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

 

[B-Riz]

Grabbed a used 5800X from the FS / FT Forum.

Right now, it is in an Asus Prime X470-Pro, Gammaxx 400, AS Ceramique 2, BIOS 5837, chipset driver 2.13.27.501, hwInfo 6.42, Win10 1903 (needs updated), balanced power plan.

Idles at ~ 26 * C, but is around 15 * F outside and is in the basement.

Reg Mode, Cinebench R20 peaks at ~84 * C multi, at ~4.5GHz. Single is ~58 * C and ~4.8GHz

ECO Mode, Cinebench R20 peaks at ~62 * C multi, at ~4.2GHz. Single is ~58 * C and ~4.8GHz

Reg Multi 6004 pts Single 622 pts

ECO Multi 5575 pts Single 619 pts

Probably just going to use it in ECO mode, unless doing something MT intensive.

 

[B-Riz]

I have a 5800x and I'm wondering if it is normal for it to be running at slightly over 4.1 GHz all-core during Prime95 v30.3 (small FFT test), and slightly over 4.5 GHz all-core during Cinebench R20 multicore test. Also the all-core CPU VID during Prime95 v30.3 (small FFT test) is lower than the all-core CPU VID during Cinebench R20 on my 5800x which may play a huge part for the higher CPU temperatures in Cinebench R20 than Prime95 v30.3 on my CPU.

I think you might be confused.
If you have " SET " your multiplier value + FSB --- you will have an " all core " overclock .

The description of the behavior you say looks like " Boost " mode.
Boost will vary in " Boost " score depending on the highest core recorded.
Boost mode frequency high scores vary in Benchmark tests.

Prime95 uses a " formula " to select a maximum FSB in Boost mode
An " all core " would override this ...

Temperature is test dependent also.
If your looking for a maximum temperature -- I recommend the
Handbrake benchmark in this forum.

This answer feels a bit off (yes, it is normal behavior for the CPU), OP was wanting to know why the max multi freq for the two programs was different.

prime95 or any software does not do anything to select a max FSB, the CPU responds to the workload and adjusts as needed.

Cinebench R20 and prime95 30.3 are different workloads and the CPU will clock differently.

prime95 Small FFT is a more intense test than Cinebench R20, so the CPU adjusts to keep it running in safe thermal limits.

The 5800X, 3900X, 3600, 1700 (RMA replacement), 1700 (launch cpu) I have and tested both behave the same way in these tests, and any Ryzen CPU should; adjust volts / clocks to keep in the thermal envelope.

 

[Noid]

This answer feels a bit off

Do you you know what an " All Core" overclock is ?
You said the same as I

Your opinion is noted .. with an ignore.

 

[repoman0]

Do you you know what an " All Core" overclock is ?
You said the same as I

Your opinion is noted .. with an ignore.

But he’s right ... he was pointing out that you said the below, which isn’t how it works.

Prime95 uses a " formula " to select a maximum FSB in Boost mode

 

[kwikgta]

My 5950x using PBO with curve optimizer.

 

[Det0x]

My 5950x using PBO with curve optimizer.
View attachment 39661View attachment 39661View attachment 39662View attachment 39663

With a 5950x you should disabled SMT in timespy (and only timespy) for a higher CPU score..

Your singlecore in cinebench seems maybe on the lowside for a 5950x.
Think you can tweak some there, maybe try to improve the cooling as boosting is very much depended on temperature with PBO CO overclock. And/or you can also either try to increase or decrease the CO offset on your 2 best cores on CCX0 and see what happen (i just use cpu-z bench as a quick test while adjusting settings as cinebench singlethread takes so long time)

Do you have global c-states disabled ? (disabled = lower singlethread boost by ~50-100mhz)

Multithread seems ~normal for a good PBO CO overclock.

*edit*

You can also check your real singlecore cinebench boosting with affinity as shown here:

Question - Zen 3 and CTR

I've just installed this and been playing with it a little bit. What are you thoughts on CTR and have you used it? The Diagnostic game me this so far. I've not done any of the tuning yet and curve optimizer is also turned off. The only thing on is PBO2 + 200Mhz...

forums.anandtech.com

 

[SK10H]

Got my 5800x finally to begin testing for a day, on Asus X570 plus bios 3405 using corsair h110i, ctr cinebench avx2 test hit 90C less than 1 min and terminated, can't really cool 142w on such a small die.
I watch the guide at

to quickly see what should be done as review doesn't discuss it in length.

