OCNG5: OC firmware for Supermicro AMD G34 platforms

58 deg Tctl is high-ish with OC. Not in context of absolute heat but, as you OC, thermal thresholds of stability ultimately go down (instead of 70 deg Tctl which is the default, you may start witnessing instability already at 60 deg Tctl).

Heat and power delivery of your board will both become factors. Be very careful not to fry your board!
 
thank you for the response. I got the 6282 se's in and did notice that on oc (230 fsb) that I started to see slowness in counterstrike go. I imagine my cores must have been overheating. The day prior when I shutdown my vfio vm it actually shut down my entire host!

I also commented on your blog page vs this forum that the only way for me to OC is to set the FSB to 230 AND set the option to set max p state to yes. Otherwise I see no OC. Which creates it's own issue (the processors are only rated to allow up to 8 cores to be running at max turbo). I honestly don't need them all at max. I'm getting as high as 4GhZ clocks on some cores, but roughly between 3.0 GhZ and 4.0GhZ but never the low p state values like 1.x or even 2.x GhZ.

Any ideas why setting FSB alone with max p state off doesn't give me a fsb overclock?

edit:

turns out lscpu shows a lower speed than clockspeed. Thank you tear!
 
Last edited:
hi Folks,

i recently flashed OCNG 5.3. After the flashing procedure (after flash and running holes)... all the CPUs came up running at DDR3 1600 just fine. i then booted into windows just fine. Ran OCNG utilities and set XMP etc. Ran fine for a bit... had a shut down. Rebooted (2 beeps followed by 5 beeps) three times... and the machine finally posted again. However, i noticed that CPU 3 (out of 4) always seems to come up as DDR3-1333 for both of its dies. All the others successfully come up running at DDR3-1600. DRAM consists of DDR3-1600 G-Skill F3-12800CL9-2GBNQ modules.

These are ES chips. Just wondering if one of my 4 CPUs simply cannot handle DDR3-1600? Anyone run across this before? Is there a fix?
 
nvm, turned out to be a single bad stick that would run DDR3-1333 CL8 just fine but not at DDR3-1600 CL9 and 1T... after replacing the bad module the system now runs at DDR3-1600 at the XMP profile timings but at 1T instead of 2T.

Running%20at%20DDR3-1600%20CL9%201T_zpstxkt8ogx.jpg


hi Folks,

i recently flashed OCNG 5.3. After the flashing procedure (after flash and running holes)... all the CPUs came up running at DDR3 1600 just fine. i then booted into windows just fine. Ran OCNG utilities and set XMP etc. Ran fine for a bit... had a shut down. Rebooted (2 beeps followed by 5 beeps) three times... and the machine finally posted again. However, i noticed that CPU 3 (out of 4) always seems to come up as DDR3-1333 for both of its dies. All the others successfully come up running at DDR3-1600. DRAM consists of DDR3-1600 G-Skill F3-12800CL9-2GBNQ modules.

These are ES chips. Just wondering if one of my 4 CPUs simply cannot handle DDR3-1600? Anyone run across this before? Is there a fix?
 
Last edited:
Glad you figured it out. I was away all weekend for business, so had no time to try and dig into any of it. A lot of our past members don't post anymore so it gets harder to find support for some of these tools.
 
Overclocked that thing must eat 750w at the wall or something crazy like that - has it ever been measured?
 
Overclocked that thing must eat 750w at the wall or something crazy like that - has it ever been measured?

I always considered overclocking my chips, considering they run just under 40C under full load.. just never got around to flashing this BIOS and figured it wouldn't give me enough of an increase anyway.
 
i think for the folks running stock (ie cpu multiplier locked) chips this is not a bad option. On my rig i still use TPC since i clock some of my cores up to their maximum... 3.8 GHz. i wish there was a way to set the CPU-NB multiplier separately... without affecting ram and cpu timings, since its so important for the performance of the K10.5 cores.

To get the ram running at these speeds i selected manual... 800 Mhz and that did the trick. Now that i've got my ram issue sorted out i'll be experimenting with raising the reference clock.

I always considered overclocking my chips, considering they run just under 40C under full load.. just never got around to flashing this BIOS and figured it wouldn't give me enough of an increase anyway.
 
Last edited:
i think for the folks running stock (ie cpu multiplier locked) chips this is not a bad option. On my rig i still use TPC since i clock some of my cores up to their maximum... 3.8 GHz. i wish there was a way to set the CPU-NB multiplier separately... without affecting ram and cpu timings, since its so important for the performance of the K10.5 cores.

