I doubt your cheap Biostar motherboard has the necessary robust VRM components to allow much more than stock. Probably only a 1 + 4 VRM design like much of the B350 motherboards.

Well it certainly didn't cost me much (less than $60) but I am not so sure I would call it completely "cheaply designed".

Apparently it will run the next generation of Ryzen cpus. At least the bios flash I did earlier today appears to support the Zen 2 even though the bios menu itself doesn't offer listed support for the Zen+ stuff.

This https://www.reddit.com/r/overclocking/wiki/database/am4#wiki_tb350-btc claims a

VRM: 4+3-phase with a heatsink on Vcore.

but I don't have any way to rate my confidence in the statement.

Can't find any numbers like that on the Website product page.

I may try some more attempts to 3.9GHz but it is smoking along reliably at 3.7GHz so maybe I will stop experimenting....

Tom

---

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

When you see a VRM design listed as 4+1 or 8+2 or 10+2 or 16+2 in the new X570 or Threadripper motherboards, it is defining the number of power stages used for the cpu and for the SoC/memory. The first number defines the number of stages for the cpu and the second number after the + sign defines the number of stages for the SoC/memory. The SoC and memory are not very power hungry compared to the cpu so they can get away with lesser amount of power stages. The lesser amount of power stages means a smaller amount of components which reduces motherboard cost.

[Edit] Your motherboard using a 4+3 is very strange because it is not an even number and I don't believe is possible to design. Stages are complementary and always in pairs.

---

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

A proud member of the OFA (Old Farts Association)

When you see a VRM design listed as 4+1 or 8+2 or 10+2 or 16+2 in the new X570 or Threadripper motherboards, it is defining the number of power stages used for the cpu and for the SoC/memory. The first number defines the number of stages for the cpu and the second number after the + sign defines the number of stages for the SoC/memory. The SoC and memory are not very power hungry compared to the cpu so they can get away with lesser amount of power stages. The lesser amount of power stages means a smaller amount of components which reduces motherboard cost.

[Edit] Your motherboard using a 4+3 is very strange because it is not an even number and I don't believe is possible to design. Stages are complementary and always in pairs.

That sounds like serious case for reducing the credibility of that number.
Found an image online that appears to be the MB in question. Is there anyway to infer from what we "see" something about the VRM?

https://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=882

Tom

---

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

The AM4 Zen2 results are in. It is likely I have to find a X570 motherboard with 6 Pciex slots (and the ability to run 9 gpus) if I want to increase my core count. The other choices all are 8c/16t's. Some fit my Biostar TB350-BTC TDP 95 watt limit. But that would effect no more than 5 threads that are running cpu only crunching.

If it turns out that I can run the gpus without the "-nobs" command line at say 0.33 gpus per cpu and still maintain a stable RAC, then going for a faster, same thread count cpu might make sense.

Tom

---

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

your system is stable now that you finally replaced the Cables. I wouldn't bother changing anything.

replacing the CPU wont meaningfully impact your performance at SETI.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

When you see a VRM design listed as 4+1 or 8+2 or 10+2 or 16+2 in the new X570 or Threadripper motherboards, it is defining the number of power stages used for the cpu and for the SoC/memory. The first number defines the number of stages for the cpu and the second number after the + sign defines the number of stages for the SoC/memory. The SoC and memory are not very power hungry compared to the cpu so they can get away with lesser amount of power stages. The lesser amount of power stages means a smaller amount of components which reduces motherboard cost.

[Edit] Your motherboard using a 4+3 is very strange because it is not an even number and I don't believe is possible to design. Stages are complementary and always in pairs.

That sounds like serious case for reducing the credibility of that number.
Found an image online that appears to be the MB in question. Is there anyway to infer from what we "see" something about the VRM?

https://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=882

Tom

Wow, it really is 4+3. Very strange design. A total one-off product.

---

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

A proud member of the OFA (Old Farts Association)

I suppose I should "feel" cheated. Someone has managed to get 12 gpus using 2 extender boards running on my Biostar model MB. I don't know if he is running mining or some other kind of processing.

I suppose I could buy another 1 to 4 extender and see if I can get "more" on there....

Tom

---

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

I suppose I should "feel" cheated. Someone has managed to get 12 gpus using 2 extender boards running on my Biostar model MB. I don't know if he is running mining or some other kind of processing.

I suppose I could buy another 1 to 4 extender and see if I can get "more" on there....

Tom

Probably running mining with PCIe lanes set to gen 2 or gen1.

I wouldn’t bother.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

I suppose I should "feel" cheated. Someone has managed to get 12 gpus using 2 extender boards running on my Biostar model MB. I don't know if he is running mining or some other kind of processing.

I suppose I could buy another 1 to 4 extender and see if I can get "more" on there....

Tom

Having flashed the bios to the "most current" version

I retried the 10 card experiment with a "known" good 1 to 4 extender card and it still was failing to start the GUI no matter what combo in the Bios I used. This is the same result I had the last time on this system as well as on my Intel box.

I did discover that with the "Above 4G" coding off but the "Above 6 Vga" mining mode on, it still will recognize 9 gpus. Wasn't brave enough to try running it that way though.

Tom

---

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

I also confirmed I have my DDR voltage at 1.35 and "everything" seems to be humming along.

You know . . . . you are not constrained to using only 1.35V. DDR4 is rated to handle 1.5V. I run my DDR4 at 1.4V with no issues. All the Ryzen high clock memory presets are set at 1.4V.

If you aren't stable at 1.35V, then bump the voltage up to 1.365V or maybe 1.38V. Don't be afraid to boost voltages, the silicon can take it.

