AMD - Zen chitchat

[CAT-THE-FIFTH]

After nearly 5 years and around 380000 views,it is time to start a new thread for Ryzen:

http://forums.hexus.net/cpus/241925-...-chitchat.html
http://forums.hexus.net/cpus/200143-...-chitchat.html

The threads for both Bulldozer and Piledriver have had over half a million,yes,half a million views over six years.

Lets,see if we can beat them with the new one!!

[Phage]

All over it.

[AETAaAS]

Given the previous threads were named after the architectures, shouldn't this be named AMD Zen chitchat?

[DanceswithUnix]

One thought has occurred, which is quite removed from the Ryzen release so best said here. There was supposed to be an ARM variant of Zen. Now I really want one to play with!

[kompukare]

One thought has occurred, which is quite removed from the Ryzen release so best said here. There was supposed to be an ARM variant of Zen. Now I really want one to play with!

I though I had read that K12 aka Project Skybridge had been cancelled or postponed, although the Wikipedia still says it's due sometime in 2017
https://en.wikipedia.org/wiki/AMD_K12

But this ExtremeTech article from April 2016 say it was cancelled.
https://www.extremetech.com/extreme/...amd-processors

[HalloweenJack]

Keller veto`d the ARM cpu from his arrival I read somewhere (although ruzen does have an ARM a5 built in )

[CAT-THE-FIFTH]

Given the previous threads were named after the architectures, shouldn't this be named AMD Zen chitchat?

First page and the pedantry has already started!! Damn it I feel the pressure already!!

[CAT-THE-FIFTH]

I am going to post a summary of where we are at regarding the SMT fixes so far.


IMPORTANT INFO.

SMT was not deactivated in many reviews leading to regressions in gaming performance:
http://i.imgur.com/UFaWvLe.jpg




The regressions are reasonably noticeable and many review sites are not aware of this.

Ryzen was also not launched with Windows drivers:

http://www.legitreviews.com/amd-ryze...view_191753/15

You hit the nail on the head on all your comments. I really wish the Windows drivers were ready before we were given the parts to review. Instead we got a statement 24 hours before launch from AMD saying that they'll be coming in 30 days if all goes well. Game optimizations will be hit or miss when they come, but they appear to be coming. That takes time though and we'll see what happens.

I put it in the conclusion on the last page. The quote came direct from AMD's John Taylor. There was talk of it coming with Ryzen 5 and then they said 'in the next month' in an official statement that was e-mailed out last night.

So there might be room for some better performance figures as time progresses.


Update!!

The SMT issue is down to Windows 10. It looks like The Stilt has done even more testing:

https://forums.anandtech.com/threads...#post-38775732

I did some 3D testing and eventhou there is not nearly enough data to confirm it, I'd say the SMT regression is infact a Windows 10 related issue.
In 3D testing I did recently on Windows 10, the title which illustrated the biggest SMT regression was Total War: Warhammer.

All of these were recorded at 3.5GHz, 2133MHz MEMCLK with R9 Nano:

Windows 10 - 1080 Ultra DX11:

8C/16T - 49.39fps (Min), 72.36fps (Avg)
8C/8T - 57.16fps (Min), 72.46fps (Avg)

Windows 7 - 1080 Ultra DX11:

8C/16T - 62.33fps (Min), 78.18fps (Avg)
8C/8T - 62.00fps (Min), 73.22fps (Avg)

At the moment this is just pure speculation as there were variables, which could not be isolated.
Windows 10 figures were recorded using PresentMon (OCAT), however with Windows 7 it was necessary to use Fraps.

http://www.pcgameshardware.de/Ryzen-...eview-1222033/

It seems the 4C versions which appear to use only one CCX might be fine.

[CAT-THE-FIFTH]

How to install Ryzen under Windows 7:

https://forums.anandtech.com/threads...500572/page-11

Copy dism.exe, boot.wim & install.wim (sources directory) files from the Windows 7 ISO image to your hard-drive.
Download the USB driver for Ryzen: https://1drv.ms/u/s!Ag6oE4SOsCmDhGIQJdHdaXC-_w-C
Extract the package to the same directory as DISM is located.
Enter the folders containing the individual driver files and check that there is not "Unblock" button visible. If there is, you need to manually toggle it for each and every file.

