Ryzen Owners Zen Garden

[Count von Schwalbe]

No warning shutdowns. PSU seems fine, according to HWiNFO. New PC, less than a year old. (I think?)

Update: Kernel event ID 41. I suspect my older UPS.

 

[Mussels]

View attachment 313344



No warning shutdowns. PSU seems fine, according to HWiNFO. New PC, less than a year old. (I think?)

Update: Kernel event ID 41. I suspect my older UPS.

software errors are generally useless, since that could be triggered by a toddler playing with a fork in your fusebox - what's the actual symptoms?

 

[R-T-B]

software errors are generally useless, since that could be triggered by a toddler playing with a fork in your fusebox - what's the actual symptoms?

Also check your fusebox for toddler corpses.

 

[ViperXTR]

My other ram stick broke the other day -__-
I was getting bunch of CTD and BSOD lately and i was initially thinking it was maybe my CPU or GPU undervolt lately or my new SSD, Set them to stock and still same issues. Then i went and tried to use a ram stress test and it imnmediately cracked, same with using windows memory test, just a few secs of test and it failed. The other mem stick is running fine but when i swap it with the other, it consistently fail (can't even boot properly now). Guess time for rma
Though i was thinking now that some memory prices has dropped since i last got this (pandemic), thinking of getting a lower latency one (currently using CL18 3600mhz) vs CL16 3600Mhz, is it worth the price for Zen3?

 

[Mussels]

My other ram stick broke the other day -__-
I was getting bunch of CTD and BSOD lately and i was initially thinking it was maybe my CPU or GPU undervolt lately or my new SSD, Set them to stock and still same issues. Then i went and tried to use a ram stress test and it imnmediately cracked, same with using windows memory test, just a few secs of test and it failed. The other mem stick is running fine but when i swap it with the other, it consistently fail (can't even boot properly now). Guess time for rma
Though i was thinking now that some memory prices has dropped since i last got this (pandemic), thinking of getting a lower latency one (currently using CL18 3600mhz) vs CL16 3600Mhz, is it worth the price for Zen3?

Rare to have RAM actually die, but it does happen. I've only had one DDR4 stick die on me, ever - and it was the first one i ever bought, ironically.

In my testing C16 vs C18 isn't much of a difference - the RAM is so cheap now that you'd want to check carefully what you're getting, getting four ranks at C18 will smash two ranks at C14, let alone 16.
At 3200 when ryzen couldnt clock higher the latency made a larger difference, but 3200C14 and 3600 C16 are basically equal - and nothing says you can't try and tune the timings down.


Keeping in mind that modern corsair LPX is fine on ryzen (it was the 000 stuff with issues), the price differences here in Au are tiny to get 3600 C16

I bought this, turned out to be single rank - great if you want four of them, sad for my ITX system with two slots.

 

[ViperXTR]

Rare to have RAM actually die, but it does happen. I've only had one DDR4 stick die on me, ever - and it was the first one i ever bought, ironically.

In my testing C16 vs C18 isn't much of a difference - the RAM is so cheap now that you'd want to check carefully what you're getting, getting four ranks at C18 will smash two ranks at C14, let alone 16.
At 3200 when ryzen couldnt clock higher the latency made a larger difference, but 3200C14 and 3600 C16 are basically equal - and nothing says you can't try and tune the timings down.


Keeping in mind that modern corsair LPX is fine on ryzen (it was the 000 stuff with issues), the price differences here in Au are tiny to get 3600 C16
View attachment 313435

I bought this, turned out to be single rank - great if you want four of them, sad for my ITX system with two slots.
View attachment 313436

Yeah, rare for ram to die on me as well, though i remember one dying on me but that was the DDR2 days, if anything it's the motherboard and not the ram. Maybe it's possible that the cause is i oc from 3600 to 3733mhz with added .1voltage i think for around 2 years if it's life? I dunno.
Also, that g.skill pic you posted is the exact ones im using, but i remember seeing it on a tool that says it's dual rank?
Edit:
Says dual rank on CPU-Z

 

[Count von Schwalbe]

software errors are generally useless, since that could be triggered by a toddler playing with a fork in your fusebox - what's the actual symptoms?

Sudden unexpected shutdown without warning. PSU rail voltages look ok and I wasn't doing anything heavy in terms of load. I would expect it to be the UPS, but the screens also plugged into the UPS didn't shut down. Possibly the fault in power was sufficiently small so as not to affect the screens.

 

[R-T-B]

Sudden unexpected shutdown without warning. PSU rail voltages look ok and I wasn't doing anything heavy in terms of load. I would expect it to be the UPS, but the screens also plugged into the UPS didn't shut down. Possibly the fault in power was sufficiently small so as not to affect the screens.

FWIW I had a UPS die in such a fashion once. It's not common but it can happen. You might also just try replacing the SLA batteries if possible.

