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

 

[gk1951]

Sorry to hear that Shmee. Hope it gets corrected.

 

[bigboxes]

Just ordered a Thermalright heatsink for the Sabrent M.2 SSD. We'll see what hardware comes with the heatsink.

 

[bigboxes]

 

[B-Riz]

I would have rather had a large NVMe drive myself. Unfortunately they're priced out of my budget.

The $250 asking price seamed fair, but I offered him $200 for it and he accepted my offer.

$250 is a good deal, I paid $346 for an open box 4TB 860 EVO.

 

[Kenmitch]

$250 is a good deal, I paid $346 for an open box 4TB 860 EVO.

Yep...But I guess you missed the part about accepting my $200 offer.

 

[B-Riz]

Yep...But I guess you missed the part about accepting my $200 offer.

Ha! An even better deal!

 

[Kenmitch]

Booted up my backup 5600x/6800 rig to play around with for a couple of days!

I'm currently testing out the new MSI ComboAM4PIV2 1.2.0.2 beta uEFI. I watched a video on using the curve optimizer and decided to play around with it and see how far I can push the 5600x. I was a little bit concerned about disabling the PBO limits until I compared the voltages, etc using HWiNFO64. I compared PBO limits set to Motherboard vs Disabled and Disabled used way less voltages and was very easy to stabilize. For comparison it was something like 1.263 vs 1.312 at the same clocks. I can't think of any reason why PBO limits being disabled would cause any harm.



I might have to play around with the negative offsets on the cores that are currently set at 20 to see if I can increase the multi-threaded clocks. I didn't think about it until I was tying this, but as I raised the Boost clock override I had to lower the offset on the other 2 cores to keep from erroring out. I think it's possible to maybe squeeze a little more performance out of multi-core loads if I can increase some more offsets....Hmm

The beta uEFI is stable, but I've found 3 quirks so far. Unable to boot up 1900 FCLK. HWiNFO64 doesn't read Core 0 T0 effective clocks. CTR 2.0 RC5 doesn't like the new uEFI version and instantly fails on stress testing with stock clocks or maxed out clocks.

Time Spy run to check the cpu clocks during gaming. Even beat my own top score for single card 5600x/6800 combo rigs. I still have to do some more testing to make sure clocks are stable, but so far so good.



It was a fun adventure!

 

[bigboxes]

Got the heatsink (8 inch heatpipe) for the Sabrent this evening. Probably install it tomorrow evening or Saturday morning. The new power supply is coming tomorrow. Looks like a fun Saturday for me!

 

[Magic Carpet]

Good to know about that improvement of 1.2.0.0. Also, an update here for my system, I am currently RMAing the 5900X. Dropped it off to ship to AMD yesterday afternoon. Reason was it was getting occasional random WHEA errors, event ID 18 and BSODs with the bugcheck 124. This was all stock, not even XMP loaded. So I am back on the 5800X for now.

Interesting. Although, I doubt the CPU was faulty. More likely the BIOS/voltages. The delta must be huge between different samples. I’d try the 5950x, if you can find it.

 

[Det0x]

So ive done with my initial testing for CTR 2.1 RC5 build 15 with automatic up and downclocking on the fly (just like PBO)

Comparing threadscaling in Cinebench R20 just like i did last time.
These were my previous results:

PBO CO benchmode: (ambient ~ 20 degrees)
Bios 3003, which have the best PBO CO boosting behavior of all asus bioses

• 1 thread = 662 points
• 2 threads = 1303 points
• 4 threads = 2444 points
• 6 threads = 3706 points
• 8 threads = 4887 points
• 10 threads = 5974 points
• 12 threads = 7022 points
• 14 threads = 7906 points
• 16 threads = 8645 points
• 20 threads = 9583 points
• ...seems like i didn't save 24 thread screenshot, but 105xx score
• 32 threads = 12238 points

CTR 2.04 hotfix: (ambient ~ 24 degrees)
Bios 3003, but dont matter since using CTR

• 1 thread = 652 points
• 2 threads = 1295 points
• 4 threads = 2525 points
• 6 threads = 3752 points
• 8 threads = 4979 points
• 10 threads = 6016 points
• 12 threads = 7171 points
• 14 threads = 8287 points
• 16 threads = 8831 points
• 20 threads = 9539 points
• 24 threads = 10217 points
• 28 threads = 11117 points
• 32 threads = 12032 points