- "DISM /mount-wim /wimfile:boot.wim /index:2 /mountdir:x:\xxx" (x:\xxx = a temporary path of your choice, make sure to have ~20GB of space available for install.wim).
- "DISM /image:x:\xxx /add-driver /driver:Ryzen_USB_W764\ /recurse /forceunsigned"
- "DISM /unmount-wim /mountdir:x:\xxx /commit"

Windows 7 install.wim files contain four different OS variants, regardless of the officially stated edition of the media you have (Home, Professional, Ultimate).
The index order within the install.wim is always the same, regardless of the edition: Index 1 = Home Basic, Index 2 = Home Premium, Index 3 = Professional, Index 4 = Ultimate.

So if your media is for Professional edition, you need to make the changes to Index 3. If it is a Ultimate media, then make the changes to Index 4, etc.

- "DISM /mount-wim /wimfile:install.wim /index:x /mountdir:x:\xxx" (x:\xxx = a temporary path of your choice, make sure to have ~20GB of space available for install.wim).
- "DISM /image:x:\xxx /add-driver /driver:Ryzen_USB_W764\ /recurse /forceunsigned"
- "DISM /unmount-wim /mountdir:x:\xxx /commit"

After you have added the drivers to both of the WIMs, you can install Windows 7 from a USB drive and using USB keyboard and mouse.
Make sure that you don't use USB ports provided by a 3rd party manufacturer (other than ASMedia), as there are still no drivers for those in the media.
After the installation, install the Relive chipset driver pack for Windows 7 (available at AMD.com).

[CAT-THE-FIFTH]

I saw this on AT forums:


https://forums.anandtech.com/threads...#post-38771400

Causes of poor gaming relative to CPU performance of Ryzen:

1. Windows is load-balancing across CCXes.
This means that a thread is being moved around on the CPU - which is normal - so that a single core isn't used more than others. On Ryzen, that needs to happen ONLY within a CCX, otherwise you will incur a massive penalty when that thread no longer finds its data in the caches of the CCX.

2. SMT hurts single threaded performance due to shared structure.
Ryzen statically partitions three structures to support SMT:
a. Micro-op queue (dispatcher)
b. Retirement queue
c. Store queue

This means that, with SMT enabled, these resources are cut, potentially, in HALF (mind you, these are just queues that impact throughput of a single thread).

3. Memory latency quirks still not worked out.
Gaming can be quite sensitive to memory latency and bandwidth. These issues will be, most likely, remedied with BIOS updates.

Combined you can see, clearly, what is happening and most of the reviews make sense.

A Windows driver update to treat each CCX almost as if it were its own CPU will help immensely. The SMT problem is likely PERMANENT... unless AMD can adjust the partitioning with microcode, which I doubt.

What this all means is simple: once the Windows update has landed, BIOSes are patched up, and SMT is disabled, an 8-core Ryzen will likely be competitive with a quad i7 in gaming while blowing past it in multi-threaded. If all you do is game, then the 1700 may well become a very valid option that will work increasingly better in future games.

This also lets us know where Zen 2 will be able to improve the most. Make the impacted queues competitively shared (or just a little larger), improve inter-CCX communications, decouple the L3 speed from core speeds (for higher core clocks), and a few other relatively simple tweaks and you have a second generation Ryzen that steals the show.

We also know why AMD hasn't released anything other than their 8-core chips - these issues need to be ironed out in production. You need thousands of eyes and testers and numerous companies each responding to their customers' needs to get a grip on what is most important to fix before finalizing Zen 2.

Looking at what he has posted in the past,he seems quite technically competent - what do people think of his analysis?

