• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

AMD Ryzen Thread: Affordable Core Act

NeOak

Member
I'm hearing rumblings that for some people there are games that won't launch with TR?

"Citation_needed".jpg
 

jrcbandit

Member
Everyone buying memory for Ryzen should be getting Samsung B-die chips single rank for maximum compatibility to run at listed speeds and best chance to overclock the RAM. This Reddit post lists most of the Samsung B-die available. Also, updating the Bios is imperative because the June and July bios updates with motherboards contain improved memory compatibility.
 
D

Deleted member 17706

Unconfirmed Member
That sounds like it's failing memory training. Make sure the retry count is set to 3 or more. Newer UEFI versions default that setting to 1.


Try gradually increasing the memory speeds. Don't just go straight to 2933. And even try something beyond 2933. Sometimes there is a "memory hole" when overclocking where you can't run a certain kit at some speeds, but it will run at faster speeds.
Try bumping the voltage from 1.35V to 1.40V and see if that helps too. It should still be safe for DDR4. (though overclocking is always done at your own risk)
Make sure that you also have the DRAM VBoot set to match, instead of leaving it on auto.


Make changes gradually and one at a time.

Thanks a lot for that write up. I'll give it a try when I get home later tonight.
 
Games so far appear to suffer more of high core to core latencies than lower IPC, and thanks to the mesh replacing the ring bus Skylake-X so far is worse there than Broadwell-E. So for gaming with more cores than 7700k, 6950x and the likes are actually a better fit.

I know, and I'd get the 6950x if it wasn't $1700. The 7900x is $1000 for the same number of cores/threads, and only some games see a performance penalty which is small enough that it's not worth $700 to me.
 

Datschge

Member
·feist·;245935438 said:
That's great. ^^

Related message from AMD employee amdmatt in the "gcc segmentation faults on Ryzen / Linux" discussion on AMD community

In that discussion responses of now multiple people indicate that Ryzen chips manufacture in June and later may be unaffected. The UA line on the chips seems to mention year and week when it was manufactured, eg. UA 1725SUS may refer to 25th week (latter part of June) in 2017.

The focus on specific heavy workloads under Linux is imo disingenuous. According to the discussion the same issues has been reproduce under BSD and in WSL under Windows. Also some people reported that (under Windows) they had issues to achieve stability in higher RAM speeds that are supposed to work normally and later were able to reproduce the segmentation faults (under Linux).

So my personal suggestion would be if your chip has a manufacturing date older than June and has stability and compatibility issues that shouldn't be there check for segmentation faults under Linux. Then if your chip is affected apply for RMA.
To follow up on that, seems like multiple people affected by the segmentation faults also manage to crash Ashes of the Singularity - Escalation - DX12 benchmark mode (DX11 is not affected). So that may be a good first check under Windows for testing if one's chip is affected.
 

Datschge

Member
How can a multithreading focused processor fail like that on Linux compilation and pass QA ffs amd
It's hard to reliable reproduce, everything points to early bad silicons that slipped through QA. Being an issue with bad silicon a fix in the microcode or workaround in software is unlikely to come (according to the Phoronix article AMD only stepped up their Linux testing/QA in reaction to this).
 

Dr.Acula

Banned
What timings are you using if you don't mind me asking? If mine doesn't hold at 2933, I'll need to try something else and don't want to default to 2133.

So I'm just using XMP and the 2667 setting holds at default 15-17-17-17-35 with a 1.35v.

As I said, this weekend I'm gonna experiment more with manual mode, I tried XMP 2933, even looser timings it didn't work, maybe 2933 manual will do the trick.

ASRock Fatal1ty AB350 Gaming K4 BIOS ver. 3.00 (7/18/2017).
 

thelastword

Banned
I'm hearing rumblings that for some people there are games that won't launch with TR?
Dirt has no issue with game mode, some of the cores are disabled and it works without issue. I'm sure it will be patched to not crash under 32 cores...It's a Dirt issue not TR tbh.......
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
Dirt has no issue with game mode, some of the cores are disabled and it works without issue. I'm sure it will be patched to not crash under 32 cores...It's a Dirt issue not TR tbh.......
Indeed. Some of the crashing software is seeing so many cores for the first time since its inception. Give devs some time to react.
 