To get the ram running at these speeds i selected manual... 800 Mhz and that did the trick. Now that i've got my ram issue sorted out i'll be experimenting with raising the reference clock.

So, a 6172 chip is multiplier locked?
 
i was experimenting with upping my ref clock... in order to overclock the northbridge and gain a bit of additional performance on the L3. i was going in small increments. 205 was fine. 210 booted but i was experiencing random shutdowns. i set things back to 209... and Windows would boot but immediately reboot the machine at the login screen. i ended up rebooting... 3 times and artificially shutting the machine down each time to reset the OCNG to get things back into a bootable state.

At this point i'm wondering how a lot of you 4P folks were able to up your ref clocks to 220+ as 210 isn't even stable for me. Perhaps the NB voltage needs to be upped to give additional stability prior to booting the OS? So how did you folks running the 61xx chips up the NB voltage? i don't see an option for the 61xx folks.

i really like OCNG because it has allowed me to finally run this 4P rig at my memory's rated speeds (slightly faster actually due to the timing merge)... but man alive given how important running decent NB speeds are to these 61xx processor dies... it would be nice to get the NB to at least 2400.
 
i was experimenting with upping my ref clock... in order to overclock the northbridge and gain a bit of additional performance on the L3. i was going in small increments. 205 was fine. 210 booted but i was experiencing random shutdowns. i set things back to 209... and Windows would boot but immediately reboot the machine at the login screen. i ended up rebooting... 3 times and artificially shutting the machine down each time to reset the OCNG to get things back into a bootable state.

At this point i'm wondering how a lot of you 4P folks were able to up your ref clocks to 220+ as 210 isn't even stable for me. Perhaps the NB voltage needs to be upped to give additional stability prior to booting the OS? So how did you folks running the 61xx chips up the NB voltage? i don't see an option for the 61xx folks.

i really like OCNG because it has allowed me to finally run this 4P rig at my memory's rated speeds (slightly faster actually due to the timing merge)... but man alive given how important running decent NB speeds are to these 61xx processor dies... it would be nice to get the NB to at least 2400.


I have 4P 6386 SE and I have the system running at 216MHz. Maybe you've reached the upper limit of your CPU. As far as going to 220, I don't know if I want to push it that hard.

You should be able to troll eBay and upgrade your processors on the cheap.
 
i was experimenting with upping my ref clock... in order to overclock the northbridge and gain a bit of additional performance on the L3. i was going in small increments. 205 was fine. 210 booted but i was experiencing random shutdowns. i set things back to 209... and Windows would boot but immediately reboot the machine at the login screen. i ended up rebooting... 3 times and artificially shutting the machine down each time to reset the OCNG to get things back into a bootable state.

At this point i'm wondering how a lot of you 4P folks were able to up your ref clocks to 220+ as 210 isn't even stable for me. Perhaps the NB voltage needs to be upped to give additional stability prior to booting the OS? So how did you folks running the 61xx chips up the NB voltage? i don't see an option for the 61xx folks.

i really like OCNG because it has allowed me to finally run this 4P rig at my memory's rated speeds (slightly faster actually due to the timing merge)... but man alive given how important running decent NB speeds are to these 61xx processor dies... it would be nice to get the NB to at least 2400.


Buy MC ES chips and try again, sounds like MC retail chip limitations TBH. ES chips you can drop the multi... retail chips IIRC are locked you can't even drop down 1x. Dropping a multi and then bumping NB and good cooling would help ;)
 
I recently upgraded to 6378's and ran 240mhz for a week. Seemed to be alright, temps were fine. Only issue is TeamViewer would no longer show me video after a while but the rig kept churning out work. Changed it to 230 and haven't had an issue.
 
Anyone by chance has ocng-utils-5.3 ? I just got the motherboard today but can't find it. Can anyone who has it attach it in the forum?
 
Anyone by chance has ocng-utils-5.3 ? I just got the motherboard today but can't find it. Can anyone who has it attach it in the forum?
I'll have a look around, but I can't promise anything. I got rid of my board years ago but I have a sickness and I save almost everything lol
 
Last edited:
The bios is still downloadable here:

http://darkswarm.org/sm-bios/ocng5-5.3-20151222.7z

i verified that the zip is the same as the one i downloaded in 2018

the ocng utils link is not on the web archive though.
I just found mine as well. If the link isn't working for you, let me know and I can email it.

edit: I tried that link, it wouldn't download. Just giving me a network error in Chrome.
 