LLC4 = upto 2% vdroop

PX high = from 1 to 2 threads @ 5050mhz
PX mid = from 3 to 4 threads @ 4950mhz
PX low = from 5 to 9 threads @ 4900mhz
P2 = from 10 to 20 threads @ 4800/4675mhz
P1 = from 21 to 32 threads @ 4700/4600mhz

Results from CTR 2.1 RC5
Latest bios 3501 for these runs, which i did all back-to-back (could gain a few points with restarts between runs)

• 1 thread = 668 points
• 2 threads = 1302 points
• 4 threads = 2528 points
• 6 threads = 3800 points
• 8 threads = 4999 points
• 10 threads = 6081 points
• 12 threads = 7187 points
• 14 threads = 8185 points
• 16 threads = 8963 points
• 20 threads = 9540 points
• 24 threads = 10031 points
• 28 threads = 11044 points
• 32 threads = 12064 points

LLC4 = upto 2% vdroop

PX high = from 1 to 2 threads @ 5050mhz
PX mid = from 3 to 4 threads @ 4950mhz
PX low = from 5 to 9 threads @ 4900mhz
P2 = from 10 to 20 threads @ 4775/4650mhz
P1 = from 21 to 32 threads @ 4700/4600mhz

Settings: (noteworthy: L3 latency is only 9.8ms and these settings survived 5 iteration of IBT very high which is a 270watt load at these speeds)


Full set of screenshots can be found here:

https://imgur.com/a/JmgrHWm

CTR 2.1 RC5 seems indeed to be working nicely even in a "medium-workload" as Cinebench, but it is in the "light-workloads" / games that will benefit the most from this update.

 

[bigboxes]

Today was M.2 heatsink day. Power supply delivery was rescheduled for Monday. I just had to make sure that I had room to secure the mounting screw. Other than rearranging the thermal tape, worrying about damaging it. I ran benchmarks using the Samsung Magician software. Used HWMonitor to monitor temps. Temps dropped from 60C to 50C under load. My 980 Pro is toasty at 73C under load. Going to have to check out the motherboard heatsink before trying another solution.

 

[JoeRambo]

The pigs flew today and I've been able to acquire this puppy, was starting to doubt they exist at all...

 

[blckgrffn]

I finally got a 5800x bought and installed.

What a pig! At stock settings it chewed right through my Noctua cooler! (referenced here: https://forums.anandtech.com/threads/noctua-14cm-fan-replacement-needs-moar-rgb.2578356/ )

Under 12 BOINC - Rosetta tasks it was chewing like 90% of it's amperage according to Ryzen master and marched right up to 90c and throttled. I wish I would have screen capped it, it was using so much power!

I went into the bios and enabled Eco Mode and XMP (which had been disabled before) and boom, fired it back up and everything was so much more sane. Power usage was sane, still saw a couple bounces to 4.6ghz+ while the cores were loading and still ~4.1 ghz with the 12 tasks.

I was just a bit floored. I mean, I have an ARGB fan (not a terrible or super cheap one) on there that had been super fine with the 3600...

Anyway, I was really glad I had read about eco mode so I could go find it.

 

[thigobr]

5800X is hot mainly because the die is so small and the power limits are the same as for the 5900X/5950X. But those CPUs have 2 chiplets to fill that power budget. It doesn't pull a lot of power though.

Playing with PBO limits you will be able to reach a better tradeoff between frequency and temperature.

 

[bigboxes]

Got the new power supply this evening. I'll probably swap it out this weekend. Seasonic is awesome.

 

[Kenmitch]

Flashed my 5900x rig with the MSI beta ComboAM4PIV2 1.2.0.2 to do some testing and optimize the performance. The GPU was pretty easy to whip in shape, but he 5900x has taken a lot of trial and error with the - offsets to get the best bang for buck clocks out of it. Using CoreCycler for testing clocks is time consuming! Using PBO limits disabled and some pretty aggressive - offsets on most of the cores gave me the best compromise of single vs multi-core clock speeds in the end.



The 6800 XT Taichi likes to overclock, but it's hit and miss with what settings one chooses in the control panel. I'm overclocked while undervolted to 1116mv's which gave me the best overall results. The current profile looks to be stable and is very consistent during the runs. The cpu clocks due tend to fluctuate a little depending on the temps which effect my overall score....I'm adding 3 more fans in pull on the radiator once they get here tomorrow.




The Taichi looked like it sagged when installed the conventional way due to the length and the way it was made so I ordered the Lian Li vertical GPU mount which looks nicer and gives lower temps to boot. I hate having RGB software running so I just have the G.Skill running at the default profile for now. The Corsair and Taichi RGB's are in default mode from a previous install with no software currently installed. It's nice the way that works!



