AMD - Zen chitchat

[CAT-THE-FIFTH]

Which is exactly what we should expect given that we already know games are not heavily affected by 4/4 vs 4/8 vs 6/12 vs 8/16.


I quite like this guys analysis of the gaming situation. He also covers fallout 4.

The thing is anything over 60FPS is pointless in FO4 due to the physics issues so if you avoid building large settlements,using certain DLC items and a certain class of mods Ryzen will be fast enough. However,no review really attempts to try and test performance in large player constructed settlements or with mods,which is where performance can really falter,and this is where the relative positions of the CPUs is more important.

Edit!!

Even with the streaming part of FO4,almost all of that done today is centred around mods too.

[shaithis]

http://uk.pcmag.com/chipsets-process...-amd-knows-why

Something for Ryzen users to be aware of.....keep checking for those BIOS updates

[Biscuit]

http://uk.pcmag.com/chipsets-process...-amd-knows-why

Something for Ryzen users to be aware of.....keep checking for those BIOS updates

Things like this are why I always wait for the second generation of major product refreshes. Looks like with Ryzen its best to wait until "version 2.0" of motherboard models start to see the light of day as current users are basically beta-testers.

[badass]

Things like this are why I always wait for the second generation of major product refreshes. Looks like with Ryzen its best to wait until "version 2.0" of motherboard models start to see the light of day as current users are basically beta-testers.

Me too. I'll never jump on something this new, no matter how much I don't like waiting. It beats doing what QA should have done in the first place.

Even enterprise grade kit is problematic when first released. The Xeon E5-2xxx v3 CPU's had an issue that caused lockups but no diagnostic data to be collected!

[MrNeil]

Some AMD Ryzen 5 1600 have already been delivered ?

http://www.guru3d.com/news-story/amd...y-selling.html

[watercooled]

Ughhh. Sorry to post my problems on this thread again, but I have a new (Asus) motherboard, and I'm still having major problems after it all working fine for a couple of days. The first sign of a problem was a strange input delay, the only way I can think to describe it, in Windows - opening programs would take notably longer than usual, the 'halo' around things like desktop icons when hovered over would lag way behind the cursor, switching tabs in task manager took longer, etc.

Then I plugged in a USB headset which caused my USB mouse (but not PS/2 keyboard) to stop responding, meaning I had to reboot using the keyboard. Reliability history reported hardware problems but without any useful information I could see.

The strange lagginess got worse, so I went into the UEFI setup to explore, and the system hung. Then I rebooted, and it got stuck at POST. Remove power and boot again, hangs again in BIOS.

Now I'm really, really fed up with this and just want a working PC. Granted I still have the same CPU but something seems horribly wrong with something. I'm not seeing masses of reports of this sort of problem but I don't have the patience or motivation to continue investigating. I'm now at the point of returning everything and going with Kaby Lake, which is a real shame.

It *could* be a faulty CPU - but what if it isn't? That's another week or so wasted messing about with a replacement that may not work, faffing about with RMA's, etc.

[kompukare]

It *could* be a faulty CPU - but what if it isn't? That's another week or so wasted messing about with a replacement that may not work, faffing about with RMA's, etc.

Does which USB port you plug it into make any difference? Or by using tree view (well View>Devices by Connection) in Device Manager, you should be able to figure out which ports are from where.

There are some USB troubleshooting programs, but since it's in the BIOS it can't really be the drivers. Have you tried a live Linux? An easy way to determine driver vs hardware (well hardware or BIOS I guess).

I wonder if one of your USB devices could be, well not faulty but maybe behaving in a way either the chipset, Ryzen SOC or the BIOS can't handle?

Don't think ASMedia have the best reputation for drivers (certainly 'asmedia usb crash' gets lots of hits mainly with 3.0 and 3.1 devices), but unsure whether the ones they supplied AMD have problems.

Once in Windows, did you look at DPC latency? (DPC Latency Checker from http://thesycon.de/eng/latency_check.shtml). But it's the slowdown in the BIOS which makes that unlikely to be of much use.

I hope that the design protects the CPU from, for example, a faulty USB device shorting the CPU.

[watercooled]

Yeah I did check DPC latency and it seems fine. I used the following program though, as IIRC the one you linked no longer works correctly post-Win7 and reports erroneous high numbers. http://www.resplendence.com/latencymon

The headset has been fine, and worked normally again after that problem, and the only other thing plugged into USB is my mouse. I've even tried different mouse+keyboard.

As far as I know, the USB 2.0 ports are supplied by the chipset, while at least some of the USB 3 ports are directly from the CPU, which is the weird part. I just don't understand how this problem carried across onto a new motherboard, unless of course there's some problem with the CPU/SoC's interface with the chipset. It was definitely 'oh not again' moment the first time the system hung, and similar to what happened on the first board, it just progressed from there.

I've not tried a live Linux distro, mainly because of the problems with POST mostly eliminating of it being a problem solely with Windows/drivers. On the last motherboard I was even troubleshooting with all of the HDDs unplugged as I was concerned with the number of unexpected resets the SSDs were being hit with.

Edit: It's not even like I have a lot of, or any unusual, devices connected to the motherboard. A HDD, two SATA SSDs and a GPU. All of which have been fine for ages on the previous platform. I've tested with different RAM too. It's all very bizarre and frustrating.

[kompukare]

Yeah I did check DPC latency and it seems fine. I used the following program though, as IIRC the one you linked no longer works correctly post-Win7 and reports erroneous high numbers. http://www.resplendence.com/latencymon

Yes, I think I have that too in this folder for hardware diagnostics but AFAIR it requires installing whereas the other one is portable. Ran fine on my Win10 installation though.