I just found mine as well. If the link isn't working for you, let me know and I can email it.

edit: I tried that link, it wouldn't download. Just giving me a network error in Chrome.

i fixed the link to the one on the web archive since that one still works.

PS: My Quad Opteron (SuperMicro H8QGi-F running OCNG, and Extra Spicy cpus) is still running strong every day.
 
i fixed the link to the one on the web archive since that one still works.

PS: My Quad Opteron (SuperMicro H8QGi-F running OCNG, and Extra Spicy cpus) is still running strong every day.
I was running a 4P 6378 for years, but sold it locally about a year ago for $300. Figured I probably wouldn't see that kind of offer again considering how cheap the chips and ram were. The board though, that's another story.
 
Hi,

Does anyone reach with this custom BIOS on H8DGU-F with dual 6386 SE or H8QGi+-F with quad 6386 SE CPU's frequency 3,7 GHz like how it was for Phenom X6 1100T to work back that time ?.
And if so, there is any feeling for real performance increase ?. I could not see so in case of Phenom X6 1100T, which was stock 3,3 as max for Opteron 6386 SE, and turbo to 3,7 GHz. Phenom X6 1100T was unstable and use to restart at frequency above 3,7 Ghz on an Asus motherboard top list that moment. That was at least in my case.

And anyone could extend this custom BIOS functionality, like for example have the only one PCIe slot on H8QGi+-F to run on 2*8x ? . I need it splited for a video card and a storage card, as integrated video card is obsolete and cannot allow to use system as workstation due to very low performance level. I could see an extra applications with a configuration menu in place, and even if OC related, my assumption is that it feature extra functionality for default BIOS anyway.
Supermicro Support informed me 3 months ago that development for H8QGi+-F was dropped long time ago and no more assistance in any way can be offered.
The Supermicro riser cards will not work outside just first slot( for example : RSC-R1UU-2E8 or RSC-R2UG-2E4E8).
There is a third party that offer a riser card supposedly compatible and provide such feature( 2*8x or 4*4x4x4x4x), but the cost of that in this point in time is the same or higher than the cost of the other motherboard same Supermicro class with extended PCIe slots (about 400 euro).

Thanks.
 
thanks to you guys, I have the file now. I'll try it and run some benchmarks.
Hi,
I am just curious. What expectations do you have for such old server CPU class ?. Ar you using it for mining Monero or so ?.
I am using it as workstation and get advantage from 1 TB ram feature as I am looking for very fast storage device and hence I am using ramdrives.
For example under my tests it show that Opteron 6386 SE is at least 4 times less performant per core than Intel I7 mobile Gen 11 CPU in Monero mining, at same CPU core frequency 3,3 GHz, while price ( for the system itself) been the same (about 2000 euro) but power consumption in matter of electricity been 10 times more in Opteron side. Also Opteron operate with RAM at DDR3 1333 MHz frequency, while Intel I7 Mobile at 3800 Mhz DDR4. This may also matter and looks obvious.
Thanks.
 
I don't think anyone in their right mind is looking at old Socket G34 CPU's for mining...
 
Hi,
I am just curious. What expectations do you have for such old server CPU class ?. Ar you using it for mining Monero or so ?.
I am using it as workstation and get advantage from 1 TB ram feature as I am looking for very fast storage device and hence I am using ramdrives.
For example under my tests it show that Opteron 6386 SE is at least 4 times less performant per core than Intel I7 mobile Gen 11 CPU in Monero mining, at same CPU core frequency 3,3 GHz, while price ( for the system itself) been the same (about 2000 euro) but power consumption in matter of electricity been 10 times more in Opteron side. Also Opteron operate with RAM at DDR3 1333 MHz frequency, while Intel I7 Mobile at 3800 Mhz DDR4. This may also matter and looks obvious.
Thanks.

i'm curious why you are running the ram at ddr3-1333 when OCNG gives you the ability to run at ddr3-1600 with fast timings?

