AMD Ryzen 3000 Builders Thread

[EXCellR8]

3rd attempt at getting another 16GB CL14 Flare X kit... UPS 'incorrectly sorted' the parcel and it got sent to Vermont...

I quit.

 

[Kenmitch]

3rd attempt at getting another 16GB CL14 Flare X kit... UPS 'incorrectly sorted' the parcel and it got sent to Vermont...

I quit.

Ouch....That sucks! How far are you from Vermont in UPS days?

 

[EXCellR8]

In NH so right next door... but this being my 3rd time ordering a kit and not getting it (either on time or not at all) is aggravating.

My RMA kit is still all the way in CA; I don't plan on seeing that for at least another week.

 

[DrMrLordX]

Important to note, it's a different formula for the temperature reading now. Whereas before it was taking the absolute highest/hottest temperature reported (i.e. hotspot temp), it is now not only ramping up boost on idle less rapidly, it is also taking a "smoothed" temperature reading more representative of the die temperature by measuring at multiple sensors over a small time interval.

So combined with the relaxed boost from idle conditions you will see 1) lower idle temps and 2) less "spiky" temps. This doesn't really change anything in terms of power draw or performance so it's basically dumbing down the temperature reporting to placate the OCD types who complained about it.

Another thing to keep in mind for people not using the new chipset drivers: if you minimize Ryzen Master (or any other monitoring program) and use a histogram, you will see lower idle temps on the histogram. The temps spike while the monitoring program is active on the desktop. Weird but true.

I also managed to get higher boost clocks while Ryzen Master was minimized.

How do you guys test your memory overclocks?

Honestly? y-cruncher and Prime95 Blend.

 

[Kenmitch]

After some testing it looks like the Crosshair VI plays nicely with the 3600 and my d-die. Only real hurdle I found was with ram speed beyond the default. Easy fix in the uEFI once you find the bootup dram voltage setting. Just need to make it match the dram voltage the memory requires once overclocked is all.

I'm doubting the Crosshair VI really supports PCIe 4.0 as shown in GPU-Z. Is there really anyway to test it?



I was also able to bump up the clocks a little bit. Boost at all core 4017 MHz and single core would fluctuate between 4217-4242 MHz during the entire CB run.

 

[Kenmitch]

Installed the new beta chipset drivers and the new ryzen master software update.

https://community.amd.com/community...te-5-let-s-talk-clocks-voltages-and-destiny-2

Before I tried the new beta chipset drivers I installed some cumulative updates for Win 1903. I'm not really sure what happened after that. I initiated the chipset driver install and all was going fine, until the green screen instant reboot. There was some AMD security device flagged in device manager. I initiated the install again which looked to make it go away, but it popped up some message stating that this software is intended to only be installed on AMD hardware and aborts. Whatever went south also killed the monitoring in GPU-Z as everything would show 0's. This AM once I noticed the GPU-Z issue I tried to just install the beta chipset again and was greeted with the same message. The non beta onces installed fine and GPU-Z is monitoring correctly again.....Ignoring that it shows my PCIe running at 4.0 x16.

 

[DrMrLordX]

I'm doubting the Crosshair VI really supports PCIe 4.0 as shown in GPU-Z. Is there really anyway to test it?

Unless you are running a beta UEFI, it probably doesn't. I think 3DMark has a PCIe bandwidth test for video cards now.

 

[Kenmitch]

Unless you are running a beta UEFI, it probably doesn't. I think 3DMark has a PCIe bandwidth test for video cards now.

It's the latest uEFI from Asus I guess it's technically a beta version, but it looks to be working good so far.

I just ran that 3DMark PCIe bandwidth test, but how do you know what's good? Usually they have a better than ??% I thought.

http://www.3dmark.com/pcie/26995

 

[Hitman928]

It's the latest uEFI from Asus I guess it's technically a beta version, but it looks to be working good so far.

I just ran that 3DMark PCIe bandwidth test, but how do you know what's good? Usually they have a better than ??% I thought.

http://www.3dmark.com/pcie/26995

Looks like what AMD showed as PCIE4 performance:

 

[Kenmitch]

Looks like what AMD showed as PCIE4 performance:

Maybe so? I'll have to run the test on my x570 MB once I put the rig back together. Will be using the same gpu so I'd imagine the results should be similar.

 

[EXCellR8]

For some reason I haven't been able to install update KB4505903 which supposedly fixes some audio issues... not sure if anyone else has had this problem. I can probably install it manually but Windows just gives an error through regular update panel.

I'm still getting a very brief 'skip' in system sound every so often so hopefully after it's installed I can reinstall my regular sound card--though the onboard on this board is actually half decent tbch.

[EDIT]: Scratch that... wont let me install manually either.

 

[PotatoWithEarsOnSide]

Anyone using an X570 Gaming Edge WiFi?
MSI seem to have a distinct lack of BIOS updates since launch, and there seems to be complete radio silence in relation to them.
Got mine being delivered today, and was expecting to have to flash the latest BIOS on first boot.

