AMD Ryzen 3000 Builders Thread

[PotatoWithEarsOnSide]

Updated my Ryzen Master since I didn't initially do so after updating BIOS.
From what I could tell, it makes little to no difference if you manually set a peak VCore since it won't abide by it anyhow...if CPUZ is anything to go by.

Talking of CPUZ. Seems as though reference benchmarks are way out, and are in need of updating. They list the 3700X as having single core of 511 when my 3600X (which also has same boost clocks) score 535.
Also, it doesn't fully list the correct AGESA details; shows only 1.0.3 but not the ABBA part.

 

[DisEnchantment]

My 3900X arrived.
Not after Mindfactory delayed it by two and half months
Not after UPS delayed by an additional 5 days
But my new rig is up.

But I was doing remote desktop to the new build and attempt to format the drive.
Turns out when I was running diskmgmt.msc I formatted my local drive instead of the remote drive.
Now running file recovery since last night but not so successful because it is an NVMe drive.


My 3900X goes to 80 degrees really fast and also comes down fast.
The fans are always screaming but I changed to a different manual profile. see pic
Chip is hitting 85 degrees frequently with H100i Pro RGB.
Chip is from Week 30 I guess.
Is it normal for the chip to be idling in the 60s ? I will try with the H110i 280mm rad if not enough with 240mm H100i
Also I use Noctua NT-H1 paste.
Are the temps on the higher side?

My chip crosses 4650+ MHz regularly on ABBA BIOS (but usually on the same two cores)
This chip is fast.
VMs very smooth. VC2019 good. Yocto good.
Games good.
I got 450+ fps in CSGO using 3900X/RX 5700XT at 1440p with medium settings. compared to 200ish with 1700X/RX480

Spoiler: PICTURES

 

[TheGiant]

My 3900X arrived.
Not after Mindfactory delayed it by two and half months
Not after UPS delayed by an additional 5 days
But my new rig is up.

But I was doing remote desktop to the new build and attempt to format the drive.
Turns out when I was running diskmgmt.msc I formatted my local drive instead of the remote drive.
Now running file recovery since last night but not so successful because it is an NVMe drive.


My 3900X goes to 80 degrees really fast and also comes down fast.
The fans are always screaming but I changed to a different manual profile. see pic
Chip is hitting 85 degrees frequently with H100i Pro RGB.
Chip is from Week 30 I guess.
Is it normal for the chip to be idling in the 60s ? I will try with the H110i 280mm rad if not enough with 240mm H100i
Also I use Noctua NT-H1 paste.
Are the temps on the higher side?

My chip crosses 4650+ MHz regularly on ABBA BIOS (but usually on the same two cores)
This chip is fast.
VMs very smooth. VC2019 good. Yocto good.
Games good.
I got 450+ fps in CSGO using 3900X/RX 5700XT at 1440p with medium settings. compared to 200ish with 1700X/RX480

/SPOILER]

you know- start norton commander- you have disk C on the left side and on the right side too...- why do I need 2 disks C? so I formatted one of them..

the temps are normal

about the FPS- now you see why noone compared the intel skylake lineup with the ryzen up to now- it is just bloody slower in high fps gaming

congratz on the new 3900X

 

[PotatoWithEarsOnSide]

Does seem like quite high idle temperature TBH. I'm at 32-36C with my 3600X and air cooling.
Temperature spikes aren't an issue though since these CPUs cool down incredibly quickly after they spike.

 

[EXCellR8]

Machine with 3700X is having a real hard time with OS updates now, so I think I might do a repair install tonight and see if it clears it up. I can't get Windows to boot when there is an update and need to manually restart in order to be able to log on and do anything. So much for hoping it would resolve itself...

I couldn't get Macrium to back up the booting volume, something about a Win 1903 bug, so I think I'll just hope for the best because it's really annoying.

 

[Topweasel]

It's important to note relatively ambient temp. But that seems a high idle temp. Running a H100i (with Noctua NF-A12x25 fans, 2 pulling from front). I get mid 30's idle or low usage. All core has to be pretty heavy for it to jump up to 80c.