PS: as far as being able to run higher frequencies - i'm running Opteron 6328s (these have an all core turbo of 3.5 GHz and single core turbo of 3.8 GHz). If i just run them with C6 normally. They'll pretty much stay at 3.5 GHz with a very rare occasional spike to 3.8 GHz. i can either lock them at this state with turionpowercontrol -psmax 1, or using OCNG. i have not yet tried upping the base clock yet past the stock 200 MHz. The main gain for these (over my previous Opteron 61xx ES at 3.0 GHz) was a slight increase of single core performance, and a big increase in memory performance.
 
i'm curious why you are running the ram at ddr3-1333 when OCNG gives you the ability to run at ddr3-1600 with fast timings?

PS: as far as being able to run higher frequencies - i'm running Opteron 6328s (these have an all core turbo of 3.5 GHz and single core turbo of 3.8 GHz). If i just run them with C6 normally. They'll pretty much stay at 3.5 GHz with a very rare occasional spike to 3.8 GHz. i can either lock them at this state with turionpowercontrol -psmax 1, or using OCNG. i have not yet tried upping the base clock yet past the stock 200 MHz. The main gain for these (over my previous Opteron 61xx ES at 3.0 GHz) was a slight increase of single core performance, and a big increase in memory performance.
I am running LRDIMMS, to fill max RAM matrix capacity as 1 TB, as per Supermicro manual specifications for RAM matrix, where it can be 1333 Mhz frequency maximum. Hence I took it and run it accordingly. I am not ready to try to burn it out if not necessary.
Yet I have on regular desktop too 1333 to 1600 Mhz and I cannot see a huge improvement in performance between these frequencies. I have attached to system a video card Asus Nvidia GeFroce GT 730 and hence UI on Windows 10 is moving just fine. I am not looking for extreme performance still, while I also want to keep system stable, but the general impression is that system is very responsive.
With default Supermicro BIOS 3.5b I can see spikes on 6386 SE to 3,5 Ghz ( on load as long on idle) for what looks half cores ( half from 16 or 12 if 6 cores disabled in BIOS) as I can see, from CPUz and CoreTemp show it both, as Windows Taks Manager seems show an average frequency. I was just wondering if under OC it will reach 3,8 GHz stable, remembring it was same generation to Phenom 1100 T. Yet I am not ready to change default Supermicro BIOS for a trial and error as I really need system stable and motherboard fully working. I was looking to see others experience first.
But 6328 are running default in boost mode to 3,8 Ghz and regular 3,2. Looks you get some average ( https://www.amd.com/en/products/cpu/6328). I don't see what advantage you get with OCNG BIOS then.
It looks 6386 SE behave the same with default Supermicro BIOS!.
 

Attachments

  • DefaultOC_.jpg
    DefaultOC_.jpg
    46.8 KB · Views: 0
  • DefaultOC_idle.jpg
    DefaultOC_idle.jpg
    46.6 KB · Views: 0
I am running LRDIMMS, to fill max RAM matrix capacity as 1 TB, as per Supermicro manual specifications for RAM matrix, where it can be 1333 Mhz frequency maximum. Hence I took it and run it accordingly. I am not ready to try to burn it out if not necessary.
Yet I have on regular desktop too 1333 to 1600 Mhz and I cannot see a huge improvement in performance between these frequencies. I have attached to system a video card Asus Nvidia GeFroce GT 730 and hence UI on Windows 10 is moving just fine. I am not looking for extreme performance still, while I also want to keep system stable, but the general impression is that system is very responsive.
With default Supermicro BIOS 3.5b I can see spikes on 6386 SE to 3,5 Ghz ( on load as long on idle) for what looks half cores ( half from 16 or 12 if 6 cores disabled in BIOS) as I can see, from CPUz and CoreTemp show it both, as Windows Taks Manager seems show an average frequency. I was just wondering if under OC it will reach 3,8 GHz stable, remembring it was same generation to Phenom 1100 T. Yet I am not ready to change default Supermicro BIOS for a trial and error as I really need system stable and motherboard fully working. I was looking to see others experience first.
But 6328 are running default in boost mode to 3,8 Ghz and regular 3,2. Looks you get some average ( https://www.amd.com/en/products/cpu/6328). I don't see what advantage you get with OCNG BIOS then.
It looks 6386 SE behave the same with default Supermicro BIOS!.
Ah, thanks for the memory explanation.
i was previously running Opteron 61xx ES (unlocked) chips. The 61xx ES chips were actually more stable that the current stock 6328s. With the stock 6328s, there are programs that will reboot the machine every time. The first one that comes to mind is the AIDA64 memory test. It gets to the portion that is testing L3, and the machine reboots. i did not have any such issues with the 61xx ES chips. I found the only way i can get it to complete without a reboot was to reduce the processor frequency down by about a GHz (ie set max power state to an artificially low value). i tried multiple things to see what affected it - artificially ran the memory slower (back to 1333), upped the voltage slightly (using Turionpowercontrol). i tried lowering the northbridge speed to 1800. Still made no difference. i don't think its the cooling as i am running the same Noctua coolers as i did before. These 6328s have fewer cores and also run cooler, but it almost seems as if something is triggering a TDP limit, which then causes the board to reboot.