I didn't notice the SSD is sitting crooked until I looked at the photo. I'll have to remount that when I install the new fans tomorrow.

 

[Dave3000]

Is there a way I can manually overclock my 5800x to a fixed 4.6GHz and a fixed 1.27v CPU vcore, but still have AVX2 power hungry apps stay within the official power limits and temperature limits of this CPU. I once overclocked it manually to 4.6GHz all-core at 1.27v for the CPU vcore and the CPU Package temperature instantly hit 106 C during Prime 95 v30.3 and I had to shutdown my PC immediately because I didn't want to risk damaging my CPU. I thought it would just thermal throttle at 90C by slowing down the CPU and lowering the CPU voltage but it didn't that time. In Cinebench 20 the CPU Package temperature was around 80C at those same settings. I used HWinfo64 to monitor CPU temperatures. Also if I don't manually overclock the CPU and just leave the CPU at stock settings will never go over 90C (usually between 86C-89C in Cinebench 20 and 65-72C in gaming, depending on ambient temperature) on me and will run automatically with less voltage and slower in Prime95 v30.3. I want it to run at a fixed multiplier and voltage but have still adjust the CPU multiplier and CPU voltage automatically and obey the 90C thermal threshold during AVX2 power-heavy apps such as Prime95 v30.3 to prevent the CPU from overheating during Prime95 v30.3 with AVX2. Also if I run Prime95 v30.3 at stock CPU settings, it will be in the mid 80s C and CPU voltage around 1.15v and CPU clocks at around 4.15 GHz.

 

[bigboxes]

Went ahead and installed the new power supply. I just hate shutting down my machine.

You guys know what I'm talking about.

You can see the difference between the older power supply and the new one. The Prime is just a tad bit longer.

The Seasonic built this PSU with the markings printed reversed from the Platinum one. It is orientated for having the fan on the underside and be able to read the ports and the backside. That's better for me.

I was able to just leave the cables in place. I cleaned them and straightened them prior to installation in the new rig so it's no big deal. I have the Prime 1000W version in my server. I have a complete set of power cables from that one too. I use custom ones instead. So, I will have no problem finding cabling and ports to plug them into if need be.

 

[Magic Carpet]

The pigs flew today and I've been able to acquire this puppy, was starting to doubt theyView attachment 43642 exist at all...

😍 What motherboard your planning to use with this unicorn?

 

[B-Riz]

I finally got a 5800x bought and installed.

What a pig! At stock settings it chewed right through my Noctua cooler! (referenced here: https://forums.anandtech.com/threads/noctua-14cm-fan-replacement-needs-moar-rgb.2578356/ )

Under 12 BOINC - Rosetta tasks it was chewing like 90% of it's amperage according to Ryzen master and marched right up to 90c and throttled. I wish I would have screen capped it, it was using so much power!

I went into the bios and enabled Eco Mode and XMP (which had been disabled before) and boom, fired it back up and everything was so much more sane. Power usage was sane, still saw a couple bounces to 4.6ghz+ while the cores were loading and still ~4.1 ghz with the 12 tasks.

I was just a bit floored. I mean, I have an ARGB fan (not a terrible or super cheap one) on there that had been super fine with the 3600...

Anyway, I was really glad I had read about eco mode so I could go find it.

Is there a way I can manually overclock my 5800x to a fixed 4.6GHz and a fixed 1.27v CPU vcore, but still have AVX2 power hungry apps stay within the official power limits and temperature limits of this CPU. I once overclocked it manually to 4.6GHz all-core at 1.27v for the CPU vcore and the CPU Package temperature instantly hit 106 C during Prime 95 v30.3 and I had to shutdown my PC immediately because I didn't want to risk damaging my CPU. I thought it would just thermal throttle at 90C by slowing down the CPU and lowering the CPU voltage but it didn't that time. In Cinebench 20 the CPU Package temperature was around 80C at those same settings. I used HWinfo64 to monitor CPU temperatures. Also if I don't manually overclock the CPU and just leave the CPU at stock settings will never go over 90C (usually between 86C-89C in Cinebench 20 and 65-72C in gaming, depending on ambient temperature) on me and will run automatically with less voltage and slower in Prime95 v30.3. I want it to run at a fixed multiplier and voltage but have still adjust the CPU multiplier and CPU voltage automatically and obey the 90C thermal threshold during AVX2 power-heavy apps such as Prime95 v30.3 to prevent the CPU from overheating during Prime95 v30.3 with AVX2. Also if I run Prime95 v30.3 at stock CPU settings, it will be in the mid 80s C and CPU voltage around 1.15v and CPU clocks at around 4.15 GHz.