Paragon

Member
To follow up on that, seems like multiple people affected by the segmentation faults also manage to crash Ashes of the Singularity - Escalation - DX12 benchmark mode (DX11 is not affected). So that may be a good first check under Windows for testing if one's chip is affected.
So I just run the DX12 benchmark and I'm good if it doesn't crash? (it didn't)
Or do I have to keep running the benchmark on a loop?
 

Datschge

Member
So I just run the DX12 benchmark and I'm good if it doesn't crash? (it didn't)
Or do I have to keep running the benchmark on a loop?
Nobody on the AMD community clarified that so far, but by the sound of it the benchmark may always crash before completion for those with affected chips.
 
D

Deleted member 17706

Unconfirmed Member
That sounds like it's failing memory training. Make sure the retry count is set to 3 or more. Newer UEFI versions default that setting to 1.


Try gradually increasing the memory speeds. Don't just go straight to 2933. And even try something beyond 2933. Sometimes there is a "memory hole" when overclocking where you can't run a certain kit at some speeds, but it will run at faster speeds.
Try bumping the voltage from 1.35V to 1.40V and see if that helps too. It should still be safe for DDR4. (though overclocking is always done at your own risk)
Make sure that you also have the DRAM VBoot set to match, instead of leaving it on auto.


Make changes gradually and one at a time.

Thanks again for this!

It may have been a combination of things. I was using a BIOS version from May, so I updated it to the latest one from July. I then went in and ignored your advice to take in steps, and bumped it up to 2933 straight away, but I set the retry count to 3 and manually set the RAM voltage, including VBoot to 1.35V.

I'm not booted into Windows and it's working just fine at 2933 Mhz!

I may end up returning those G.Skill sticks since they are quite a bit more expensive. Not sure if it's worth the money to get 200 Mhz more and tighter timings (tested latency listed at: 14-14-14-34 vs. 15-17-17-35).
 
Interesting. Without adjusting BCLK?

Would you tell me what timings and voltage settings you are using? I know I can't expect it to fit exactly with my setup, but might be a good starting point.

I didn't do anything with the BLCK. Timings I'm using: 16-17-17-35-69-2T. So fairly slow, but I haven't messed around too much, yet. Currently using 1.27V on the RAM (BIOS) with HW Monitor reading 1.264V as the actual value in Windows. Keep in mind when using XMP (also 2933, but better timings) I couldn't get that RAM stable at 1.35V!

Edit: I see you fixed it already ^^ Try Prime95s Blend test, that's what crashed my RAM when using XMP 2933 Mhz settings.
 

JohnnyFootball

GerAlt-Right. Ciriously.
I will say that I do feel AMD could have done more to entice people to buy the X370 chipsets. Aside from SLI there is virtually nothing on those boards distinguishing them from the 350.

I dunno I feel like maybe if they had added additional PCI Express lanes, I'd be enticed.
 

SpotAnime

Member
Well, good news. XMP profile running DDR4-3000 at 2933MHz was stable all night. On a side note, playing games all night and core temps didn't even crack 50C. On top of that, this Noctua is wisper quiet. My old PC sounded like a jet engine, this one stays cool and so silent the HD is louder than the fans.

So happy I decided to treat myself and finally upgrade.
 

Jrs3000

Member
Ok GAF need some thoughts on the matter. Ryzen 3 1200 for $99 or 5 1600 for $192. I'm doing a budget build and plan to overclock to 4ghz. I haven't found any benches to see how big of a performance delta there is between them in Fps and strategy games. Thoughts?
 

JohnnyFootball

GerAlt-Right. Ciriously.
Ok GAF need some thoughts on the matter. Ryzen 3 1200 for $99 or 5 1600 for $192. I'm doing a budget build and plan to overclock to 4ghz. I haven't found any benches to see how big of a performance delta there is between them in Fps and strategy games. Thoughts?

Is this a serious question? Or a typo? cuz it seems odd that someone would be torn between a 1200 and 1600 as they're two very different CPUs.

If you have the option of getting a 1600 you get a 1600.

A 1600 should be the bare minimum when considering Ryzen.

I can't see any reason to get a 1200. Even if you're strapped for cash just save up a little while longer.
 

Jrs3000

Member
Is this a serious question? Or a typo? cuz it seems odd that someone would be torn between a 1200 and 1600 as they're two very different CPUs.

If you have the option of getting a 1600 you get a 1600.

A 1600 should be the bare minimum when considering Ryzen.