 

[ViperXTR]

Yeah, rare for ram to die on me as well, though i remember one dying on me but that was the DDR2 days, if anything it's the motherboard and not the ram. Maybe it's possible that the cause is i oc from 3600 to 3733mhz with added .1voltage i think for around 2 years if it's life? I dunno.
Also, that g.skill pic you posted is the exact ones im using, but i remember seeing it on a tool that says it's dual rank?
Edit:
Says dual rank on CPU-Z
View attachment 313483

got new kit of ram, CL16, booted and working fine now. Yup its that one broken stick

 

[Nordic]

I am glad I did upgrade from a 5800x3d to a 7800x3d. In my performance testing, I had performance improve between ~15% and 50% depending on where I was in the game. Minimum FPS is now no lower than 30 and average fps is no lower than 60 in the worst performing locations. Overall average FPS is now high enough that I can forget about it most of the time.

I might upgrade to the 8800x3d depending on how it performs on release. I expect an incremental upgrade. That is a long while off though.

Some of you might enjoy seeing this graph showing FPS, CPU load, GPU load, and sensor data. Overclocking my 7800x3d shows with an average 5040mhz while playing the game. I could show an area with worse performance but this graph is fairly representative of overall performance.

 

[AusWolf]

Some of you might enjoy seeing this graph showing FPS, CPU load, GPU load, and sensor data. Overclocking my 7800x3d shows with an average 5040mhz while playing the game. I could show an area with worse performance but this graph is fairly representative of overall performance.

How is that overclocking? The max PBO on the 7800X3D is 5050 MHz.

 

[Nordic]

How is that overclocking? The max PBO on the 7800X3D is 5050 MHz.

Without messing with curve optimizer settings, I would have about 4600mhz average while gaming. You can get technical on semantics but my performance is better than stock.

 

[AusWolf]

Without messing with curve optimizer settings, I would have about 4600mhz average while gaming. You can get technical on semantics but my performance is better than stock.

Okay, but is your gaming clock higher now? I know you said 5040 MHz average, but that could mean clock on only 1 core, same as stock. I'm not aiming to be pedantic, just trying to understand what you mean by "overclocking" a locked chip.

 

[Nordic]

Okay, but is your gaming clock higher now? I know you said 5040 MHz average, but that could mean clock on only 1 core, same as stock. I'm not aiming to be pedantic, just trying to understand what you mean by "overclocking" a locked chip.

The average clock speed is higher during gaming by approximately 500mhz. The software and screenshot used is not the only means I have measured that improvement. The 5040mhz reported by CapFrameX is an "CPU max clock" so I assume it is some sort of average of peak core speed. The reported max clock speed is 5050mhz which isn't informative at all because it only reflects one core. The game is multithreaded and does benefit from multithreaded improvements more than single core for average FPS. I have recorded a similar 500mhz clocks peed increase in single core and multithreaded workloads in benchmarks like Cinebench or OCCT.

Curve Optimizer effectively is a mild undervolt to decrease temperature so that the CPU will push clock speeds higher. I personally view this as overclocking, at least with modern hardware. Overclocking is not the same as it was ten years ago. The max clock speed listed by AMD and Intel is a theoretical speed if the conditions are right. The best cooling isn't enough to achieve the listed max clock speed. One has to tinker with Curve Optimizer on AMD cpu's to create the conditions to get closer to that clock speed. Stability testing is necessary. If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck. This is what overclocking looks like today even if it might not meet the technical definition from years passed. Skatterbench would also describe this as overclocking. "The traditional overclocking approach for AMD Ryzen CPUs is to set a negative curve optimizer."

Are you following me now?

I wish I purchased a motherboard with eclk options. My CPU has headroom to overclock further. I might even be able to reach 5400mhz single core stable with a different motherboard.

 

[AusWolf]

The average clock speed is higher during gaming by approximately 500mhz. The software and screenshot used is not the only means I have measured that improvement. The 5040mhz reported by CapFrameX is an "CPU max clock" so I assume it is some sort of average of peak core speed. The reported max clock speed is 5050mhz which isn't informative at all because it only reflects one core. The game is multithreaded and does benefit from multithreaded improvements more than single core for average FPS. I have recorded a similar 500mhz clocks peed increase in single core and multithreaded workloads in benchmarks like Cinebench or OCCT.

Curve Optimizer effectively is a mild undervolt to decrease temperature so that the CPU will push clock speeds higher. I personally view this as overclocking, at least with modern hardware. Overclocking is not the same as it was ten years ago. The max clock speed listed by AMD and Intel is a theoretical speed if the conditions are right. The best cooling isn't enough to achieve the listed max clock speed. One has to tinker with Curve Optimizer on AMD cpu's to create the conditions to get closer to that clock speed. Stability testing is necessary. If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck. This is what overclocking looks like today even if it might not meet the technical definition from years passed. Skatterbench would also describe this as overclocking. "The traditional overclocking approach for AMD Ryzen CPUs is to set a negative curve optimizer."

Are you following me now?

