My preliminary guess is that at best the GPU on the Amd 5 2400G is the near equivalent of about 1/2 of a Gtx 750Ti running under Windows 10.

I have two Gtx 750Ti's running under Windows on Intel boxes. When we are running "average" task complexity they take around 15 minutes. The 2200G is taking around 22-30 minutes so far. That is with no command line parameters. I just added my current best guess a little while ago.

I am running 75% cores and running the gpu with 0.33 cores which means it doesn't count against my core limit. So I have 6 cpu threads and 1 gpu thread running.

The memory is set to CL16?/3200 for the frequency. The chips are CL14/3200.

Things should begin to settle a little more now (I hope). It is beginning to look like the cpu task 8.00 is running faster than any of the others.

Tom

---

A proud member of the OFA (Old Farts Association).

My preliminary guess is that at best the GPU on the Amd 5 2400G is the near equivalent of about 1/2 of a Gtx 750Ti running under Windows 10.

I have two Gtx 750Ti's running under Windows on Intel boxes. When we are running "average" task complexity they take around 15 minutes. The 2200G is taking around 22-30 minutes so far. That is with no command line parameters. I just added my current best guess a little while ago.

I am running 75% cores and running the gpu with 0.33 cores which means it doesn't count against my core limit. So I have 6 cpu threads and 1 gpu thread running.

The memory is set to CL16?/3200 for the frequency. The chips are CL14/3200.

Things should begin to settle a little more now (I hope). It is beginning to look like the cpu task 8.00 is running faster than any of the others.

Tom

There are two issues that I know about for sure.
1) With the memory I am running the system crashes if you don't set the default DRAM voltage to 1.35 volts instead of "auto".
2) The screen refuses to come back on if you let Windows 10 turn it off after some time period. So it is set to "never" turn off.

On a side note, I was able to get the Internal gpu and two discrete gpus running on this motherboard (Asus B350M-E) using a PCIe expansion card to setup the gpus outside the case. Yes, I needed a bigger PSU.

Tom

---

A proud member of the OFA (Old Farts Association).

Tom, interesting findings. Thanks for sharing your results and troubleshooting. Did you see any throttling when running the integrated GPU like we witnessed with the older APUs?

Tom, interesting findings. Thanks for sharing your results and troubleshooting. Did you see any throttling when running the integrated GPU like we witnessed with the older APUs?

I am clearly having some kind of issues. It is still crashing. I have been running it at 6 out of 8 threads and trying to see if a full thread or partial thread makes a difference on the speed of the gpu processing.

A little while ago I turned it from 75% cpu to 100% but still have 1 full thread driving the gpu.

I also have the MB "Asus Optimum" setting (eg. one of the possible automated overclocks). I also have the memory set to auto and dram voltage set to auto. So the ram is running at 2400 with 1.2volts. The DRAM is rated at CL14/3200 at 1.35volts.

The cpu is running near 3.9GHz. I can't tell what the gpu is running at. I can say that any attempt to use the builtin gpu overclocking vs. manual overclocking causes gpu computation errors.

I have not been able to even manually overclock the gpu at the 1400-1600 range. I just finished 8+ hours where it ran at 1000 and didn't apparently stumble.

So right now my system is not running as fast as most of the reviews talk about. I can't figure out some of how the voltages settings work on this entirely different AM4 motherboard.

I have had to reset the cmos a couple of times :(

So I don't think I am really making much progress except that I am not replicating the apparent review performances :)

Tom

---

A proud member of the OFA (Old Farts Association).

Just for giggles, I have just switched my 2400G machine to an all GPU diet. Once it runs out of CPU tasks, maybe I will get a clear idea of what speed the GPU can process at.

I originally thought it would be running near 50% of a gtx 750ti. But either I was wrong, my settings are slowing it down or the latest mix of gpu tasks are running much slower than the starting estimate of near 30 minutes per task.

Tom

---

A proud member of the OFA (Old Farts Association).

Just for giggles, I have just switched my 2400G machine to an all GPU diet. Once it runs out of CPU tasks, maybe I will get a clear idea of what speed the GPU can process at.

I originally thought it would be running near 50% of a gtx 750ti. But either I was wrong, my settings are slowing it down or the latest mix of gpu tasks are running much slower than the starting estimate of near 30 minutes per task.

Tom

I got to thinking about this and realized that I could use <project_max_concurrent> to do the same thing via the app_config.xml file.

