Ryzen: Strictly technical

Page 30 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

Rifter

Lifer
Oct 9, 1999
11,522
751
126
I have a source inside the PC testing circles, he had access to Ryzen from the first samples to QS samples and he posted two interesting things about Ryzen.






basically you are actually BETA testing.

So you would recommend waiting for next stepping? it seems you insider thinks there will be fixed silicon out before zen+, as he said few months, and zen+ is a year+ away.
 

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
for stability it can, for the ability to run higher clocks stable it can, for the ability to ru higher frequency through fabric it can. by large difference, will it boost say Tombraider from 112FPS to 140FPS, probably not but it will increase that number enough to be noticeable.
It is still hard locked to memory clock, bear that in mind.

I mean, once they add some control over memory, that may/will become relevant. But now now.
 

Wall Street

Senior member
Mar 28, 2012
691
44
91
lynfield came first, on 1156 which went up to clarksdale only Celerons, i5's and i3's, the socket LGA1366 came out for only i7 Gulftown and Nahelem. 1156 was replaced by 1155 and 1366 replaced by 2011

That isn't right. Socket 1366 arrived before 1156.
 

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
What does it even matter how much better the yields get or waiting for a next stepping this and that when the darn software doesnt even support the hardware its running on!! Thats the point. :mad:Its like a car company saying oh look you can upgrade your gas hog rust bucket of a beater for this here shiny new car for less then your car is worth and it runs great on this here new fuel called jelly 205.but after you buy the new car there is no place that even carries jelly 205.. Give it a few months lots of gas stations will carry jelly 205. really?? when everyone is like what the hell is jelly 205?? is this a fair comparison or am I just way out in left field here?
 

bjt2

Senior member
Sep 11, 2016
784
180
86
They still have some interesting options for Zen2, which in part didn't make it into Zen1 due to complexity and available time and other resources.

As I wrote before, they might use Zen1 as a more general purpose core to be used from top to bottom, servers to mobile. But Zen2 might be added to the portfolio as a more specialized core, improving on the remaining weaknesses of Zen1. For example a likely K12 related AMD patent (covering an AArch64 CPU) showed a third AGU. Schedulers, FPRF read ports for FMA, renamer, buffer sizes, SMT partitioning, etc. could still be improved on. Mind you, that Ryzen is where it is with all those trade offs.

Integer Zen performance isn't bad. It's FP pefrormance that can be improved. For instance with 4 symmetric 128 bit FMACs that can do also FMUL, FADD and ivec. Maybe specialization only on FDIV, SQRT and some IVEC instructions, but better if they are fully symmetric to fully exploit the advantage of 4 separate FP pipes...

EDIT maybe have also 6 full AGLU instead of 4+2 (Ryzen can perform 4 LEAs yet), maybe more branch, idiv and imul pipelines, even if IDIV is faster than SKL/KBL...
 

dnavas

Senior member
Feb 25, 2017
355
190
116
I have a source inside the PC testing circles, he had access to Ryzen from the first samples to QS samples and he posted two interesting things about Ryzen.

While on the one hand that's frustrating, on the other hand, this is almost exactly why I bought this CPU rather than one of Intel's. I can upgrade this one. More than once. I am not going to have the socket I just bought abandoned two weeks after I bought it (not that I'm bitter or anything). What wasn't clear beforehand was whether the upgrades were going to give good performance boosts. Now we have some evidence that at least one of them will ;> I assume this is why the 4 cores are coming out later....
That does make it easier to accelerate returning the 16GB sticks, as it seems clearer to me that four of them are not likely to work anytime soon.
 

Magic Hate Ball

Senior member
Feb 2, 2017
290
250
96
Not sure what the later silicon will have different but one thing I know on final bios they will return memory subtimings, they were on final QS platforms then AMD removed them.
Haven't had a chance, but my ASRock Taichi has a beta BIOS available for memory timing control.
 
  • Like
Reactions: looncraz

looncraz

Senior member
Sep 12, 2011
722
1,651
136
For Ryzen owners.
Please test Ryzen with WinRAR benchmark in single-channel mode and dual-channel mode.
We need WinRAR benchmark results with both modes: Multithreading option off / on.

With dual-channel mode:
Test 1 - WinRAR
Test 2 - WinRAR with Affinity to cores 0-7 (CCX0)
Test 3 - WinRAR with Affinity to cores 8-15 (CCX1)

Then run benchmark with single-channel mode.
You must remove one RAM module. So system will use only one channel.
Test 4 - single-channel - WinRAR
Test 5 - single-channel - WinRAR with Affinity to cores 0-7 (CCX0)
Test 6 - single-channel - WinRAR with Affinity to cores 8-15 (CCX1)

If results of single-channel tests are better, then Ryzen uses separated memory controllers with one memory controller per one CCX.

I am just going to test latency and bandwidth in single-vs-dual channel mode. I want to know the latency cost between the two PHYs (and, hence, the data fabric).
 
  • Like
Reactions: Drazick

imported_jjj

Senior member
Feb 14, 2009
660
430
136
I am just going to test latency and bandwidth in single-vs-dual channel mode. I want to know the latency cost between the two PHYs (and, hence, the data fabric).

Have you looked at BF1 DX11 vs DX12?
Check Computerbase and focus on 7700k, 6900k and Ryzen https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps
DX11 scales better to more cores and has higher FPS
DX12 scales poorly and has lower FPS
Was wondering if AMD's GPUs behave like that too.
 

looncraz

Senior member
Sep 12, 2011
722
1,651
136
I am starting to feel the same way. There isn't really any point in supporting the underdog, if they become just a bad as the monopoly. If I am going to buy from lying sleazebags, then I might as well buy intel and nvidia.

