Trying to OC my PC and ran in a few issues, so help appreciated

[salatji]

Ok, my PC is a msi x370 gaming pro, 1600x, 16gb 3200 cl16 rgb trident z and a Gigabyte 1070 Gaming.

 

I have run into several issues/questions.

 

Question 1: Can I use XMP for the memory (currently manages the 2933 xmp) if I manually OC the CPU? I am asking this because I did not even get to OC the 1600x to 3.8/3.9 GHz, it always kept crashing/failed to boot.

 

Question 2: Is there a possibility that MSI Afterburner somehow manages to OC (and my 1070 seemed to run stable in kombuster with a OC of 150 mhz core and 350 mhz memory) and then fails to apply it in games? I tried to launch cities skylines after that, and at first I got in the menu at 10 fps and failed to load a map and later the game did not even launch properly. After restarting, everything seems to be fine again.

 

With both attempts of OC so far I ran in a wall, so I am kinda annoyed because I "went by the books" and still failed in CPU OC (never got even 3.8 GHz running), GPU OC (ran in kombuster at the above clocks but failed to load cities skylines) and the only thing that somehow worked out so far is the XMP-selection for 2933 mhz (and even then I am annoyed because I was hoping to get ideally the 3000 mhz or maybe even the 3200 mhz out of my memory...

 

If someone can help me out with tips, videos/tutorials or some info if there are programs who have issues with e.g. afterburner, feel free to post, appreciate every help I can get.

 

youtube tutorials about OC I watched and tried to use for overclocking: 

 and 

 

 

and 

 

 

[rEkmAInc1csBnlM6C85H]

Questions 1:

 

What is your memory configuration?

• 2 x 8GB?
• 4 x 4GB?

With 4 x 4GB you will have a very hard time getting to the memory rated spec with XMP unless you got silicon lottery lucky (My go to is 1 DIMM per channel). If you are running 4 DIMMs try running only two it near guarantees memory stability.

 

Question 2:

 

It is allot easier to run a synthetic than a game.

 

Compared to Kombuster in a game you are jumping between power states and loading and unloading the ram allot more often. It is not uncommon for OCs to be rock stable in synthetics (Super position, Heaven, Time spy) but fail in real world test cases.

[salatji]

9 minutes ago, Andrew 1337 said:

Questions 1:

 

What is your memory configuration?

• 2 x 8GB?
• 4 x 4GB?

With 4 x 4GB you will have a very hard time getting to the memory rated spec with XMP unless you got silicon lottery lucky (My go to is 1 DIMM per channel). If you are running 4 DIMMs try running only two it near guarantees memory stability.

 

Question 2:

 

It is allot easier to run a synthetic than a game.

 

Compared to Kombuster in a game you are jumping between power states and loading and unloading the ram allot more often. It is not uncommon for OCs to be rock stable in synthetics (Super position, Heaven, Time spy) but fail in real world test cases.

It's 2x8. Does XMP-profiles affect the results of CPU overclocking? Or are memory and CPU overclocks independent?

 

 

About the GPU-overclock: So if I tone down the OCs, they might still work and I just have to test around?

[rEkmAInc1csBnlM6C85H]

5 minutes ago, salatji said:

It's 2x8. Does XMP-profiles affect the results of CPU overclocking? Or are memory and CPU overclocks independent?

 

 

About the GPU-overclock: So if I tone down the OCs, they might still work and I just have to test around?

 

On Ryzen, RAM speed effects infinity fabric speeds internally on the CPU.

 

I know for Gen 3 Ryzen you can decouple those, but if I recall you can not decouple them on Gen 1. So you could have the CPU or the RAM being the issue in this case.

[salatji]

2 minutes ago, Andrew 1337 said:

 

On Ryzen, RAM speed effects infinity fabric speeds internally on the CPU.

 

I know for Gen 3 Ryzen you can decouple those, but if I recall you can not decouple them on Gen 1. So you could have the CPU or the RAM being the issue in this case.

Well, guess I'll get myself a 3600 for black friday/christmas then... Was already thinking about it, but this reaffirms my decision

[rEkmAInc1csBnlM6C85H]

2 minutes ago, salatji said:

Well, guess I'll get myself a 3600 for black friday/christmas then... Was already thinking about it, but this reaffirms my decision

You may want to check out this video if you are going Gen3.

 

[salatji]

4 minutes ago, Andrew 1337 said:

You may want to check out this video if you are going Gen3.

 

Well, I only intended to swap out the CPU as I am fairly limited by my budget and the 200€ is probably the limit of my personal christmas present, so even if 3600 mhz memory would be better, I probably will still stick with my current one.