So I did. Didn't appear to change anything so I did.
1) cleared the command line files.
2) Used the Amd Master app to see how fast the gpu was running.
3) Successfully used the Bios gpu setting of "exterme" to bump the gpu speed to 1500 (it was running around 1100 on "auto").
4) I am now running 4 threads, 3 cpu and 1 gpu.

Some of the latest results are reasonably fast for a change. The question is if those are outliers or not.

If I try to use the Amd Master app to set for the "default" creator settings the system won't boot and I have to clear the CMOS before it will.

I got busy and installed a proposed beginners Web Dev setup that included enabling the Linux file system under Windows, installing Ubuntu from the Windows app store. Installing git under both windows and under linux. And installing the vscoding editor under windows and setting it up so it is the default under Linux. This was all from googling "beginner web dev" and following a "medium" article.

Tom

---

A proud member of the OFA (Old Farts Association).

The "averages" haven't dropped (yet) but with the Gpu set at 1500 GHz's I am seeing at least some elapsed times in the 30 minute or lower range.

The cpu is at baseline (3.6Ghz) but running with cpu boost enabled so windows is reporting speeds of about 3.88GHz most of the time.

The other interesting result is when I select the gpu under the Windows task manager it says there are 3 different computation areas but only 1 is in use. I wonder if it would use a 2nd area if it had 2 tasks? Naw, that would be too easy...

Tom

---

A proud member of the OFA (Old Farts Association).

The other interesting result is when I select the gpu under the Windows task manager it says there are 3 different computation areas but only 1 is in use. I wonder if it would use a 2nd area if it had 2 tasks? Naw, that would be too easy...

Tom

Just tried that and while it pushed the one gpu "compute" box to near 100% it didn't start something in one of the other "compute" boxes.

Darn.

Without trying it, I am assuming that if I run 2 gpu tasks the speed of processing will drop in half or slower. That has been a reliable result for years.

Tom

---

A proud member of the OFA (Old Farts Association).

GPU processing speed is entirely dependent on system memory other than obvious core clock speed. So if you are running system memory at 2400Mhz for stability, you are going to see longer compute times compared to running the system memory at the rated 3200Mhz XMP frequency at spec 1.35 volts.

Your memory can be run at 1.4V but it shouldn't be necessary for just 3200 clocks. I would think 1.365V - 1.38V would be sufficient and what the DRAM calculator shows normally. The Vsoc voltage should be around 1.05 - 1.1V but I have seen reports in the forums that the new Ryzen APUs work better with a little more SoC voltage compared to a Ryzen by itself because of the extra work the IMC has to do with the APUs. I would still stay below the previous maximum allowed spec for Vsoc of 1.2V

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

GPU processing speed is entirely dependent on system memory other than obvious core clock speed. So if you are running system memory at 2400Mhz for stability, you are going to see longer compute times compared to running the system memory at the rated 3200Mhz XMP frequency at spec 1.35 volts.

Your memory can be run at 1.4V but it shouldn't be necessary for just 3200 clocks. I would think 1.365V - 1.38V would be sufficient and what the DRAM calculator shows normally. The Vsoc voltage should be around 1.05 - 1.1V but I have seen reports in the forums that the new Ryzen APUs work better with a little more SoC voltage compared to a Ryzen by itself because of the extra work the IMC has to do with the APUs. I would still stay below the previous maximum allowed spec for Vsoc of 1.2V

I need to understand my MB bios better. It has "offsets" for things like the Vsoc voltage so I can't figure out exactly how to get from the voltage being shown to the higher voltage.

I have not had any crashes that seemed to be related to DRAM itself since I started using 1.35 volts.

When trying the offsets I have managed to reach "will not boot" very easily :( I guess I need to google MB bios details better. So far I haven't found any.

Tom

---

A proud member of the OFA (Old Farts Association).

An offset is just that. Either subtract or add from the base value depending on the sign. You use it by turning off Auto and selecting Offset or Manual. Put in the desired value in the offset field and select the sign. The report back value reflects the change from the base value you either accept as Auto is turned off or the value you input as your choice. Google some YT videos on BIOS manipulation if you are still confused. The motherboard manual covers the basics of changing the BIOS but manual quality differs among brands it seems. Offset is handy in that it still allows automatic tracking of loading just like Auto but you decide where the start point is and not whatever the BIOS developers decided for the value in Auto. Manual is just that, manual and static and never changes, so makes a poor choice for a varying load.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

An offset is just that. Either subtract or add from the base value depending on the sign. You use it by turning off Auto and selecting Offset or Manual. Put in the desired value in the offset field and select the sign. The report back value reflects the change from the base value you either accept as Auto is turned off or the value you input as your choice. Google some YT videos on BIOS manipulation if you are still confused. The motherboard manual covers the basics of changing the BIOS but manual quality differs among brands it seems. Offset is handy in that it still allows automatic tracking of loading just like Auto but you decide where the start point is and not whatever the BIOS developers decided for the value in Auto. Manual is just that, manual and static and never changes, so makes a poor choice for a varying load.