With both my ES chips and the 6328s, the OCNG BIOS has allowed me to run considerably better memory timings. i'm running Ballistix tactical ddr3-1866. Unfortunately, though the processors support running at 1866, the board does not :( Memory performance before and after is dramatically better, and i notice it in memory bound applications. With the 61xx i would just run a turionpowercontrol script to overclock the processors to 3GHz. i also had another script that would overclock every other core to 3.6 GHz that i would occasionally run. The upper limits for the 61xx was my cooling, and its memory performance. i have not attempted to run the 6328s at a higher frequency using OCNG yet.

The 6328s are better than the 61xx ES, as the memory controller seems to be faster. Northbridge is also running at 2000 instead of 1800. Single thread performance at 3.5 GHz is about equal to the 61xx running at 3 Ghz. i can force them to run at 3.5 GHz all the time using turionpowercontrol -psmax 1, however when running them stock i have found no way to force them to 3.8 GHz. i have tried using WSMR to force threads on every other core and they almost never go to 3.8 GHz. Core for core these Piledriver cores are weaker than the K10s, but the memory performance is better which seems to even it out.

PS: do you think you could try running the AIDA64 cache and memory benchmark on your 6386s? i am curious if you have the same reboot issue.

https://aida64.co.uk/user-manual/tools/cache-and-memory-benchmark
 
Last edited:
Ah, thanks for the memory explanation.
i was previously running Opteron 61xx ES (unlocked) chips. The 61xx ES chips were actually more stable that the current stock 6328s. With the stock 6328s, there are programs that will reboot the machine every time. The first one that comes to mind is the AIDA64 memory test. It gets to the portion that is testing L3, and the machine reboots. i did not have any such issues with the 61xx ES chips. I found the only way i can get it to complete without a reboot was to reduce the processor frequency down by about a GHz (ie set max power state to an artificially low value). i tried multiple things to see what affected it - artificially ran the memory slower (back to 1333), upped the voltage slightly (using Turionpowercontrol). i tried lowering the northbridge speed to 1800. Still made no difference. i don't think its the cooling as i am running the same Noctua coolers as i did before. These 6328s have fewer cores and also run cooler, but it almost seems as if something is triggering a TDP limit, which then causes the board to reboot.

With both my ES chips and the 6328s, the OCNG BIOS has allowed me to run considerably better memory timings. i'm running Ballistix tactical ddr3-1866. Unfortunately, though the processors support running at 1866, the board does not :( Memory performance before and after is dramatically better, and i notice it in memory bound applications. With the 61xx i would just run a turionpowercontrol script to overclock the processors to 3GHz. i also had another script that would overclock every other core to 3.6 GHz that i would occasionally run. The upper limits for the 61xx was my cooling, and its memory performance. i have not attempted to run the 6328s at a higher frequency using OCNG yet.

The 6328s are better than the 61xx ES, as the memory controller seems to be faster. Northbridge is also running at 2000 instead of 1800. Single thread performance at 3.5 GHz is about equal to the 61xx running at 3 Ghz. i can force them to run at 3.5 GHz all the time using turionpowercontrol -psmax 1, however when running them stock i have found no way to force them to 3.8 GHz. i have tried using WSMR to force threads on every other core and they almost never go to 3.8 GHz. Core for core these Piledriver cores are weaker than the K10s, but the memory performance is better which seems to even it out.