[C17H23NO2]

What you could do is having everything at stock settings , xmp first - try gaming. 
works? good, cpu overclock next - try some cpu benchmark + gaming. 

works? good, gpu overclock next. 
If you run into a problem of whatever kind you know what it is cause you changed everything bit by bit. 
If gaming with xmp on works but then the pc fails after overclocking, for example, you then gotta find out what's the cause. Yes probably cpu OC , but too high temperature ? not enough voltage? generally unstable cause too high frequency? 
But you can also start with cpu oc and then xmp to see how that works together, and if Ram is a problem. If you know what I mean.  
Still the 3600 as a present is a nice thing to look forward to.  

[salatji]

58 minutes ago, C17H23NO2 said:

If gaming with xmp on works but then the pc fails after overclocking, for example, you then gotta find out what's the cause. Yes probably cpu OC , but too high temperature ? not enough voltage? generally unstable cause too high frequency? 

Well, is the 1st gen Ryzen really that bad in OCing? I think I tried 3.8 or 3.9 at like 1.4 volts or maybe even the official "save voltage" quoted by ryzen with 1.45 and it still failed...

[Tristerin]

2 hours ago, salatji said:

It's 2x8. Does XMP-profiles affect the results of CPU overclocking? Or are memory and CPU overclocks independent?

 

 

About the GPU-overclock: So if I tone down the OCs, they might still work and I just have to test around?

Yes absolutely does, because your CPU memory controller only guarantees 2667mhz, anything over that, is luck, and potentially unstable.

 

Start your OC with your CPU then move to memory for Ryzen gen 1 in my experience (I have 2 1700s I play with).  Put the CPU to max turbo boost settings, and then start with a voltage of 1.375vCore.  If it boots, and passes CineR20 and a Firestrike run you can either choose to increase MHZ, or reduce voltage.  Id increase CPU by 100mhz increments, until unstable and either decide if Im willing to move to 1.4vCore or not.

 

After you get a stable CPU OC, then work on your RAM.  My system is 99% stable, but sometimes it crashes, as I fine tune the 3600mhz Im able to push on my memory controller.

 

 

[salatji]

36 minutes ago, Tristerin said:

Yes absolutely does, because your CPU memory controller only guarantees 2667mhz, anything over that, is luck, and potentially unstable.

 

Start your OC with your CPU then move to memory for Ryzen gen 1 in my experience (I have 2 1700s I play with).  Put the CPU to max turbo boost settings, and then start with a voltage of 1.375vCore.  If it boots, and passes CineR20 and a Firestrike run you can either choose to increase MHZ, or reduce voltage.  Id increase CPU by 100mhz increments, until unstable and either decide if Im willing to move to 1.4vCore or not.

 

After you get a stable CPU OC, then work on your RAM.  My system is 99% stable, but sometimes it crashes, as I fine tune the 3600mhz Im able to push on my memory controller.

 

 

ok, victim found (you seem to have some experience with ryzen OCing

 

Questions:

 

- is it possible that a previously badly mounted CPU cooler could've caused my previous OCs to fail? I replaced it with a different one that was imo easier to mount and apparently for the first time I instantly was able to run my 1600x at 3.9 ghz with 1.4V through a R20 without crashing

 

- Why is HWmonitor showing that the CPU-voltage is at like 1.35 while running R20 even though I did put 1.4 in the bios? Is this a concern/a issue with my PSU or perfectly normal?

 

- if you have a MSI board: how exactly does the offset-OC work? I don'T quite understand what values I have to insert (currently looking  that up, but if you know the answer, I would appreciate it. )

[Tristerin]

55 minutes ago, salatji said:

ok, victim found (you seem to have some experience with ryzen OCing

 

Questions:

 

- is it possible that a previously badly mounted CPU cooler could've caused my previous OCs to fail? I replaced it with a different one that was imo easier to mount and apparently for the first time I instantly was able to run my 1600x at 3.9 ghz with 1.4V through a R20 without crashing

 

- Why is HWmonitor showing that the CPU-voltage is at like 1.35 while running R20 even though I did put 1.4 in the bios? Is this a concern/a issue with my PSU or perfectly normal?

 

- if you have a MSI board: how exactly does the offset-OC work? I don'T quite understand what values I have to insert (currently looking  that up, but if you know the answer, I would appreciate it. )

1.) Yes CineR20 will crash me in about 2 seconds if I don't have a proper way to carry the heat away (thermal crash)

2.) Anything over 1.375 (iirc) can potentially cause silicone degredation in Ryzen first gen.  What is your Windows power plan?  Make sure its on Maximum Performance so that you don't fail a OC due to Windows thinking you need less juice.  It can fluctuate but when I put load on mine it hits my max.  Id say perfectly normal, get your Windows Power Plan corrected.