CTR2.0 show my core as silver sample


Adjusting auto oc to 200 give me a max of 175 at 5025mhz with higher cppg rating, which every core is capable of hitting according to hwinfo.



With auto LLC, CTR warn of vdroop, so I am using level 2 at the moment only with CPU vcore at auto in BIOS.

The problem seem to be that two of the cores with no auto oc, no curve optimizer, disabled pbo, can't even pass p95 smallest fft at 2 threads by setting process affinity to the associated threads.

+200 test AVX2 p95 smallest fft, 1T ~4725, 2T ~4625

C01,2,4,5,7,8 pass a simple 0.5-1 hr avx2 1 or 2 thread affinity p95 test, but C03,06 fail pretty much immediately. The app simply crash without error which is great, at least it doesn't get BSOD this way.

C02 – T2/3 , -15 curve pass p95 smallest fft 0.5 hr 1/2 thread

C03 – T4/5, +5 2T 4550 fail, +10 2T 4480-4510 ok

C06 – T10/11 0 fail

So I actually have to set to +10 on C03 to make p95 happy by bending the frequency curve down when setting affinity to stress test a particular core at 2T level, will have to do the same for C06.

It looks like even though Windows Scheduler will assign low level tasks to my best core C02 C04 which I can do probably at -15 and more, but I have to add some positive evil volt to the curve on my worst core. So either Asus auto vcore is not enough or AMD latest QC isn't good as default volt/speed fail on some cores?

edit:
With LLC lvl2, +10 would still fail at C06. Setting LLC back to auto make it work. So it's LLC being too weak. +1 to the video, -1 to CTR, doesn't seem to be Asus/AMD fault.
Now I will have to start from the beginning with auto LLC. This is going to take a long time to play.
With auto LLC, weakest C03 is still failing without adding positive curve.

 

[kwikgta]

With a 5950x you should disabled SMT in timespy (and only timespy) for a higher CPU score..
View attachment 39665
Your singlecore in cinebench seems maybe on the lowside for a 5950x.
Think you can tweak some there, maybe try to improve the cooling as boosting is very much depended on temperature with PBO CO overclock. And/or you can also either try to increase or decrease the CO offset on your 2 best cores on CCX0 and see what happen (i just use cpu-z bench as a quick test while adjusting settings as cinebench singlethread takes so long time)

Do you have global c-states disabled ? (disabled = lower singlethread boost by ~50-100mhz)

Multithread seems ~normal for a good PBO CO overclock.

*edit*

You can also check your real singlecore cinebench boosting with affinity as shown here:

Question - Zen 3 and CTR

I've just installed this and been playing with it a little bit. What are you thoughts on CTR and have you used it? The Diagnostic game me this so far. I've not done any of the tuning yet and curve optimizer is also turned off. The only thing on is PBO2 + 200Mhz...

forums.anandtech.com

I'm all air cooled, using a Noctua DH-15 with 170cfm Silverstone fans, Thermaltake Core x5 case with 4/200mm fans, 2/140mm, and 1/120mm. I'm as far as I can go on air and have no desire to water cool

 

[Dave3000]

Is there a way to make a Ryzen 5000 series CPU to stay in boost mode the whole time but still vary the boost depending on CPU temperature and CPU wattage consumed? My 5800x frequently drops down to as low 3775 MHz during idle and that's on the Windows 10 High Performance power plan and the latest BIOS for my motherboard. On my i7-4930k, choosing the High Performance power plan in Windows 10 would prevent my 4930k from dropping down to base clock even during idle and would stay in turbo boost mode and idle would be using the single core turbo boost clock and the 2nd highest turbo bin but the same does not happen to my 5800x. Is this just the way Ryzen 5000 series is designed to run and I can't change it unless I manually set the CPU clock multiplier in the BIOS and lose speed in programs that are lightly threaded?