2 important factors. My cooler is installed with an intel bracket (my board uses 115x cooler platform to save space) and I did 3 peas approach with the peas over each approximate die location, with the io pea being pretty large.

 

[DisEnchantment]

My ambient temperature is 20 degrees Celsius.
I run a H100i with two Corsair PWM SP140L with NH-T1 thermal paste.
Using default AM4 bracket.
But I have a very weak fan profile, I will tune it more aggressively to the limit of my noise tolerance.
Ryzen temperature spikes are very rapid, goes up and down really fast. When I switched the monitoring interval temperature it is moving all over the place.
I am concerned if I have a particularly leaky chip

 

[Markfw]

I would try setting vcore at 1.2 fixed and fan fixed as high as you can stand it. Also, Corsair fans are notoriously loud, I would consider replacing them.

You cpu speeds and average vcore look like mine. You also might try a fixed cpu speed if you use it a lot.

 

[Kenmitch]

Updated my Ryzen Master since I didn't initially do so after updating BIOS.
From what I could tell, it makes little to no difference if you manually set a peak VCore since it won't abide by it anyhow...if CPUZ is anything to go by.

Talking of CPUZ. Seems as though reference benchmarks are way out, and are in need of updating. They list the 3700X as having single core of 511 when my 3600X (which also has same boost clocks) score 535.
Also, it doesn't fully list the correct AGESA details; shows only 1.0.3 but not the ABBA part.

I haven't tried and locking the voltage in the uEFI and all core clocks in the ABBA uEFI yet.

I haven't noticed any abnormal vcore if I cap it in Master, but I'll take a look.

On another note....How do you like that cooler?

 

[PotatoWithEarsOnSide]

My own cooler is pretty good for £30.
A bit fiddly to mount the fans due to bizarre mounting clips, but nothing too annoying.

 

[DrMrLordX]

@DisEnchantment

If I'm reading your HWiNFO64 screenshots correctly, you got a temp of 86C with a current draw of 46 amps. That isn't right at all. Anyway, if your idle temps are 60C, that is WAY too high. Something is wrong. I would look at either reseating the cooler, or trying to turn DOWN the volts. Also consider installing the stock cooler for testing purposes. It'll give you a baseline.

 

[PotatoWithEarsOnSide]

Asus TUF GAMING X570-PLUS (WI-FI) Performance Results - UserBenchmark

www.userbenchmark.com

My best Userbench run from today.
Pretty impressed that I got my 3600X to score 99.2% which puts it in the 100th percentile.

edit: why settle for 99.2% when https://www.userbenchmark.com/UserRun/20436669 100% is possible.
https://cpu.userbenchmark.com/AMD-Ryzen-5-3600X/Rating/4041#Benchmarks which unofficially makes it the highest scoring 3600X benched with SMT enabled.
Another of my benches posted the highest score on the Extreme metric: https://www.userbenchmark.com/UserRun/20435942

 

[Despoiler]

[ PSA ]

I grabbed the ABBA BIOS for my Asus x370 board a couple days ago and have been testing. The good news is that it's clearly faster in benchmarks. It also runs SO much cooler. I have a custom WC loop for my CPU. With ABB BIOS I was peaking at 75C CInebenchR20 MT. ABBA I'm getting 54C. AMAZING! The bad news is that I still wasn't seeing 4400mhz on ST workloads of any kind. It was closer than ABB, but still topped out at 4382.

