AMD Ryzen 5000 Builders Thread

[B-Riz]

	Cores Threads	Base Freq	Turbo Freq	L3 Cache	TDP	MSRP
Ryzen 9 5950X	16c/32t	3400	4900	64 MB	105 W	$799
Ryzen 9 5900X	12c/24t	3700	4800	64 MB	105 W	$549
Ryzen 7 5800X	8c/16t	3800	4700	32 MB	105 W	$449
Ryzen 5 5600X	6c/12t	3700	4600	32 MB	65 W	$299

5600X comes with cooler, the rest do not.

above from https://www.anandtech.com/show/1614...en-3-on-nov-5th-19-ipc-claims-best-gaming-cpu

 

[lightmanek]

I forgot the base clock on the 5900 is 3.7

Because my stock multi is 37x and I run 100.5MHz FSB.

 

[Hail The Brain Slug]

So it looks like OC'ing the CPU is pretty pointless for gaming performance (GN and HU reviews showing no difference, even reduction). Would one of the cheaper ASUS X570 boards still allow for tweaking the last bit out of PC3600 RAM?

That would be true, yes, for every motherboard out there except the new, yet to be released Crosshair VIII Dark Hero.

Asus have added some hardware to support a new feature called Dynamic OC Switcher that lets you set a current limit to switch between factory boost/PBO and your pre-set all core overclock. So for a lower power draw like gaming load you can have top boost blocks with PBO.

It gives the best of both worlds, it would seem.

I'm waiting for it.

 

[MrTeal]

PBO off, XMP On - 4.45GHz, 1.244V, 118.1W, 84.3°C
PBO off, XMP Off - 4.44GHz, 1.238V, 118.1W, 84.7°C

Not a whole lot of difference, though that test isn't particularly memory intensive. 85°C is ridiculous for stock operation with that cooling setup though.

 

[Rigg]

That would be true, yes, for every motherboard out there except the new, yet to be released Crosshair VIII Dark Hero.

Asus have added some hardware to support a new feature called Dynamic OC Switcher that lets you set a current limit to switch between factory boost/PBO and your pre-set all core overclock. So for a lower power draw like gaming load you can have top boost blocks with PBO.

It gives the best of both worlds, it would seem.

I'm waiting for it.

Do you have a link that that points to this being a hardware feature? I was under the impression (assumption) that this was a software feature that could be implemented into bios updates for older motherboards. I was pretty excited at the prospect of seeing this feature pop up in a new BIOS for my C8H. That's a bummer if this requires a hardware implementation to work.

 

[Hail The Brain Slug]

Do you have a link that that points to this being a hardware feature? I was under the impression (assumption) that this was a software feature that could be implemented into bios updates for older motherboards. I was pretty excited at the prospect of seeing this feature pop up in a new BIOS for my C8H. That's a bummer if this requires a hardware implementation to work.

Nothing concrete except a few techtubers asking Asus if it will come to older boards and being told it cannot (and that they tried) and alluding to pcb/hardware changes.

I imagine they don't really want to share what they've done to keep it a secret sauce/exclusive feature.

 

[Rigg]

Nothing concrete except a few techtubers asking Asus if it will come to older boards and being told it cannot (and that they tried) and alluding to pcb/hardware changes.

I imagine they don't really want to share what they've done to keep it a secret sauce/exclusive feature.

Sad news. Any links to the vids/posts where the rumor's are being discussed? All I saw was the initial video from der8auer on launch day.

 

[Hail The Brain Slug]

Sad news. Any links to the vids/posts where the rumor's are being discussed? All I saw was the initial video from der8auer on launch day.

In the comments of this video, second comment (by Nicholas Graves) nested discussion from the youtuber.

I've also seen it discussed elsewhere by other youtuber/ reviewers but I don't recall where.

I think it's likely that they had to make hardware changes of some sort in order to support fast, reliable switching on the fly under load.

 

[Rigg]

In the comments of this video, second comment (by Nicholas Graves) nested discussion from the youtuber.

I've also seen it discussed elsewhere by other youtuber/ reviewers but I don't recall where.

I think it's likely that they had to make hardware changes of some sort in order to support fast, reliable switching on the fly under load.

Thanks for the link. That's disappointing. Here's to hoping they can offer an alternative way to squeeze more performance out of these things. I'm pretty sure I tried every possible way to induce throttling under heavy load on my 3900x and couldn't find anything that worked in OC mode. Seriously, it shouldn't be that hard to implement clock and voltage offsets that work along side a power and/or temp limit to throttle the CPU under high current loads.

 

[Kenmitch]

Got my hands on a 5600x today to do some playing around. I'm currently trying to see how high I can push my b-die and keep the voltage reasonable. Not bad results for 1.49v's.

I still have to do some more extensive memory stability testing, but so far it's looking very doable. I have not tried to push the clocks any higher yet. The Dram Calculator is of no help too me as it won't generate any results beyond 3866 MHz. I used the 3866MHz timings mostly and just bumped it up to 4000MHz to see what happens.

 

[JoeRambo]

Got my hands on a 5600x today to do some playing around. I'm currently trying to see how high I can push my b-die and keep the voltage reasonable. Not bad results for 1.49v's.

Awesome results, for sure! Can you run Intel(R) Memory Latency Checker for us?

 

[Kenmitch]

Awesome results, for sure! Can you run Intel(R) Memory Latency Checker for us?

I'll look into it. I've never used that software before. Is it easy to do?