3.) Are you talking about SOC Offset?  I have ASUS and ASRock boards I use.  If its the SOC Offset (located within the same fields as OCing the CPU) there is a max, iirc, of 1.2v, anything over that will fry your Memory Controller on the CPU.  I have HEARD that you should put it at 1.15v for better OCing of your RAM, however I have left mine on Auto because so far so good for me.  Once I have time Im tempted to play more with this to get my 1% crashes down to 0% perfect stability.

 

3.1) If you are talking about CPU voltages more in depth options, you will need to speak with someone more fluent than I in overclocking, sorry.  Im actually quite a baby at this, but through trial and error I am starting to understand.

 

Ill summon @Zando Bob he may have something to say, typically knows something about everything.   For some reason it wont let me tag @nick name.  He also is very versed in Ryzen.

 

[Tristerin]

@nick name if you get this read my post above, summoning people who may know

[salatji]

11 minutes ago, Tristerin said:

Anything over 1.375 (iirc) can potentially cause silicone degredation in Ryzen first gen.

I'll leave it at 1.375 and 3899 mhz for now. Question about Aida 64: If I stresstest with it, at what temperature should I start to worry?

 

Also which one is the relevant temperature in aida: CPU or CPU Diode?

 

I'll try if I can run some games for now, if everything works, I'll try do a aida64 over night and hope for the best

[Tristerin]

41 minutes ago, salatji said:

I'll leave it at 1.375 and 3899 mhz for now. Question about Aida 64: If I stresstest with it, at what temperature should I start to worry?

 

Also which one is the relevant temperature in aida: CPU or CPU Diode?

 

I'll try if I can run some games for now, if everything works, I'll try do a aida64 over night and hope for the best

I don't use Aida often so cant speak to that.  I use HWMonitor to log my temps.  From what I have read about first Gen Ryzen, 95c max temp or more can cause degredation of the CPU.  Personally I don't like above 80c. 

[Zando_]

"Question 1: Can I use XMP for the memory (currently manages the 2933 xmp) if I manually OC the CPU? I am asking this because I did not even get to OC the 1600x to 3.8/3.9 GHz, it always kept crashing/failed to boot."

 

- Yes. XMP can give people issues sometimes, I never experienced any with my R5 1600 and G.Skill Ripjaws V or Patriot something RAM. Trident Z should play nice with Ryzen as well. If you're worried about your kit being problematic, just run it at bone stock speeds while you tweak the CPU. 

"Question 2: Is there a possibility that MSI Afterburner somehow manages to OC (and my 1070 seemed to run stable in kombuster with a OC of 150 mhz core and 350 mhz memory) and then fails to apply it in games? I tried to launch cities skylines after that, and at first I got in the menu at 10 fps and failed to load a map and later the game did not even launch properly. After restarting, everything seems to be fine again."

 

- Also yes. You can apply totally wack OCs that are unstable, once the GPU comes under load and tries to hit those it just crashes. I usually run Unigine Valley at like 1440p (enable DSR or just run it at 1080p, uncheck fullscreen so you can have apps on top of it) with MSI Afterburner on top. Then I slowly increase clocks, wait a minute or two to see if there's artifacting or a crash, then increase them again. Keep going till it crashes then revert back to the last stable OC, and let it run for 20 minutes or so, then boot up your favorite games as a final test. OCs can be stable in benchmarks but not in games. For example, I can push around 2050Mhz on my Radeon VII in both Valley and the 3DMark benches like FireStrike and Time Spy, but in my games it'll crash. 

Also, only ever change one thing. I'd usually recommend getting your highest stable core clock, then starting to tweak memory. 

 

1 hour ago, Tristerin said:

1.) Yes CineR20 will crash me in about 2 seconds if I don't have a proper way to carry the heat away (thermal crash)

2.) Anything over 1.375 (iirc) can potentially cause silicone degredation in Ryzen first gen.  What is your Windows power plan?  Make sure its on Maximum Performance so that you don't fail a OC due to Windows thinking you need less juice.  It can fluctuate but when I put load on mine it hits my max.  Id say perfectly normal, get your Windows Power Plan corrected.

3.) Are you talking about SOC Offset?  I have ASUS and ASRock boards I use.  If its the SOC Offset (located within the same fields as OCing the CPU) there is a max, iirc, of 1.2v, anything over that will fry your Memory Controller on the CPU.  I have HEARD that you should put it at 1.15v for better OCing of your RAM, however I have left mine on Auto because so far so good for me.  Once I have time Im tempted to play more with this to get my 1% crashes down to 0% perfect stability.