I can't see any reason to get a 1200. Even if you're strapped for cash just save up a little while longer.

Yup serious. Only so much funds available. I can have it up an running now with a 1200. I know a 1600 is better but is it double the price better? If I'm getting 70fps vs 100fps in games at 1080p I'm still over 60 and would be fine for Now with the lesser option.
 

Colbert

Banned
Ok GAF need some thoughts on the matter. Ryzen 3 1200 for $99 or 5 1600 for $192. I'm doing a budget build and plan to overclock to 4ghz. I haven't found any benches to see how big of a performance delta there is between them in Fps and strategy games. Thoughts?

Look into this channel. You should find benchmarks for both CPUs there:

https://www.youtube.com/channel/UCI8iQa1hv7oV_Z8D35vVuSg

Ryzen R3 1200 Test: https://www.youtube.com/watch?v=IoR0RDETGZI
Ryzen R5 1600 Test: https://www.youtube.com/watch?v=UfNMn7RWgLw
 

cyen

Member
Yup serious. Only so much funds available. I can have it up an running now with a 1200. I know a 1600 is better but is it double the price better? If I'm getting 70fps vs 100fps in games at 1080p I'm still over 60 and would be fine for Now with the lesser option.

Thats the good thing about AM4 since you can latter upgrade to Zen2 if you need a better cpu in a few years.
 

Jrs3000

Member

Khaz

Member
Just what I needed to see. It's viable OC for 1080p 60fps gaming with a strong gpu. I'll stick with it until zen 2 is out. I'll put my bucks into another part of my system.

I'd say get the 1300X instead, for $30 more. It should give you about the same clock speed as better Ryzens, but without the additional cores and thread. Given that games used to favour Hz more than threads, it is a good cheap starting point for current and past gaming.
 

Dr.Acula

Banned
So I tried some extra manual timing tests with my RAM, and my PC just refused to power on, just dead-dead, I had to let it sit with the powerbar off for a minute before it finally booted up... aaaaand I'm scared to try and get my ram to the 2933 mark. 2667 it is :p
 

nubbe

Member
So I tried some extra manual timing tests with my RAM, and my PC just refused to power on, just dead-dead, I had to let it sit with the powerbar off for a minute before it finally booted up... aaaaand I'm scared to try and get my ram to the 2933 mark. 2667 it is :p

the hardware wont take damage unless you over voltage too much.
Just changing memory settings and timings will do no harm, but the the computer will most like not be able to operate since the electric signal will be too weak to react properly to the switching in the logic.

Modern motherboards will auto reset to default settings after a few failed boots
 
Well, despite my earlier misgivings about the 1950x that I posted about in this thread, I went ahead and purchased one! I have a G-Sync monitor and I really don't care about frame rates above 100, I just want smooth gameplay and lots of processing power to run multiple apps at the same time even while gaming.

I had considered waiting for the 7920x but with the 7900x already using about the same amount of power as the 1950x and with Intel's awful TIM and the cooling struggles skylake-x has had, I decided the extra few frames per second just weren't worth it.

Plus I'm more optimistic about the upgrade possibilities on my x399 motherboard in a couple years versus what I'd have with the x299 platform.

Now I'm just hoping my current DDR4 3200 will work because I really can't afford another $350 for new RAM.
 

Newboi

Member
Well, despite my earlier misgivings about the 1950x that I posted about in this thread, I went ahead and purchased one! I have a G-Sync monitor and I really don't care about frame rates above 100, I just want smooth gameplay and lots of processing power to run multiple apps at the same time even while gaming.

I had considered waiting for the 7920x but with the 7900x already using about the same amount of power as the 1950x and with Intel's awful TIM and the cooling struggles skylake-x has had, I decided the extra few frames per second just weren't worth it.

Plus I'm more optimistic about the upgrade possibilities on my x399 motherboard in a couple years versus what I'd have with the x299 platform.

Now I'm just hoping my current DDR4 3200 will work because I really can't afford another $350 for new RAM.

Congrats, let us know how the 1950X performs for you!
 
D

Deleted member 17706

Unconfirmed Member
Anyone else been using the Ryzen Master utility for overclocking?

It seems like overclocking with the utility still retains all of the idle downclocking features that the CPU utilizes when it's not overclocked. I think this has to be all set manually in the various BIOS settings when overclocking via the BIOS.