 

[JoeRambo]

I'll look into it. I've never used that software before. Is it easy to do?

yeah, download from Intel https://software.intel.com/content/www/us/en/develop/articles/intelr-memory-latency-checker.html
extract archive and run as admin in windows folder of that archive.

 

[Kenmitch]

yeah, download from Intel https://software.intel.com/content/www/us/en/develop/articles/intelr-memory-latency-checker.html
extract archive and run as admin in windows folder of that archive.

I downloaded, extracted, ran as admin, it does it's thing, but vanishes in the end. Where is the output? Do I have to do some command line stuff?

 

[JoeRambo]

I downloaded, extracted, ran as admin, it does it's thing, but vanishes in the end. Where is the output? Do I have to do some command line stuff?

To see full output -> you need to run "cmd" with admin rights ( find it in windows start menu, right click, run as admin)
then change directory with for example if you extracted it to "mlc" directory on your desktop => copy paste and run "%USERPROFILE%\Desktop \mlc\windows\mlc.exe"

then it will run, but output will stay in cmd window.

 

[Kenmitch]

To see full output -> you need to run "cmd" with admin rights ( find it in windows start menu, right click, run as admin)
then change directory with for example if you extracted it to "mlc" directory on your desktop => copy paste and run "%USERPROFILE%\Desktop \mlc\windows\mlc.exe"

then it will run, but output will stay in cmd window.

While waiting for a response I figured it out using the Windows Power Shell. I redid it using the CMD prompt as Admin in case the results were different.



I've never run it before so I have no idea if it's good, bad. or even ugly?

The above results were using the following. I did bump the memory voltage a tiny bit to 1.51v's just in case.



I was able to boot into Windows at 4200/2100 which looks to be beyond the memory controllers threshold for pain....Crackling sounds of death.

 

[GoodRevrnd]

You got to 4000 CL15 stable? Wow

 

[Kenmitch]

You got to 4000 CL15 stable? Wow

Not 100% tested for stability yet, but CL16's probably worse case. I saved the profile and moved on to do some further testing to see how strong the memory controller is.

I'm currently playing around with 4133MHz, but haven't even tried any tweaking of it yet. I just used the Try It feature in the uEFI to see what would happen. I'm going to see if I can monkey around with the timings, but it's a ??? as the Dram Calculator doesn't go that far.



Wish me luck....I'll probably need it. 2066MHz look to be my FCLK limit.

 

[JoeRambo]

View attachment 33886

I've never run it before so I have no idea if it's good, bad. or even ugly?

Thanks a lot, i have asked for MLC cause it measures a lot of things. Results are real good for AMD system, compared to 3950x with 3600CL15 handtuned, it advanced at least 12ns and 10ns were shaved from HIT / HITM latencies. ( ignore loaded latencies, cause 6 cores with increasing delay simply can't generate 16 core load).

 

[richierich1212]

Got my hands on a 5600x today to do some playing around. I'm currently trying to see how high I can push my b-die and keep the voltage reasonable. Not bad results for 1.49v's.

I still have to do some more extensive memory stability testing, but so far it's looking very doable. I have not tried to push the clocks any higher yet. The Dram Calculator is of no help too me as it won't generate any results beyond 3866 MHz. I used the 3866MHz timings mostly and just bumped it up to 4000MHz to see what happens.

Ken, what sticks do you have?

 

[Det0x]

Thanks a lot, i have asked for MLC cause it measures a lot of things. Results are real good for AMD system, compared to 3950x with 3600CL15 handtuned, it advanced at least 12ns and 10ns were shaved from HIT / HITM latencies. ( ignore loaded latencies, cause 6 cores with increasing delay simply can't generate 16 core load).

View attachment 33898

Here is my 3950x

 

[Kenmitch]

Ken, what sticks do you have?

I have the G.Skill Flare-X 3200 CL14 16GB kit. I bought it just before AMD launched Zen 2.

 

[Hitman928]

PBO off, XMP On - 4.45GHz, 1.244V, 118.1W, 84.3°C
PBO off, XMP Off - 4.44GHz, 1.238V, 118.1W, 84.7°C

Not a whole lot of difference, though that test isn't particularly memory intensive. 85°C is ridiculous for stock operation with that cooling setup though.

Temp definitely seems excessive at those frequencies/voltages. Can you try running Cinebench r20 and checking temps/freq/voltages as well?

 

[Kenmitch]

It took some trial and error and not really a whole lot of tweaking to stabilize in the end....I had it really close in the beginning, but would get 1 or two random errors after a couple of passes. Stress testing memory is time consuming! About 5-10 minutes of tweaking settings vs hours of waiting for results!



The other two results in the chart are from my other memory profiles. 3600 CL14 and 3800 CL15 results.

 

[lightmanek]

Monster memory speed and latency, wow!
What volts are you feeding your sticks with?

 

[Kenmitch]

Monster memory speed and latency, wow!
What volts are you feeding your sticks with?

I only pump 1.5v's into my b-die. It's probably the safest high memory clock my 5600x will do. My 5600x FCLK craps out at 2100MHz, but 2066MHz might be doable if I get the urge to tinker around with it some more. I did some initial testing, but decided to go back to 4000MHz.

Maybe I'll revisit it? I was just looking at latency and bandwidth and decided it might not be worth the effort in the end. Having a buffer zone on the FCLK limit was more of a concern then possibly a max higher memory clock.