I've been doing a lot of memory latency benching with AIDA64 as well. I've notice that even with my Ryzen 1700 AIDA64 would report 99.8 BCLK instead of 100. I always thought this was weird because BCLK was auto in my BIOS. I did some research and the Internet is well....the Internet. Some people say it's a BIOS bug. Some people say it's normal because AIDA64 would read a lower value from spread spectrum. I guess it was time to find out what was really going on. Given that 99.8 was always what was reported across two chips I didn't buy the spread spectrum premise. I can buy a BIOS bug, but the alternative is that it was intentional by ASUS. I've changed just about every setting in the BIOS except one. I've always used their DOCP. This is their name for AMD compatibility settings for XMP. I set memory control to manual and re-applied all values. I fire up AIDA64 and low and behold the BCLK is reported as 100 and my chip hits 4400mhz. What I surmise is that ASUS in their amazing wisdom thought it was a good idea to slightly downclock the BCLK when using DOCP to avoid the stability issues early Ryzen might have with memory. That's why ASUS became know for being able to run 3200 memory without issue. They were essentially cheating. Currently they've either forgotten they did that and never normalized the BCLK or they've knowing continued as a strategy to mitigate stability issues. I don't know how many other ASUS boards do this, but mine does. If you have an ASUS board and AIDA64 is reporting anything lower than 100 BCLK stop using DOCP and manually set your memory. I would be curious to know the results from others.

 

[PotatoWithEarsOnSide]

Mine also reports 99.8MHz BCLK.
I'm not very bothered by it since I'm not having any issues hitting the advertised clocks, and my RAM had no issues going from 3200MHz CL16 16-18-18-36 to 3200MHz CL14 14-17-17-34 with FCLK of 1800MHz and no voltage changes.

 

[DrMrLordX]

I disabled spread spectrum in my UEFI and it still reads 99.8 MHz bclk sometimes.

 

[DisEnchantment]

I reseated the cooler head and reapplied some paste. I also tweaked the fan profile to be a bit more agressive in the lower temperature ranges and now idling at 40 degree celsius with 18 degrees celsius ambient.

I ran prime95 for 1 hour and the max temp was 83 degrees celsius.
I did not enable any fancy stuffs in the BIOS except XMP. I just need the build to be stable.

VMs are really responsive.

One thing I noticed is that QEMU did not really improve by much. In fact multiarch/qemu-user-static is much faster than qemu with A72 cores simulation. I dont know what is going wrong but the processor is underutilized when I run qemu natively, and the BogoMIPS it shows me is a paltry 125.
Running Ubuntu VMs in Windows 10 Pro.

I will try Native Ubuntu today and see if I have a similar issue.

 

[Despoiler]

I changed my memory settings from 3200 C14 to 3600 C16. The lowest latency with 3200 was 70.5ns and a lot of boots it was a couple ns higher. 3600 is 67ns every time. Same 1.35v too. Pretty happy with that.

 

[CHADBOGA]

I changed my memory settings from 3200 C14 to 3600 C16. The lowest latency with 3200 was 70.5ns and a lot of boots it was a couple ns higher. 3600 is 67ns every time. Same 1.35v too. Pretty happy with that.

Have you run any benchmarks to show the differences?

 

[Despoiler]

Have you run any benchmarks to show the differences?

AIDA64 only. I don't think I play any games that have engines which are sensitive to memory latency.

 

[Despoiler]

https://www.amd.com/en/support/chipsets/amd-socket-am4/x370

AMD Chipset Drivers
Revision Number 1.9.27.1033
File Size 50 MB
Release Date 9/27/2019

Supports:

Windows® 7/10 for
2nd Gen AMD Ryzen™ Desktop Processor
7th-Gen AMD A-Series Processors
AMD Ryzen™ Desktop Processor
AMD X470 Chipset
AMD X370 Chipset
AMD B450 Chipset
AMD B350 Chipset
AMD A320 Chipset
Windows® 10 only for
3rd Gen AMD Ryzen™ Desktop Processor
2nd Gen AMD Ryzen™ Threadripper™ Processor
AMD Ryzen™ Threadripper™ Processor
AMD Ryzen™ Desktop Processor with Radeon™ Vega Graphics
AMD Ryzen™ Mobile Processor with Radeon™ Vega Graphics
AMD X399 Chipset
AMD X570 Chipset

This is an all-new chipset software installer with updated UI, optimized package size, package versioning unique to the chipset drivers, and improved installer performance.