PS: do you think you could try running the AIDA64 cache and memory benchmark on your 6386s? i am curious if you have the same reboot issue.

https://aida64.co.uk/user-manual/tools/cache-and-memory-benchmark
Please see down below a screenshot with result of run for Aida64 Cache and Memory benchmark. It was no reboot. I made the test twice one after another. However and once again, please note that I have everything stock, including last available BIOS version(3.5b) from Supermicro website for H8QG6. Just to pin-point, while not sure if this matter. From my side I am not ready to go to OCNG just for fun and in case I do not see real advantages.
I am using at this time in my side ridiculous cheap cooling system yet, like 3 euro Arctic 64 AM4 aluminium coolers adapted for G34, just to do the job, because I did not decided yet what cooling coolers and method to use. I am not satisifed at all with what I found on the market these days, wherever expensive or not, and I am working to implement a custom water cooling method for all 4 CPU based DeepCool Maelstrom water. I already got 4 of it, latest on market here, for older model ( see down below picture) as these will cover an G34 Opteron processor size, and for which I have made an adapter to G34 to keep on place. These are cooling indeed very very good in my tests in comparison with an aircooler( I have a Noctua air cooler too and use to test on different AMD processors) , but I want to implement a custom solution with a much bigger liquid recipient and cooler connected in series. I am still working on that, but all my tests show that indeed watercooling with such custom solution provide a higher level and stability for cooling. You may use such alternative for testing purpose if you have it around. If you are proficient, you can use any other liquid based radiator to check if overheating is indeed the problem.
What I got for overheating on this Supermicro motherboard is that when one of the CPU's reach 65 C, motherboard will beep long and stop working with no display, but fans still rotate. You have to hard restart it to work again. To get better signal for iminent overheat, you may activate Early Alert in BIOS for overheat and motherboard will give short beeps before overheat limitation is reached, while that will be early indeed and useful if heating is done not so fast. But then you will know that it is from overheating and not something with CPU chip itself.

By this way, it may be indeed the AMD CPU's based generations. Slight off topic, I used recently to have fun with an old AMD Athlon II X4 620e on an Asus motherboard that I have it. This is based clock 2,7 Ghz, but with OC it goes just easy to 3,4 Ghz with full stability. Checking in a game, performance was indeed increased. So nice almost 25% performance increased just like that from old times AMD chips. A Phenom X6 1100T will reboot at a frequency set above 3,9 Ghz in a stress test. At least that is what I've got on it. As I can't overclock with default Supermicro BIOS an 6386 SE, I can state the fact, but perhaps someone who already did it may have said it already. Yet one of my question previous if anyone identified the limits for OC for this top list Opterons back then.
 

Attachments

  • Aida64_Cache_and_Memory_test.jpg
    Aida64_Cache_and_Memory_test.jpg
    53.5 KB · Views: 0
  • Deepcool Maelstrom cooler.jpg
    Deepcool Maelstrom cooler.jpg
    29.7 KB · Views: 0
Please see down below a screenshot with result of run for Aida64 Cache and Memory benchmark. It was no reboot. I made the test twice one after another. However and once again, please note that I have everything stock, including last available BIOS version(3.5b) from Supermicro website for H8QG6. Just to pin-point, while not sure if this matter. From my side I am not ready to go to OCNG just for fun and in case I do not see real advantages.
I am using at this time in my side ridiculous cheap cooling system yet, like 3 euro Arctic 64 AM4 aluminium coolers adapted for G34, just to do the job, because I did not decided yet what cooling coolers and method to use. I am not satisifed at all with what I found on the market these days, wherever expensive or not, and I am working to implement a custom water cooling method for all 4 CPU based DeepCool Maelstrom water. I already got 4 of it, latest on market here, for older model ( see down below picture) as these will cover an G34 Opteron processor size, and for which I have made an adapter to G34 to keep on place. These are cooling indeed very very good in my tests in comparison with an aircooler( I have a Noctua air cooler too and use to test on different AMD processors) , but I want to implement a custom solution with a much bigger liquid recipient and cooler connected in series. I am still working on that, but all my tests show that indeed watercooling with such custom solution provide a higher level and stability for cooling. You may use such alternative for testing purpose if you have it around. If you are proficient, you can use any other liquid based radiator to check if overheating is indeed the problem.
What I got for overheating on this Supermicro motherboard is that when one of the CPU's reach 65 C, motherboard will beep long and stop working with no display, but fans still rotate. You have to hard restart it to work again. To get better signal for iminent overheat, you may activate Early Alert in BIOS for overheat and motherboard will give short beeps before overheat limitation is reached, while that will be early indeed and useful if heating is done not so fast. But then you will know that it is from overheating and not something with CPU chip itself.