Try setting your PPT to ~120 and the other two settings to 0 in the AMD OC BIOS settings; that has been the easiest way to back off my 5800X from the "high" temps; 80c down to 76c or less, no loss in perf.

 

[JoeRambo]

😍 What motherboard your planning to use with this unicorn?

X570 Gigabyte Aorus Pro, waiting for the dust to settle on F33h BIOS with Agesa 1.2.0.2 before i jump. Complicating factor is the 64GB of DRAM machine needs, so i won't be breaking OC records for sure.
DRAM is 2 kits of F4-3200C14D-32GTZ , so quad dual rank DIMMs total.

 

[Magic Carpet]

X570 Gigabyte Aorus Pro, waiting for the dust to settle on F33h BIOS with Agesa 1.2.0.2 before i jump. Complicating factor is the 64GB of DRAM machine needs, so i won't be breaking OC records for sure.
DRAM is 2 kits of F4-3200C14D-32GTZ , so quad dual rank DIMMs total.

Awesome. Curious if the difference between dual rank and quad dual rank would be significant in your workloads. You’d be the first to bench 5950x in this scenario 😍

 

[Noid]

Is there a way I can manually overclock my 5800x to a fixed 4.6GHz and a fixed 1.27v CPU vcore, but still have AVX2 power hungry apps stay within the official power limits and temperature limits of this CPU.

Same issue with my BIOS.
I can limit EDC ... but not PPT with " all core"
( "All core" PPT limit is ignored, and thermal protection is non-existing )

Thru ALOT of testing, setting LLC Mode to 3 ( power plan ) keeps mine around 90c using P95.
( doing this limits your overclock - I use LLC mode 3 for benching ( p95 specifically - and the handbrake benchmark in this forum ) - LLC mode 1 for daily operation ) ( my day - to - day stuff is not CPU intensive )

Curious about your P95 behavior under different multipliers ....
Mine over-heats ~220+ PPT at ~13 minutes - for one minute - on multi 45.5 ....
... then drops back down to 180~160 PPT.
( cycling every 13 minutes )
However, the timing changes on different multipliers.
( IIRC - my 46 multi cycled at 14 min 30 seconds )
Does yours do the same ..?

So, try LLC mode 3 for P95, and anything else that has long CPU duration.
( for testing stability )

 

[B-Riz]

Same issue with my BIOS.
I can limit EDC ... but not PPT with " all core"
( "All core" PPT limit is ignored, and thermal protection is non-existing )

Thru ALOT of testing, setting LLC Mode to 3 ( power plan ) keeps mine around 90c using P95.
( doing this limits your overclock - I use LLC mode 3 for benching ( p95 specifically - and the handbrake benchmark in this forum ) - LLC mode 1 for daily operation ) ( my day - to - day stuff is not CPU intensive )

Curious about your P95 behavior under different multipliers ....
Mine over-heats ~220+ PPT at ~13 minutes - for one minute - on multi 45.5 ....
... then drops back down to 180~160 PPT.
( cycling every 13 minutes )
However, the timing changes on different multipliers.
( IIRC - my 46 multi cycled at 14 min 30 seconds )
Does yours do the same ..?

So, try LLC mode 3 for P95, and anything else that has long CPU duration.
( for testing stability )

What version of BIOS are you on?

I just put F33j on the X570 Master with the 3900X, 120W PPT, seems ok so far.

I have not setup the B550 / 5800X yet.

 

[blckgrffn]

So last night I cycled back to the main rig in my sig as I have been building a number of PCs and using my gaming/workbench to set them up. It was maybe off for four days?

Cold boot problems.

Sat their forever not posting, so I hit reset, took another minute while my heart sank and it finally posted but RGB was reset and I knew the BIOS had reset itself after a failed boot.

Booted up fine, then I rebooted and changed these things back:

CPU OC - 65W
XMP - On
RGB (Important!) - Reset to Radeon Red

I dug around a bit and saw that the PPT OC (pretty sure) was set to "advanced" and had a ton of values set. I backed it down to "enabled" as I wasn't sure turning to disabled would completely remove the smart boost features. I was surprised that it had so much data in there.

In any case, hoping that it works fine now AND there will be a non "a" bios drop with the USB fix soon. I used three of these boards recently (like, this week) updated all their bios and didn't see this issue. Of course "unplugging and let sit for several days before using again" is an extremely annoying scenario to test.