[Phage]

"once the Windows update has landed, BIOSes are patched up, and SMT is disabled, an 8-core Ryzen will likely be competitive with a quad i7 in gaming while blowing past it in multi-threaded. If all you do is game, then the 1700 may well become a very valid option that will work increasingly better in future games."

Exactly this. I'm going to give it a month or so, and see what we get. If its approx equal to a 7700K in gaming - It's a definte sale of two machines.

[CAT-THE-FIFTH]

"once the Windows update has landed, BIOSes are patched up, and SMT is disabled, an 8-core Ryzen will likely be competitive with a quad i7 in gaming while blowing past it in multi-threaded. If all you do is game, then the 1700 may well become a very valid option that will work increasingly better in future games."

Exactly this. I'm going to give it a month or so, and see what we get. If its approx equal to a 7700K in gaming - It's a definte sale of two machines.

Well I do hope it also means we can get a few percent out of FO4 while they are at it - I tend to play a reasonable amount of FO4 and Skyrim,so I would want an upgrade in them too. Its why I am so enthused by AMD working with Bethesda to get Vulkan into their games. The next Elder Scrolls with Vulkan would be an utter win.

Bye! Bye! poor CPU utilisation.

[kalniel]

A lot of the reviews have hinted quietly that there were quite a few patches from AMD/motherboard makers in the lead up to NDA lift. Rather than thinking they've missed something obvious, I wonder if it's more of a side effect of them fixing a more serious problem(s) in the lead up?

Well I do hope it also means we can get a few percent out of FO4 while they are at it - I tend to play a reasonable amount of FO4 and Skyrim,so I would want an upgrade in them too. Its why I am so enthused by AMD working with Bethesda to get Vulkan into their games. The next Elder Scrolls with Vulkan would be an utter win.

Bye! Bye! poor CPU utilisation.

It'll be for doom/prey/rage though init.

[CAT-THE-FIFTH]

A lot of the reviews have hinted quietly that there were quite a few patches from AMD/motherboard makers in the lead up to NDA lift. Rather than thinking they've missed something obvious, I wonder if it's more of a side effect of them fixing a more serious problem(s) in the lead up?

Remember that rumour of a bug which cratered performance:

https://www.reddit.com/r/Amd/comment...ils_about_zen/

It'll be for doom/prey/rage though init.

Sadly most likely the case - its large open world games which really would take advantage of it though.

[DanceswithUnix]

- what do people think of his analysis?

His point 1 is certainly true, though also almost certainly incomplete as life isn't usually that simple.

Not convinced by the rest though.

If the queues are statically partitioned and too narrow, then single thread performance would suck (like it did for early construction cores), which it clearly doesn't. If performance is suffering when threading is turned on, then that implies contention which means the threads must already be competing for any resources that matter.

That is backed up by looking at the graphs you posted above. WatchDogs 2 hurts Intel roughly as much as AMD, in the majority of games threading hurts Intel as well just not as much as AMD. To me the threading doesn't appear that far from what Intel is doing.

There is clearly a problem with thread mobility, which will require a fix to the Windows scheduler. That is the same problem AMD had with Piledriver, and it is probably something that they are at the mercy of Microsoft to get right. I have seen scheduler issues tank program performance way more than the amount we are getting here, and tuning programs to remove that can be rather hard. Perhaps some of that tuning has happened on the Intel platform for some of the games, we don't know. Fixing the process migration cost model in the scheduler would be the first broad step though, until that is done we can't really see any deeper.

Sad to say, what I really want to see is database benchmarks. Databases are boring as heck, but the world uses lots of them and they showed some bad scaling problems in Piledriver in accessing shared locks (probably down to the cache structure) which I would hope are removed in Zen but would like to know for sure.

Edit: Tied to the scheduler is power saving features and the whole issue of when to shut down some cores and boost the remaining ones with the saved power budget. That must be quite hard to get right, and I suspect it isn't working yet.

[Xlucine]

With the issues in moving between CCX's, maybe the 6 core parts would be better off with 2 cores turned off on one CCX and the other unmolested? Then in a game you'll have a full CCX to run the game, with the crippled CCX for background tasks