I wish I purchased a motherboard with eclk options. My CPU has headroom to overclock further. I might even be able to reach 5400mhz single core stable with a different motherboard.

I don't know. "Slap the biggest cooler on it that fits in your case" is my definition of overclocking modern CPUs.

My 7800X3D under the Dark Rock 4 can run at it's maximum boost at all times because it never reaches its temperature limit. That's good enough for me.

 

[Count von Schwalbe]

I don't know. "Slap the biggest cooler on it that fits in your case" is my definition of overclocking modern CPUs.

My 7800X3D under the Dark Rock 4 can run at it's maximum boost at all times because it never reaches its temperature limit. That's good enough for me.

Does it have direct contact heatpipes or is it a baseplate?

There are a lot of varying models...

 

[freeagent]

Do PPT/TDC/EDC changes do anything with Zen4 X3D?

 

[Nordic]

I don't know. "Slap the biggest cooler on it that fits in your case" is my definition of overclocking modern CPUs.

My 7800X3D under the Dark Rock 4 can run at it's maximum boost at all times because it never reaches its temperature limit. That's good enough for me.

I have custom water cooling and at stock mine would run about 4600mhz on all cores. With CO optimization I pretty much get 5050 in most situations.

Do PPT/TDC/EDC changes do anything with Zen4 X3D?

Increasing each value provided no benefit in my testing. The CPU never limited by those values. Decreasing them can improve make your CPU more efficient not unlike AM4.

 

[freeagent]

Increasing each value provided no benefit in my testing.

Boooo.

 

[AusWolf]

Does it have direct contact heatpipes or is it a baseplate?

There are a lot of varying models...

It has a normal base plate. With Ryzen, it's probably better as it can spread the heat more evenly (I guess).

I have custom water cooling and at stock mine would run about 4600mhz on all cores.

How? Mine does 4800 MHz all-core and 5 GHz single-core at stock.

Increasing each value provided no benefit in my testing. The CPU never limited by those values. Decreasing them can improve make your CPU more efficient not unlike AM4.

Yep. The 7800X3D comes with a 162 W PPT by default, but only eats about 80 W max. That's also why I'm saying one just needs a big-ass cooler and call it a day.

 

[Nordic]

How? Mine does 4800 MHz all-core and 5 GHz single-core at stock.

I wouldn't know how to answer that. I have quality paste applied well. Maybe a slight manufacturing difference in ihs.

 

[AusWolf]

I wouldn't know how to answer that. I have quality paste applied well. Maybe a slight manufacturing difference in ihs.

Does it reach 89 °C at stock? If so, at what power consumption?

 

[RJARRRPCGP]

My other ram stick broke the other day -__-
I was getting bunch of CTD and BSOD lately

Did Windows start saying "PAGE_FAULT_IN_NON_PAGED_AREA"? IIRC, that's a classic one for losing RAM stability. (Whether it's the RAM, the RAM subsystem or of course one of the RAM slots)

 

[Nordic]

Does it reach 89 °C at stock? If so, at what power consumption?

I would have to go back to stock to test that. I don't see the point.

 

[tabascosauz]

The average clock speed is higher during gaming by approximately 500mhz. The software and screenshot used is not the only means I have measured that improvement. The 5040mhz reported by CapFrameX is an "CPU max clock" so I assume it is some sort of average of peak core speed. The reported max clock speed is 5050mhz which isn't informative at all because it only reflects one core. The game is multithreaded and does benefit from multithreaded improvements more than single core for average FPS. I have recorded a similar 500mhz clocks peed increase in single core and multithreaded workloads in benchmarks like Cinebench or OCCT.

Curve Optimizer effectively is a mild undervolt to decrease temperature so that the CPU will push clock speeds higher. I personally view this as overclocking, at least with modern hardware. Overclocking is not the same as it was ten years ago. The max clock speed listed by AMD and Intel is a theoretical speed if the conditions are right. The best cooling isn't enough to achieve the listed max clock speed. One has to tinker with Curve Optimizer on AMD cpu's to create the conditions to get closer to that clock speed. Stability testing is necessary. If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck. This is what overclocking looks like today even if it might not meet the technical definition from years passed. Skatterbench would also describe this as overclocking. "The traditional overclocking approach for AMD Ryzen CPUs is to set a negative curve optimizer."

Are you following me now?

I wish I purchased a motherboard with eclk options. My CPU has headroom to overclock further. I might even be able to reach 5400mhz single core stable with a different motherboard.

afaik that's still the same as "core clock' in most software that's not Ryzen Master or HWInfo. Ergo, pretty useless unless game is loading said core at close to 100% at all times, unrepresentative of actual core behaviour.

I'm sure a hefty CO offset as is possible on AM5 results in gains, I'm just not sure it's 400MHz good. 5050 Fmax is low for the product family in general and there should not be that much practical difference, and if there is then you should be seeing a similarly massive improvement in Cinebench ST.

Might just be piss poor performance at stock due to board issues/quirks that is magnifying the effect of CO.