I think I'm just going to use the utility for the time being. The only downside is that you have to activate it every time you restart Windows.
 

NeOak

Member
Anyone else been using the Ryzen Master utility for overclocking?

It seems like overclocking with the utility still retains all of the idle downclocking features that the CPU utilizes when it's not overclocked. I think this has to be all set manually in the various BIOS settings when overclocking via the BIOS.

I think I'm just going to use the utility for the time being. The only downside is that you have to activate it every time you restart Windows.
Make a shortcut for the Start up? 🤔
 

Datschge

Member
It seems like overclocking with the utility still retains all of the idle downclocking features that the CPU utilizes when it's not overclocked. I think this has to be all set manually in the various BIOS settings when overclocking via the BIOS.
That's the P-states and Cool 'n' Quiet (which needs to be enabled so the P-states are made use of). Overclocking shouldn't disable C'n'Q but it seems plenty boards do it anyway. Just re-enable it.
 

bomblord1

Banned
I've seen a bit of conflicting info so I thought I'de ask here. Can a 1700 (not X) overclock to 4ghz? Some videos say not to because it's only rated at 65W but I've found other videos/forum posts where people claim to have done it.
 
I've seen a bit of conflicting info so I thought I'de ask here. Can a 1700 (not X) overclock to 4ghz? Some videos say not to because it's only rated at 65W but I've found other videos/forum posts where people claim to have done it.

Overclocking capability doesn't have anything to do with the rated TDP, i.e. the 1700 can easily exceed its 65W TDP. It appears like there is a tendency that the 1700 doesn't overclock as well as either the 1700X or 1800X, but the difference is 200Mhz at most. ~4 Ghz is the wall for Ryzen, so you might only get it to ~3.8 or so.
 

bomblord1

Banned
Overclocking capability doesn't have anything to do with the rated TDP, i.e. the 1700 can easily exceed its 65W TDP. It appears like there is a tendency that the 1700 doesn't overclock as well as either the 1700X or 1800X, but the difference is 200Mhz at most. ~4 Ghz is the wall for Ryzen, so you might only get it to ~3.8 or so.

Thanks
 

Wozzer

Member
Put together a new Ryzen build on the weekend but have unfortunately been encountering constant BSOD crashes, often only minutes after having boooted into Windows.

My build is; https://pcpartpicker.com/list/HmF4bj
BIOS screens; http://imgur.com/a/YAzrO

I have attempted;
  • Reseating all components,
  • Removing all optional drives (leaving m.2),
  • Updating all drivers to latest,
  • Updating Bios to latest,
  • Formatting and reinstalling Windows fresh

Even with the above, on a fresh clean install with latest manufacturer drivers I'm seeing the same behaviour. The machine will noticeably hitch, frequently noticeable when playing a game for example, and eventually freeze into a reboot or BSOD.

I'd consider the GPU a culprit, but it's the one component I kept from my past build and has always been completely stable.

Most probable culprit seems to be RAM, especially since Ryzen is known to be sensitive and it's not a kit on the boards QoL (though before buying I've seen others have success). Aware the RAM is known to have corruption issues, but haven't seen and aren't using RGB software.

Failing that it could be a motherboard/PSU issue, which I'm hoping not.

Considered next steps;
  • Manually clock the RAM, particularly raising voltage (advice here appreciated, see BIOS screens above),
  • Try to stress test with single RAM (s) inserted,
  • Run another mem check (last one was error free at 95%, then froze on the reboot so I missed the results, and took several hours -_-),
  • Try my prior build PSU,
  • Cry into a pillow and return parts (whilst I can)

Anyone faced similar issues or have any possible solutions I'm not currently considering?
 
Put together a new Ryzen build on the weekend but have unfortunately been encountering constant BSOD crashes, often only minutes after having boooted into Windows.

My build is; https://pcpartpicker.com/list/HmF4bj
BIOS screens; http://imgur.com/a/YAzrO

I have attempted;
  • Reseating all components,
  • Removing all optional drives (leaving m.2),
  • Updating all drivers to latest,
  • Updating Bios to latest,
  • Formatting and reinstalling Windows fresh

Even with the above, on a fresh clean install with latest manufacturer drivers I'm seeing the same behaviour. The machine will noticeably hitch, frequently noticeable when playing a game for example, and eventually freeze into a reboot or BSOD.