1.) CB20 runs the full bigg boy AVX workloads AFAIK, it'll crash on most unstable CPUs or ones with bad cooling, not quite as hard as Prime95 does though (small FFT is best for a 10 minute temp test, IIRC the blend option is for stability testing over a longer period). I often use it as a first test to see if my OCs are stable, since it's quick to open after boot, and usually crashes unstable OCs quickly. If it passes CB20, then I move on to Prime95 or Realbench and other stress/stability testing software, then games. (again, had my CPU be fine in benches then crash in games at those same clocks)

2.) I think max safe for Zen is 1.4v, it was 1.38 or 1.375v for Zen+, though other people say 1.4v is fine for both and yet more say to never go over 1.38v. I'd call 1.38v the cautious man's max, if you aren't too worried about your CPU and have good thermals, feel free to slap 1.4v into it. Defo wouldn't go over that, more voltage past 1.4 usually doesn't help anyways. My 2700X hit a pretty hard cap at 4.2Ghz, even 1.45v wouldn't get 4.25Ghz stable. My 1600 did 4Ghz stable at 1.4v, trying 4.1Ghz caused an instant crash. 

Max safe is always weird though, can never really get a solid number. For example, my Haswell-E i7 5820K: Intel's max is 1.35v, a lot of the community says they degrade at anything above 1.32v, some very cautious people say 1.3v max, and there's a heck ton who claim 1.4v to be fine with good cooling. 

 

3.) IDK much about the SOC offset, didn't mess with RAM when I had Ryzen. Don't push it on my X99 system either due to hearing about Haswell-E CPUs toasting their IMCs when pushing RAM overclocks. Only really pushed it on my X58 system, but that's a whole different beast. Also uses cheap af Xeons and DDR3 so if something toasts it's not really a big deal. 

Edited September 30, 2019 by Zando Bob
added more deets about my former Ryzens

[nick name]

I'm not sure I'd be much help with a first gen Ryzen.  Don't know what voltages to expect at which multipliers.  

 

I can say that I never recommend HWMonitor with Ryzen and you should instead use HWiNFO.  

[salatji]

16 minutes ago, Zando Bob said:

I think max safe for Zen is 1.4v

I mean mine crashed at 1.4 and 4GHz anyway, so I'll just leave it at that. Is Aida64 fine aswell? Or should I check with prime95?

 

I won't play around with offset.. while on auto, the bios told me that the CPU pulled 1.44V, so with 1.375 that's a lot lower, so I do think it won't matter if I lower the voltage even more when idle...

 

How do you set your fan curve? I have a Dark Rock 4, kinda maxing out at 75°C right now (I think I have a faily "slow" fan curve that kinda kicks in at those temps), but in Aida one of the two CPU-temps was closing in to 90ish, so I got a bit worried that I might have a wrong fan curve...

[Zando_]

1 minute ago, salatji said:

I mean mine crashed at 1.4 and 4GHz anyway, so I'll just leave it at that. Is Aida64 fine aswell? Or should I check with prime95?

 

I won't play around with offset.. while on auto, the bios told me that the CPU pulled 1.44V, so with 1.375 that's a lot lower, so I do think it won't matter if I lower the voltage even more when idle...

 

How do you set your fan curve? I have a Dark Rock 4, kinda maxing out at 75°C right now (I think I have a faily "slow" fan curve that kinda kicks in at those temps), but in Aida one of the two CPU-temps was closing in to 90ish, so I got a bit worried that I might have a wrong fan curve...

I ran it on a EVGA CLC 120 (120mm AIO) in an H200 case. Messed around with OCing my R5 1600 a bit before I shipped it off to a friend in need of a good PC (lowkey wish I hadn't given away my 980 Ti like that though, lol). I think it was hitting around the high 70s (I did have the fan on max and it wasn't an extended, few hour long test to see heat soak on the AIO), it wasn't really hot at all. Unless you got a really garbage chip, you should hit 3.8GHz at least though. 

[salatji]

1 hour ago, Zando Bob said:

I think it was hitting around the high 70s

In HWinfo the Tdie or the Tctl temperature @nick name ? Cuz the Tdie with 72° would be fine, but the Tctl did hit 92°C...

 

Google result told me that Tdie is the important temp, did I find the correct response? 

[Zando_]

11 minutes ago, salatji said:

In HWinfo the Tdie or the Tctl temperature? Cuz the Tdie with 72° would be fine, but the Tctl did hit 92°C...

 

Google result told me that Tdie is the important temp, did I find the correct response? 

I used HWMonitor, CPU core temps and CPU package temp. 

[Tristerin]

Thanks @Zando Bob and @nick name!  Appreciate your alls insight as always!

[nick name]

3 hours ago, salatji said:

In HWinfo the Tdie or the Tctl temperature @nick name ? Cuz the Tdie with 72° would be fine, but the Tctl did hit 92°C...

 

Google result told me that Tdie is the important temp, did I find the correct response? 

First gen adds to the Tctl and Tdie is what it actually is.