 

[scannall]

Is there a way to make a Ryzen 5000 series CPU to stay in boost mode the whole time but still vary the boost depending on CPU temperature and CPU wattage consumed? My 5800x frequently drops down to as low 3775 MHz during idle and that's on the Windows 10 High Performance power plan and the latest BIOS for my motherboard. On my i7-4930k, choosing the High Performance power plan in Windows 10 would prevent my 4930k from dropping down to base clock even during idle and would stay in turbo boost mode and idle would be using the single core turbo boost clock and the 2nd highest turbo bin but the same does not happen to my 5800x. Is this just the way Ryzen 5000 series is designed to run and I can't change it unless I manually set the CPU clock multiplier in the BIOS and lose speed in programs that are lightly threaded?

Eh, if my computer is idling not doing anything I don't care if it dropped down to 1 Ghz or lower.

 

[Dave3000]

Eh, if my computer is idling not doing anything I don't care if it dropped down to 1 Ghz or lower.

Well while playing X-Plane 11, it will frequently drop down to 3775MHz on all cores at the same time, and I occasionally see some drops in frames that last about 2 seconds when panning around the cockpit and then it smooths out during flight staying at least at 30 fps (I capped the frame rate to 30 for X-Plane 11, so maybe my 5800x doesn't need to run faster to maintain 30 fps during those instances). If I uncap the frame rate in the drivers and disable vsync in X-Plane 11 then my 5800x will see slightly above 4.8GHz on 1-4 cores at the same time while the other cores stay at slightly above 3.8GHz. I'm just wondering if those few seconds of stuttering during panning in the cockpit during flight is because possible latency of the cores going up to speed to meet the demand? I used HWiNFO64 to check the clocks while in X-Plane 11.

 

[B-Riz]

Is there a way to make a Ryzen 5000 series CPU to stay in boost mode the whole time but still vary the boost depending on CPU temperature and CPU wattage consumed? My 5800x frequently drops down to as low 3775 MHz during idle and that's on the Windows 10 High Performance power plan and the latest BIOS for my motherboard. On my i7-4930k, choosing the High Performance power plan in Windows 10 would prevent my 4930k from dropping down to base clock even during idle and would stay in turbo boost mode and idle would be using the single core turbo boost clock and the 2nd highest turbo bin but the same does not happen to my 5800x. Is this just the way Ryzen 5000 series is designed to run and I can't change it unless I manually set the CPU clock multiplier in the BIOS and lose speed in programs that are lightly threaded?

Well while playing X-Plane 11, it will frequently drop down to 3775MHz on all cores at the same time, and I occasionally see some drops in frames that last about 2 seconds when panning around the cockpit and then it smooths out during flight staying at least at 30 fps (I capped the frame rate to 30 for X-Plane 11, so maybe my 5800x doesn't need to run faster to maintain 30 fps during those instances). If I uncap the frame rate in the drivers and disable vsync in X-Plane 11 then my 5800x will see slightly above 4.8GHz on 1-4 cores at the same time while the other cores stay at slightly above 3.8GHz. I'm just wondering if those few seconds of stuttering during panning in the cockpit during flight is because possible latency of the cores going up to speed to meet the demand? I used HWiNFO64 to check the clocks while in X-Plane 11.

You can edit the Windows Power Plan to have 100% min and max state.

Have you tried disable V-Sync and set a frame rate target to a few frames above 30?

Are you using an NVME drive?

How much RAM?

 

[Det0x]

Been playing around with CTR2.1 beta2, but still haven't decided if this is really usable or not (stability wise).. These are my results so far:
Upto 6/6 cores/threads on CCX0 @ 5ghz effective clocks

 

[Kenmitch]

Been playing around with CTR2.1 beta2

Where did you find that version?

CTR2.0 is showing my 5600x as a golden sample....I had to up the load line calibration to eliminate the excessive vdrop message.



I'm still waiting to test my 5900x, but MSI is lagging on the 1.2.0.0 update for the MB.

 

[Det0x]

Where did you find that version?