As for your problem, I think you should post over on OCUK or AT (either in the AMD RYZEN Builders Thread, the Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion, or on the motherboard forum), as it seems there are not enough Hexus'ers building Ryzen.

After two boards from both Gigabyte and Asus, I'm beginning to suspect that something is wrong with your CPU. The danger of a SOC is that a faulty CPU could have problems which would normally be considered motherboard issues.

[kompukare]

@watercooled
Saw this 'help me' post over on AT and while they seems to mainly getting crashes related to memory, they too had issues with the keyboard and mouse in the BIOS:
https://forums.anandtech.com/threads...ubles.2502149/
Strange that their Gigabyte tried running their 1.35V rated memory at 1.2V though. While I love undervolting, doing so by default is asking for trouble!

[[DW]Cougho]

Ughhh. Sorry to post my problems on this thread again.......

Have you installed the Intel LAN drivers? I had the same problem with them installed, I uninstalled them and I'm using the default Windows 10 LAN drivers and the system is working as it should be. A lot of people report the same thing with drivers newer than 18.5.

[watercooled]

@watercooled
Saw this 'help me' post over on AT and while they seems to mainly getting crashes related to memory, they too had issues with the keyboard and mouse in the BIOS:
https://forums.anandtech.com/threads...ubles.2502149/
Strange that their Gigabyte tried running their 1.35V rated memory at 1.2V though. While I love undervolting, doing so by default is asking for trouble!

I actually ended up trying two different sets of memory because the Corsair memory I originally used would not run at XMP clocks @2400 - it would boot-loop and eventually report a failed overclock if I manually set it to 2400. I swapped it for the other set of memory I have, from Crucial, which ran at 2400 by default. Both sets have also worked fine for extended periods in my i3 system.

It could easily be something simple, but as you say I'm also suspecting the CPU given the only two unproven parts are CPU and mobo, and I'm already on the second mobo. However I could also have been plain unlucky and received two faulty motherboards in a row.

If it were an inherent problem it would be very obvious by now. However my problem is still somewhat reminiscent of the one the guy over at OCUK had, in that it went bad, then worked again after a delay. It could simply be coincidental that the mobo swap got it working again, but I'm not sure why the BIOS recovery would have temporarily fixed it in that case.

Have you installed the Intel LAN drivers? I had the same problem with them installed, I uninstalled them and I'm using the default Windows 10 LAN drivers and the system is working as it should be. A lot of people report the same thing with drivers newer than 18.5.

Both motherboards had Realtek LAN chipsets. In my case it seemed to be USB which was misbehaving, but it's the USB2.0 ports which is especially weird as they're supplied from the chipset rather than the SoC itself AFAIK.

[kompukare]

So it begins to look like the rumoured 16C/32T quad channel platforms is for real. The chipset's not X399 though according to the latest rather X390 with X399 being the Naples dual-CPU chipset.

The threads Ryzen Strictly Technical and AMD X399 have a collection of details.

Also, CB did an article:
https://www.computerbase.de/2017-03/...les-x390-x399/ (Google Translated)
which puts the info together.
Other sources CPC who have been tweeting:

and Reddit:
https://www.reddit.com/r/Amd/comment...and_x390_leak/
The X390 diagram is a lot more detailed than the X399 though and it includes a mobo schema.

But as far as I can see, only the X399 has any kind of management (IPMI) which seems a strange omission from the X390 if it's meant to be a high-end workstation platform:


The diagram also name the CPU series as RZ2700 for the 16C/32T and RZ4700 for the Naples. Surely they won't use the Ryzen name for servers though?

[kompukare]

And a few minutes later a rumour of a 12C/24T processor for X390:
https://videocardz.com/67649/amd-ryz...hreads-spotted
Allegedly from an Alienware Area-51 R3 (the R2 is a Broadwell-E 6800K/6850K).
And Videocardz decoded its marking:

Plus they have some scores:

[kompukare]

Gaming performance keeps getting more attention and some of the youtubers have been busy. One managed to get PC3600 DDR4 working and even against a 5GHz 7700K (although only with PC3200 DDR4: they said there's little difference for Kabylake between 3200 and 3600 and I guess they didn't want to rerun those tests) it seems their Ryzen 1700X @ just under 4GHz performs very well.

GTA V was a surprise as I though when it came out it only used a few threads. Guess they fixed that.

Crysis 3 is something else which likes extra cores. They have this ranked by average FPS but the min 1% and 0.10% are far more interesting with the 4C/8T i7 doing very badly:

Of course this RAM speed requires very specific. Didn't see them listing the RAM used though but I think it has to Samsung chips and single-ranked. (No strike-through here I believe but that last sentence is not true.)
No it is mentioned: G.Skill Tridentz 3000MHz Samsung B-Die RAM F4-3600C16D-16GTZKW which Skinflint lists at £208. Not cheap stuff

Wonder how much they can change with microcode and BIOS updates versus what has to wait for Ryzen+ or minor revision.

[watercooled]

I wonder how much of that is due to the actual RAM bandwidth, and how much is inter-CCX latency? I agree that the fabric seems like low-hanging fruit for Zen2, and it would be interesting to know what led them to choose memory clocks rather than core (or half-core) clocks.

Any idea if the CCX's are clocked independently? If so, that could be a reason they didn't couple the fabric with core clocks on this generation.

WRT the >8 core models, it will be interesting to see how they're positioned in the market. I.e. I wonder if they'll be branded Ryzen or e.g. Opteron (assuming they keep that name).