Package Includes:

AMD Chipset Drivers
AMD Ryzen™ Power Plans (required for UEFI CPPC2 in Windows® 10 May 2019 Update)
Contains desktop idle optimizations for 3rd Gen AMD Ryzen™ Processors

 

[DisEnchantment]

I would try setting vcore at 1.2 fixed and fan fixed as high as you can stand it. Also, Corsair fans are notoriously loud, I would consider replacing them.

You cpu speeds and average vcore look like mine. You also might try a fixed cpu speed if you use it a lot.

I was wondering and scratching my head why my Cinebench score were always 165 ST/2600 MT and when I read online it is mostly ~200ST/~3000 MT
After some backtracking I discovered that locking the Vcore to 1.2 caused this performance loss.

~18% drop is pretty significant, so I wonder if I could remove this Vcore locking without damaging the hardware
Also my Fan profiles are better now, I let the fans idle at higher rpm so that I dont get the temperature spikes which caused a lot of noise. And really give it the beans when the CPU hits 85 degrees celsius.
So far I have never hit more than 83 degrees even doing prime95

 

[EXCellR8]

repair install seemed to have cleared up my issues with Windows Update; now the PC is not throwing cryptic errors for me to attempt (and fail) to decipher.

 

[Seba]

I upgraded my main PC from i5-4440 (Haswell) to Ryzen 5 3600.

Motherboard: MSI B450 Tomahawk MAX.
RAM: 2x8GB kit G.Skill Ripjaws V Black DDR4-3200 CL16 1.35V (F4-3200C16D-16GVKB).
Stock cooler (AMD Wraith Stealth).

The motherboard came with 7C02v30 (v.3.0) BIOS.

Windows 10 Pro.
I installed the latest AMD chipset drivers directly from AMD (1.09.27.1033).

All defaults (no overclock), except XMP profile for RAM (also tried XMP and memory voltage manually set to 1.35V).

I've tested with Cinebench R20, Blender Benchmark, Windows 10 Memory Diagnostic, memtest86 v.8.2 (Passmark UEFI version), 3DMark (demo version from Steam), Unigine Heaven and Valley benchmarks. All OK (meaning no errors, nor crashes).

I haven't tried any game yet.

But in Prime95 (Version 29.8 build 6) I get rounding errors almost right away. Usually for Worker 4 and 3.

If the worker numbers correspond with core numbers (so if workers 1 and 2 are Core 1 and so on), that would mean my Core 2 is with issues. Core 02 is the weakest core from the first CCX (as reported by Ryzen Master).

"FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file."

If I disable AVX2 and AVX, I do not get this error. But I only let it run for about 10 minutes at a time, because the CPU temperature gets to about 90 degrees in this case.

Also, if I run just 2 workers, I can let AVX2 and AVX enabled and I don't get any error. At least not right away.

Disabling the memory XMP profile does not help with Prime95.

Although I read it had other issues, I flashed the MB to the latest BIOS/UEFI: 7C02v33 (v.3.3) - first BIOS with 1.0.0.3ABBA for my motherboard.
No change with Prime95.
But with this BIOS I had some of the known issues: Windows would remove/reinstall the Nvidia (GTX 1060 6GB) driver at start. Also once when I started the PC I got a corrupted image during boot. After a shutdown it was normal again.

I tried then the second available BIOS 7C02v31 (v.3.1, with 1.0.0.3AB). No change in Prime95.

I have not tried 7C02v32 (v.3.2) BIOS, because this one apparently also has various issues.

The Prime95 rounding error does not seem to be caused by my RAM, but I don't have any other DDR4 to test with (I only have some DDR3 modules).

Could it be because the stock cooler is not adequate for this Prime95 task?

Or I have a bad CPU?

Hard to figure this out, since I don't have a better cooler to try with it and also I don't have another Ryzen CPU to swap and see.

 

[Shmee]

Hard to say for sure, but you could run memtest86 to help eliminate RAM errors. If you have 4 passes no errors, it is likely the CPU.

 

[Seba]

I did run memtest86 (with XMP enabled) first thing with 3.0 BIOS. 4 passes. No errors.