CTR2.0 is showing my 5600x as a golden sample....I had to up the load line calibration to eliminate the excessive vdrop message.

View attachment 39792

I'm still waiting to test my 5900x, but MSI is lagging on the 1.2.0.0 update for the MB.

You have to subscribe to CTR patreon (pay) to get access..
He just release beta3 which seems much more stable



Not often you see Zen3 with 6 cores running at effective 5000mhz clocks
And/or 3 cores @ 5050mhz
This will be sick in low threaded games..

Default curve coefficients
CORE#1 14 CPPC 212
CORE#2 11 CPPC 190
CORE#3 12 CPPC 203
CORE#4 12 CPPC 199
CORE#5 12 CPPC 212
CORE#6 9 CPPC 185
CORE#7 12 CPPC 208
CORE#8 12 CPPC 194
CORE#9 6 CPPC 154
CORE#10 3 CPPC 163
CORE#11 4 CPPC 181
CORE#12 6 CPPC 149
CORE#13 3 CPPC 172
CORE#14 0 CPPC 158
CORE#15 3 CPPC 176
CORE#16 3 CPPC 167


CORES ORDER (from the best to the worst)
### CPPC VID FIT
1 С05 С01 С04
2 С01 С05 С06
3 С07 С07 C11
4 С03 С03 C13
5 С04 С04 С08
6 С08 С08 С02
7 С02 С02 С03
8 С06 С06 С05
9 C11 С09 С09
10 C15 C12 C14
11 C13 C11 C15
12 C16 C10 С07
13 C10 C15 C12
14 C14 C13 C16
15 С09 C16 С01
16 C12 C14 C10

 

[ashFTW]

Considering getting Lenovo P620 with Threadripper Pro 3995wx for processing large amounts of genomics data. Is the Zen3/Milan version of this chip due soon? Thanks!

 

[moinmoin]

Considering getting Lenovo P620 with Threadripper Pro 3995wx for processing large amounts of genomics data. Is the Zen3/Milan version of this chip due soon? Thanks!

Doesn't look like it. AMD still hasn't officially launched Milan likely due to scarcity since cloud providers are getting all chips as is. A new Threadripper gen comes way after (and the possibility is high that the Pro version will be exclusive again for some time).

 

[ashFTW]

Doesn't look like it. AMD still hasn't officially launched Milan likely due to scarcity since cloud providers are getting all chips as is. A new Threadripper gen comes way after (and the possibility is high that the Pro version will be exclusive again for some time).

ok thank you! I’ll go ahead and order now

 

[Golgatha]

Overland Park Micro Center has 5600X and 5800X in copious quantities.

They had around 10 5950X and 5900X yesterday (2/16) for like an hour (if that) as well.

 

[Det0x]

Have been running real (not aida64) memory benchmarks and pushing my 4x8gigs memory to the max, these are my results
Seems to be able to ~match or atleast come close to 10900k's @ 5500mhz, 5000mhz uncore and ddr 4600cl6 in memory intensive benchmarks.
What works in Aida64 dont don't necessary works in real world / games..

Tomb Raider 1080 medium cpu+memory bench = 261 average CPU Game fps
Free game on stream if anyone else want to try
Only interested in "CPU Game" score, so gpu dont matter here.

Far cry 5 1080 ultra cpu+memory bench = 202 average fps



All benchmarks performed on my old windows install with lots of bloatware... Think its gimping my scores. And my stock air cooler + stock bios(450w max) 3090 is also not helping.

 

[Kenmitch]

Tomb Raider 1080 medium cpu+memory bench = 261 average CPU Game fps
Free game on stream if anyone else want to try
Only interested in "CPU Game" score, so gpu dont matter here.

MB defaults on 5900x, 6800XT default clocks, Memory 3600 CL14. I went for the easy memory overclock, but haven't tried anything higher yet as MSI just dropped 1.2.0.0 on the 18th for my MB. Not sure what the CPU boost was during the run? 6800 XT has about 300Mhz of OC headroom currently left on the table.

 

[Justinus]

5950x at bone stock with 3800CL14 tuned timings. Also 261 avg cpu game.