That seems perfectly logical. I guess I am missing the "report back" step to see what the output changes to.

Tom

---

A proud member of the OFA (Old Farts Association).

What I meant by the"report back" is the field in front of the entry field for the parameter you are adjusting in the BIOS. Every BIOS I have ever seen shows the current value next to the entry field. When you change the parameter in the entry field, you see the new value reflected on the same line.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

What I meant by the"report back" is the field in front of the entry field for the parameter you are adjusting in the BIOS. Every BIOS I have ever seen shows the current value next to the entry field. When you change the parameter in the entry field, you see the new value reflected on the same line.

Let me take a look. TY.

Tom

---

A proud member of the OFA (Old Farts Association).

What I meant by the"report back" is the field in front of the entry field for the parameter you are adjusting in the BIOS. Every BIOS I have ever seen shows the current value next to the entry field. When you change the parameter in the entry field, you see the new value reflected on the same line.

Let me take a look. TY.

Tom

It crashed when I tried a tiny bit of SOC positive offset. It took some seti files with it. Had to select Seti again under the manager. And now the Stats graph is not displaying. It took the app_config.xml file too. :(

I did find a mention of an "Amd balanced power plan" that would speed up games an average of 5%. It installs with a chip set download. So I found the chipset for my motherboard and downloaded it.

I have turned off the C states on the motherboard. And just now turned off the PSS on the motherboard. Who knows, maybe I can get under 2 hours for a cpu task :)

Tom

---

A proud member of the OFA (Old Farts Association).

You should not have tried a large positive offset on SoC voltage. It runs stock on Auto around 1.05 -1.1V The absolute maximum recommended SoC voltage per AMD specs is 1.2V or bad things will happen. Seems you found out that is true.

AMD balanced power plan is not recommended now. It was a stopgap measure by AMD on 1st generation Ryzen until Windows fixed its scheduler. The recommended power plans for Windows is either high-performance or Windows Balanced Power plan with minimum processor state from 20- 50%

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

You should not have tried a large positive offset on SoC voltage. It runs stock on Auto around 1.05 -1.1V The absolute maximum recommended SoC voltage per AMD specs is 1.2V or bad things will happen. Seems you found out that is true.

AMD balanced power plan is not recommended now. It was a stopgap measure by AMD on 1st generation Ryzen until Windows fixed its scheduler. The recommended power plans for Windows is either high-performance or Windows Balanced Power plan with minimum processor state from 20- 50%

It wasn't, I don't think, a large offset. I bumped it up "2 steps" which translated into something like +0.000345? it was incremental. Anyway, I clearly need to do a lot more reading on the details of my motherboard.

Tom

---

A proud member of the OFA (Old Farts Association).

Typically in most BIOS, the minimum step size is 6mV or 0.00625V At least for the MSI and ASUS and Gigabyte boards I have used. I don't think any motherboard can offer 0.3mV step size.

Just stay clear of Vsoc voltage of 1.20V at least for basic Ryzen. I have read in the forums that the new APU's can use up to 1.25V because of the extra loading on the IMC to control both the memory and gpu cores. You just don't want to run there if you don't have to. Every mV added to the SoC voltage increases temps and degrades both cpu clocks and gpu clocks. You always want to run with the minimum voltage necessary for the CPU and the SoC to keep temps down and that still achieves stability under load.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Well, my 2400G cpu and probably MB have crapped out. Going to have to replace them.

The 2400G cpu chip will not boot in another MB. And when I dropped a different cpu into the original MB it booted once and now won't. I now suspect the MB may have fried both cpus :(

Tom

---

A proud member of the OFA (Old Farts Association).

I would never run a Ryzen of any kind on a 4 +1 VRM motherboard. Just not enough power and not high enough quality. I don't cheap out on parts and they don't fail on me. You get what you pay for always applies.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)