I'd consider the GPU a culprit, but it's the one component I kept from my past build and has always been completely stable.

Most probable culprit seems to be RAM, especially since Ryzen is known to be sensitive and it's not a kit on the boards QoL (though before buying I've seen others have success). Aware the RAM is known to have corruption issues, but haven't seen and aren't using RGB software.

Failing that it could be a motherboard/PSU issue, which I'm hoping not.

Considered next steps;
  • Manually clock the RAM, particularly raising voltage (advice here appreciated, see BIOS screens above),
  • Try to stress test with single RAM (s) inserted,
  • Run another mem check (last one was error free at 95%, then froze on the reboot so I missed the results, and took several hours -_-),
  • Try my prior build PSU,
  • Cry into a pillow and return parts (whilst I can)

Anyone faced similar issues or have any possible solutions I'm not currently considering?

What error do you get on your BSOD? Have you tried using a BSOD viewer to look at them a little more closely? They might provide a clue.
 
Put together a new Ryzen build on the weekend but have unfortunately been encountering constant BSOD crashes, often only minutes after having boooted into Windows.

My build is; https://pcpartpicker.com/list/HmF4bj

FWIW, I've read about problems with Trident Z RGB memory possibly becoming corrupted in Ryzen builds due to software conflicts involving RGB software. Here's one such thread I could find on the issue. Not sure if these issues have been resolved by now or if your symptoms match up, but it's something to consider.
 

3x0

Neo Member
All voltage configuration posted in the BIOS screens here; http://imgur.com/a/YAzrO

I'd verify the vCore values with HWiNFO under Windows. Start a stress test like RealBench, Prime95 or Cinebench R15 single/multi-core test, and double click the vcore line in HWiNFO to get a value diagram.
It's possible the board isn't supplying enough voltage regardless of the BIOS setting (excessive voltage droop under load for ex.).
Have you saved any previous BSOD dumps? If yes, are the BugCheck codes the same always (0x0000003b in this case)
It's very weird since you're pretty much running stock BIOS settings, so there shouldn't be issues like this.
 
Put together a new Ryzen build on the weekend but have unfortunately been encountering constant BSOD crashes, often only minutes after having boooted into Windows.

My build is; https://pcpartpicker.com/list/HmF4bj
BIOS screens; http://imgur.com/a/YAzrO

I have attempted;
  • Reseating all components,
  • Removing all optional drives (leaving m.2),
  • Updating all drivers to latest,
  • Updating Bios to latest,
  • Formatting and reinstalling Windows fresh

Even with the above, on a fresh clean install with latest manufacturer drivers I'm seeing the same behaviour. The machine will noticeably hitch, frequently noticeable when playing a game for example, and eventually freeze into a reboot or BSOD.

I'd consider the GPU a culprit, but it's the one component I kept from my past build and has always been completely stable.

Most probable culprit seems to be RAM, especially since Ryzen is known to be sensitive and it's not a kit on the boards QoL (though before buying I've seen others have success). Aware the RAM is known to have corruption issues, but haven't seen and aren't using RGB software.

Failing that it could be a motherboard/PSU issue, which I'm hoping not.

Considered next steps;
  • Manually clock the RAM, particularly raising voltage (advice here appreciated, see BIOS screens above),
  • Try to stress test with single RAM (s) inserted,
  • Run another mem check (last one was error free at 95%, then froze on the reboot so I missed the results, and took several hours -_-),
  • Try my prior build PSU,
  • Cry into a pillow and return parts (whilst I can)

Anyone faced similar issues or have any possible solutions I'm not currently considering?


my biosd looks exactly the same is yours. I have a b350 F gaming from ASUS.

I have the 3200 trident RGB version .

AS of today, I am able to reach the speed of 3066 just fine. there is not tweaking. The only thing you need to do is change to DOCP and pick the 3066 speed. the voltage will be increased automatically to 1.3. Everything else I left the same. Basically just change DOCP and pick 3066, should work
also, make sure you are using the correct ram slots for 2 sticks. check the mobo manual
 
D

Deleted member 17706

Unconfirmed Member
That's the P-states and Cool 'n' Quiet (which needs to be enabled so the P-states are made use of). Overclocking shouldn't disable C'n'Q but it seems plenty boards do it anyway. Just re-enable it.

Thanks. This should be an option in the BIOS settings, right? I'll do some searching.
 
Top Bottom