 

[PotatoWithEarsOnSide]

Got home just in time for my motherboard to be delivered by Amazon.
A few more hours for the 5700XT to arrive.

 

[DrMrLordX]

I just ran that 3DMark PCIe bandwidth test, but how do you know what's good? Usually they have a better than ??% I thought.

It does look like you're getting PCIe 4.0 bandwidth.

 

[Kenmitch]

Anyone using an X570 Gaming Edge WiFi?
MSI seem to have a distinct lack of BIOS updates since launch, and there seems to be complete radio silence in relation to them.
Got mine being delivered today, and was expecting to have to flash the latest BIOS on first boot.

I have the MSI MPG x570 Gaming plus. I never tried the out of the box uEFI as I flashed it to the Agesa 1.0.0.3a right after I got it booted up. The lack of rushing uEFI updates I really see no issue with. I'd much rather they take their time on the next update and get it right the 1st time. I get boosting @ or very close to the advertised boot speeds. I've tried a 3600 and it would boost to 4.175-4200 MHz during CB R20 single core loads. I just dropped in a 3700x last night and it looks to follow a similar pattern. Haven't done much playing around with the 3700x yet. Single core boost during the CB R20 run looked to be hovering at 4375 MHz if I remember correctly.

 

[ozzy702]

Anyone know when a decent bios for the X370 Tiachi will be released? My buddy just ordered up a 3900x.

 

[Markfw]

Anyone know when a decent bios for the X370 Tiachi will be released? My buddy just ordered up a 3900x.

5.60 is OK, will run the chip. 5.61 somebody found on some site, and referenced here, not sure what thread.

 

[lopri]

I flashed to 1003ABB and the high vcore problem at 3600 MT/s memory is gone. I will see what else has changed and report back.

Meanwhile has anyone managed a Super PI 1M run under 9 seconds? I was so excited to see the first sub-10 sec number from an AMD CPU with my 3700X. (9.217 secs) Now that I see how close it is to 9 second mark, I am wondering what kind of frequency it takes to go below.

 

[Pandora's Box]

nvm

 

[Kenmitch]

Meanwhile has anyone managed a Super PI 1M run under 9 seconds? I was so excited to see the first sub-10 sec number from an AMD CPU with my 3700X. (9.217 secs) Now that I see how close it is to 9 second mark, I am wondering what kind of frequency it takes to go below.

What clocks got you that score?

I think the last time I ran that benchmark was on my 2550k overclocking adventure. With a crazy 1.7v's I was able to do some benchmarks @5.6GHz on custom water.

 

[lopri]

1.7V I don't think I have ever given that much to any CPU! What kind of 1M time did 5.6 GHz Sandy post?

The best 1M time I've got on this 3700X (9.168) is with bclk 102 and stock Precision Boost. Memory @3466/CL14.

 

[Kenmitch]

What kind of 1M time did 5.6 GHz Sandy post?

I showed mercy and only rapes 2 cores.

 

[lixlax]

I installed my new MB, SSD and RAM already 3 weeks ago, but what I can't buy is a 3900X. Rest of the lineup have been available since the launch

 

[Kenmitch]

I installed my new MB, SSD and RAM already 3 weeks ago, but what I can't buy is a 3900X. Rest of the lineup have been available since the launch

That sucks!

Bright side is maybe your MB manufacturer will launch a stable uEFI version that's not buggy.

 

[DrMrLordX]

So who here has been playing with the new combination of Ryzen Master, AMD chipset drivers, and the latest AGESA variant?

Thus far, I'm largely unimpressed. I've only been able to get boost clocks as high as 4.55 GHz in default settings and setting core affinity to one of the CPU's fastest cores with polling intervals in Ryzen Master of 4s or longer. Scored a 9.035s in SuperPi that way, which is interesting. PBO is not increasing my ST clocks at all. In MT CBR20, default is scoring around 7143 while PBO scores 7183. Not much difference there.

I went for a lighter workload, CBR10. In this workload, I got a better score thanks to conformity of core clocks. The bench finished its multiple threads closer to each other, whereas with default mode I had a few lingering render threads waiting for completion near the end of the bench (most were already finished). That resulted in about +4000-6000 points in the benchmark for PBO.

I also discovered that core affinity in Win10 task manager can be used to exclude certain cores if you set affinity on an MT application that is already "in flight". Good luck figuring out which core in Task Manager corresponds to which core on which CCX of your Matisse since it can get a little weird in MT workloads. You might be able to take advantage of that in games that can't spawn 24 threads. Unfortunately, I can't positively demonstrate that clockspeeds will be much higher in such scenarios, and the Windows scheduler is supposed to be picking those cores anyway.

All testing up to this point ran in Ryzen Performance mode (not Ryzen Balanced). I haven't tried negative vcore offsets yet.