---

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

A proud member of the OFA (Old Farts Association)

Hmmm, the people in that thread don't seem to be able to read the manual. It's pretty clear about using three different Power Supplies for the 3 different groups of GPUs.
You don't connect all the GPUs to One Power Supply. Also, the P106 is basically a 1060 running at PCIe Gen 1, https://www.youtube.com/watch?v=O1nok2VlF1M
Actually, that thread is quite encouraging. It's just I'm not up to moving things around to add a third PS right now. I'll probably get another 1060 first, to replace the last 1050.

no configuration in the manual lists a setup with >13 normal GPUs, only when using mining cards to get it up to 19, but good luck.

a P106 is comparable to a 1060, but i don't know if anyone has tried them with SETI. you may run into driver and/or app/boinc issues recognizing it as a valid card. buy one and let us know if it works.

I saw that video months ago. they had to edit the drivers, but unfortunately they only showed modifying the drivers on a windows platform. I'm not sure if anyone has done the same on the Linux drivers.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

I didn't say anything about hacking drivers, I merely pointed to a related video where it was stated a couple of times that the P106 & GTX1060 is the exact same hardware. I have a handful of dual fan Gigabyte 1060's, they are the exact same card as the Gigabyte P106 except the display ports and backplate is removed from the P106. Right now things seem to be working just fine with 5 slots left open between the middle and lower groups, and not a P106 to be seen;

setiathome_CUDA: Found 14 CUDA device(s):
  Device 1: GeForce GTX 1060 3GB, 3016 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 1, pciSlotID = 0
  Device 2: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 4, pciSlotID = 0
  Device 3: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 8, pciSlotID = 0
  Device 4: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 10, pciSlotID = 0
  Device 5: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 12, pciSlotID = 0
  Device 6: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 13, pciSlotID = 0
  Device 7: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 14, pciSlotID = 0
  Device 8: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 15, pciSlotID = 0
  Device 9: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 16, pciSlotID = 0
  Device 10: GeForce GTX 1060 3GB, 3019 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 9 
     pciBusID = 18, pciSlotID = 0
  Device 11: GeForce GTX 970, 4043 MiB, regsPerBlock 65536
     computeCap 5.2, multiProcs 13 
     pciBusID = 17, pciSlotID = 0
  Device 12: GeForce GTX 1050, 2000 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 5 
     pciBusID = 7, pciSlotID = 0
  Device 13: GeForce GTX 1050, 2000 MiB, regsPerBlock 65536
     computeCap 6.1, multiProcs 5 
     pciBusID = 9, pciSlotID = 0
  Device 14: GeForce GTX 960, 2002 MiB, regsPerBlock 65536
     computeCap 5.2, multiProcs 8 
     pciBusID = 11, pciSlotID = 0

Things did slow down a bit, as usual when adding cards. I'll just have to buy a few 1070s, or so ,to replace those older cards some day.
The nice thing is things are running a bit cooler and quieter with the slowdown.

All I can say TBar is . . . . . WOW! Congratz on sticking with that motherboard. !Kudos!!

---

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

A proud member of the OFA (Old Farts Association)

Yes, I understand it's some sort of record or something? The only thing I'm sure of is my back hurts and I need more PCIe MB connectors ;-)
I had to give up on the Spoofed client, too much lag at startup and it only ran 12 tasks. The 'normal' client has a bit of lag too, I had to go to 'Show active tasks', maybe just too many tasks. We'll see what happens around 3k tasks.

good job getting it working. hope it lasts, since every other time you tried 13 you had system issues shortly after. but yes, i've never seen anyone get >13 on that board. I have no experience with it since I've never used one personally. looking again, i see some vague references that 14 is possible in some cases, but most of the comments i see say that 13 is the magic number for stability and support. try for 15. see what happens.

and the P106-100 is not the "exact" same hardware as a 1060. the P106 lacks any NVENC for example. but for the purposes of SETI and other computational work, they should perform similarly. After reading some more about them on various forums, it seems you might not need to do any driver modification for the drivers to work with them on linux (but may require an older version). I'm going to China in a few weeks, maybe i'll bring one back to play with.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Hi,
This is a re-direct on Intel Motherboards/Mining Motherboards to here: https://setiathome.berkeley.edu/forum_thread.php?id=83307 where high gpu count Intel MB's and Intel Mining MB's are more on topic.

Tom

---

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

It still looks like trying to run a Ryzen 3000 cpu with 16c/32t's on a Biostar TA350-BTC probably won't work. That is because even the Ryzen 3900x (12c/24t) has a TDP higher than the TA350-BTC claims it could work on.

I suspect the safest way to test this proposition is to buy another TA350-BTC and install at Ryzen 3000 16t/32c on it, add a gpu or two and just run it. That way I won't put my highest performing Rig's MB at risk just for a test.
If the Ryzen 3000 does blow that MB then I can save up my penny's for an X570 MB with at least 6 gpu slots....

Future Proof? Whats that? ;)

Tom

---

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

In case you didn't read it someplace else my Biostar TB350-BTC with a Ryzen 7 2700 (@least 3.7GHz) and 9 gpus (3 different groupings) has cracked the 400,000 RAC mark.
It did that sometime last Saturday (6/1/2019).

Here is the listing on the Leaderboard (look for Tom M) https://setiathome.berkeley.edu/top_hosts.php

Tom

---

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

Just discovered that the TDP for a Ryzen 7 2700x is the same as the top Ryzen 3000 cpus. Or at least that is the number on the "cpu support" list on the website for my Biostar TB350-BTC.

Which means I can attempt to run a Ryzen 9 3900x without assuming it will burn out the MB. If I want to spend $500 on upgrading my cpu instead of replacing 3 Gtx 1060 with Gtx 1070's :)

Tom

---

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