The scheduler issues only show up in a few cases - and AMD is patching those with those app developers.

The Windows scheduler would need a considerable reworking to behave well with Ryzen - there's no way Microsoft would be willing to put in the effort.

There are a few bugs related to the scheduler and Ryzen, but those should be patched with relative ease and no fanfare.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Just found another game that sees Ryzen as 16 cores.
i'm too lazy to list all specs http://www.legitreviews.com/cpu-bot...ed-on-amd-ryzen-versus-intel-kaby-lake_192585
Note on the right, 16 cores, 16 threads, the reviewer confirmed that it's an in game tool.
wildlands-cpu-usage-amd.jpg

wildlands-amd-versus-intel.jpg
 
Last edited:

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Rambus has made DDR4 and HBM2 memory controllers for GF 14lpp. I can't imagine other GF's customer than AMD to need one.

And as third party memory controllers are usually DFI-compliant they naturally share that standardized interface register level.

The DRAM controller firmwares are very similar in structure and size as well.
Otherwise I could believe that the IP was supplied by Rambus, but AFAIK Rambus doesn't have any GDDR5 IPs.
Steamroller based parts had DDR3-GDDR5 hybrid controller in them.
 

looncraz

Senior member
Sep 12, 2011
722
1,651
136
Have you looked at BF1 DX11 vs DX12?
Check Computerbase and focus on 7700k, 6900k and Ryzen https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps
DX11 scales better to more cores and has higher FPS
DX12 scales poorly and has lower FPS
Was wondering if AMD's GPUs behave like that too.

Will test when I swap back to the Windows 10 image. But I've only rarely seen DX12 be beneficial since the first few games to use it... usually buggy and unstable - and often slower.
 
  • Like
Reactions: Drazick

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Will test when I swap back to the Windows 10 image. But I've only rarely seen DX12 be beneficial since the first few games to use it... usually buggy and unstable - and often slower.

I think the core scaling would be the most interesting to check.
 

OrangeKhrush

Senior member
Feb 11, 2017
220
343
96
What I want to know is, is the 7700K stock or stock locked, if not locked then it is running up to 4.5Ghz with a 200mhz advantage already. I like the 3Ghz benches, that is Ryzens most stable clockspeed and it is an acid test on Intels IPC being real or clockspeed related.
 
  • Like
Reactions: ZGR and looncraz

formulav8

Diamond Member
Sep 18, 2000
7,004
522
126
Aside from the fact that our very own Roybotnik showed that the scheduler really is harming performance.

They are going by an AMD post apparently. I haven't personally looked into it myself as I won't have one to test until R5 comes out. I'll be using Win7 anyways.
 

Snarf Snarf

Senior member
Feb 19, 2015
399
327
136
Aside from the fact that our very own Roybotnik showed that the scheduler really is harming performance.


I think what the PR blog is saying is that Microsoft is not going to put the effort of rewriting the scheduler behavior for AMD. They don't have the market share or the money to send their own engineers to Microsoft to hold their hands through the code. Optimizations being done on a per app basis is probably AMD's strategy right now, it's easier to have a handful of engineers float around to different devs giving them help where it's needed vs being forced to do the heavy lifting for a project as big as an OS scheduler. The other option could be through microcode updates changing CPUID information that the OS sees, then they can give a short list of priorities to Microsoft to implement once the OS sees that a chip is Ryzen based.
 

looncraz

Senior member
Sep 12, 2011
722
1,651
136
What does it even matter how much better the yields get or waiting for a next stepping this and that when the darn software doesnt even support the hardware its running on!! Thats the point. :mad:Its like a car company saying oh look you can upgrade your gas hog rust bucket of a beater for this here shiny new car for less then your car is worth and it runs great on this here new fuel called jelly 205.but after you buy the new car there is no place that even carries jelly 205.. Give it a few months lots of gas stations will carry jelly 205. really?? when everyone is like what the hell is jelly 205?? is this a fair comparison or am I just way out in left field here?

That's just fine when your new shiny car also runs on normal gas just fine - just not as well as you know it can.

Ryzen isn't a bad performer at all - it's really really good. Crazy fast in some respects, very energy efficient, and quite stable. We're just waiting on updates that will make it even better.

Even when Ryzen is performing "badly" it is performing like a beast. Unless you run CPU Queen or memory latency benchmarks all day long, you'll never know there's a deficit. Ryzen is strongest where it matters most.

And then you get eight of them bad-boy cores!
 
Last edited:

Rifter

Lifer
Oct 9, 1999
11,522
751
126
I think what the PR blog is saying is that Microsoft is not going to put the effort of rewriting the scheduler behavior for AMD. They don't have the market share or the money to send their own engineers to Microsoft to hold their hands through the code. Optimizations being done on a per app basis is probably AMD's strategy right now, it's easier to have a handful of engineers float around to different devs giving them help where it's needed vs being forced to do the heavy lifting for a project as big as an OS scheduler. The other option could be through microcode updates changing CPUID information that the OS sees, then they can give a short list of priorities to Microsoft to implement once the OS sees that a chip is Ryzen based.

This might be believable, if linux hadn't fixed theirs in less than a week. But that fact that they did makes me think this should not take alot of man hours for MS to fix. Something else is going on here.
 
Status
Not open for further replies.