of Ryzen. I cannot wait to get my hands on 160 bucks quad core with SMT, or even better the 12T part.

 

[Markfw]

Ok, I'll give you that. But with AMD's decision to basically make their 8 core cpu from two 4 core cpus slapped together with slow infinity fabric communication, well if you aren't going to run your memory over 2666mhz, you are seriously gimping the build. All mobo manufacturers are claiming 3200mhz support for a reason, and even some and AMD are claiming above 3600mhz support in May. Higher mem frequency is the only way to get infinity fabric/L3 cache latency lower. It would be FOOLISH to buy a 2666mhz memory kit now, when you will then be buying a 3600mhz or thereabouts kit in a month or two.

OK, so I did some reading FIRST, then picked reasonable parts that are supposed to work, and in 2 hours, I have a perfectly working and stable system with a 400 mhz overclock, and at the memory settings advertised. So if you have a problem getting things to work, maybe you should go do some more reading.

Edit, and Yes I have been building systems and tweaking since the mid 1980's.