 

[Dave3000]

A few days ago I started having issues with my 5800x running hotter than usual. I reapplied the thermal compound and reseated the heatsink onto the CPU. It only improved CPU full load temperatures by 2C, CPU temps now take longer to climb and are about 2C cooler during full load since doing that. My 5800x does hit 87C during the OCCT Linpack test during a 5-minute duration. It hits 88C in CB11.5 and CB20. It used to hit up to 84C in CB11.5, CB20, and OCCT Linpack (5-minute duration). I did update my BIOS about a week ago to 1804 for my Asus B550-F Strix (Wi-Fi) but did not notice an increase in CPU temperatures right away, so I don't think it's the BIOS. Before reapplying the thermal paste and reseating the heatsink onto the CPU, my 5800x would reach 90C in CB11.5 near the end of the 1st run. In ECO 65W mode, my 5800x hits 63C in CB20 instead of 88C but I lose MT performance using ECO mode and I don't get the full performance I paid for if I use ECO mode. About 5540 in ECO 65W mode and 6079 with stock settings in CB20 CPU benchmark. So basically, now my 5800x is hitting upper 80's in CB11.5 and CB20 and 87C in OCCT Linpack and CPU tests (5-minute duration) at stock settings. My CPU cooler is a Noctua U12A and I'm using the stock thermal paste that came with this CPU cooler. My case is a Fractal Meshify C with 2 120mm intakes and 1 120mm exhaust, I replaced the single stock intake with dual CoolerMaster 120mm fans and the stock exhuast with a CoolerMaster 120mm fan a few days after I built the system but that probably only improved CPU temps by 1C along with the temperatures of my GPU and CPU climbing slower but still hitting about the same temperatures as it did with the single stock intake and the single stock exaust fan configuration of my Fractual Meshify C case.

 

[IEC]

@Dave3000 Enable PBO2 and set your PPT to 105W (+/-). Depending on silicon lottery, you should find performance closer to stock but with at least 10C cooler load temperatures.

 

[B-Riz]

A few days ago I started having issues with my 5800x running hotter than usual. I reapplied the thermal compound and reseated the heatsink onto the CPU. It only improved CPU full load temperatures by 2C, CPU temps now take longer to climb and are about 2C cooler during full load since doing that. My 5800x does hit 87C during the OCCT Linpack test during a 5-minute duration. It hits 88C in CB11.5 and CB20. It used to hit up to 84C in CB11.5, CB20, and OCCT Linpack (5-minute duration). I did update my BIOS about a week ago to 1804 for my Asus B550-F Strix (Wi-Fi) but did not notice an increase in CPU temperatures right away, so I don't think it's the BIOS. Before reapplying the thermal paste and reseating the heatsink onto the CPU, my 5800x would reach 90C in CB11.5 near the end of the 1st run. In ECO 65W mode, my 5800x hits 63C in CB20 instead of 88C but I lose MT performance using ECO mode and I don't get the full performance I paid for if I use ECO mode. About 5540 in ECO 65W mode and 6079 with stock settings in CB20 CPU benchmark. So basically, now my 5800x is hitting upper 80's in CB11.5 and CB20 and 87C in OCCT Linpack and CPU tests (5-minute duration) at stock settings. My CPU cooler is a Noctua U12A and I'm using the stock thermal paste that came with this CPU cooler. My case is a Fractal Meshify C with 2 120mm intakes and 1 120mm exhaust, I replaced the single stock intake with dual CoolerMaster 120mm fans and the stock exhuast with a CoolerMaster 120mm fan a few days after I built the system but that probably only improved CPU temps by 1C along with the temperatures of my GPU and CPU climbing slower but still hitting about the same temperatures as it did with the single stock intake and the single stock exaust fan configuration of my Fractual Meshify C case.

Do you have any fans blowing from top down on to the HSF / VRMs? I have one doing that.

Am running 5837 (same AGESA as your BIOS) on the PRIME X470-PRO, right now, prime95 Small FFT, I am getting ~80*C with a Gammaxx 400 and Ceramique 2 with the 5800X.

Try prime95 small fft to see if it stays around 80*C; IMHO I think prime95 is the 'best' stress test for a cpu.