By this way, it may be indeed the AMD CPU's based generations. Slight off topic, I used recently to have fun with an old AMD Athlon II X4 620e on an Asus motherboard that I have it. This is based clock 2,7 Ghz, but with OC it goes just easy to 3,4 Ghz with full stability. Checking in a game, performance was indeed increased. So nice almost 25% performance increased just like that from old times AMD chips. A Phenom X6 1100T will reboot at a frequency set above 3,9 Ghz in a stress test. At least that is what I've got on it. As I can't overclock with default Supermicro BIOS an 6386 SE, I can state the fact, but perhaps someone who already did it may have said it already. Yet one of my question previous if anyone identified the limits for OC for this top list Opterons back then.

Thanks so much for running that. Perhaps one of the thermal measurement devices on one/both of my chips is bad. Tear had mentioned that they had run across this before. i will be replacing these 6328s with others in about two weeks. i am running NH-U9DO A3 coolers. Heat was indeed a problem with my 61xx (K10 based) Opteron ES samples. i could run all of them at 3.0 GHz day in and day out. With IBT, things would hit about 55C or so. Wattage at the wall was about 1000 or so - so getting close to the 60C limit. Voltage at 3.0 GHz for complete stability was 1.2v. The 6328s run extremely cool in comparison. i think the max i've seen yet in HwInfo has been about 27C. i will try activating the motherboard overheat warnings.

That is a very interesting cooling solution you've got there though. You have got me thinking what i might do to mount those in my SC748 case. Since i'm currently only running 2 of the Opterons, perhaps one solution would be to mount the fans above the bottom two sockets. Was it easy to make an adapter to mount them to the G34 mounting nuts?

As far as your OT comment. i myself like this older machinery for some reason. i've been thinking of replacing it with a SuperMicro Dual EPYC mobo, but i think i will just tinker with it for now. The K10 cores in my Opteron 61xx i was able to run up to about 3.8 GHz (on some cores only). Others could only run 3.6 GHz. i would run 3 of the cores on each of the 6 core dies at the higher speed and the other 3 at 1.4 GHz to spread the heat out. i would then use WSMR to assign the applications to only those fast cores. In the coming weeks i'll be reporting more here as far as how high i can get the 63xx chips running.

One other thing interesting is that when i do "turionpowercontrol -htstatus" i see that with the Opteron 6328s, that the HT speed is only 2600 vs 3200 with the 61xx ES chips. Makes me wonder if any improvements can be had by upping that.
 
Thanks so much for running that. Perhaps one of the thermal measurement devices on one/both of my chips is bad. Tear had mentioned that they had run across this before. i will be replacing these 6328s with others in about two weeks. i am running NH-U9DO A3 coolers. Heat was indeed a problem with my 61xx (K10 based) Opteron ES samples. i could run all of them at 3.0 GHz day in and day out. With IBT, things would hit about 55C or so. Wattage at the wall was about 1000 or so - so getting close to the 60C limit. Voltage at 3.0 GHz for complete stability was 1.2v. The 6328s run extremely cool in comparison. i think the max i've seen yet in HwInfo has been about 27C. i will try activating the motherboard overheat warnings.

That is a very interesting cooling solution you've got there though. You have got me thinking what i might do to mount those in my SC748 case. Since i'm currently only running 2 of the Opterons, perhaps one solution would be to mount the fans above the bottom two sockets. Was it easy to make an adapter to mount them to the G34 mounting nuts?

As far as your OT comment. i myself like this older machinery for some reason. i've been thinking of replacing it with a SuperMicro Dual EPYC mobo, but i think i will just tinker with it for now. The K10 cores in my Opteron 61xx i was able to run up to about 3.8 GHz (on some cores only). Others could only run 3.6 GHz. i would run 3 of the cores on each of the 6 core dies at the higher speed and the other 3 at 1.4 GHz to spread the heat out. i would then use WSMR to assign the applications to only those fast cores. In the coming weeks i'll be reporting more here as far as how high i can get the 63xx chips running.

One other thing interesting is that when i do "turionpowercontrol -htstatus" i see that with the Opteron 6328s, that the HT speed is only 2600 vs 3200 with the 61xx ES chips. Makes me wonder if any improvements can be had by upping that.
From my side I see just not efficient enough air and water coolers for this in what was top G34 Opterons CPU's, as it looks it reach closed to limit temperature very fast, like in seconds or just minutes, in any case of coolers that I tested, for a considerable processing load as demanded by specific applications. I do not see acceptable to run a desktop CPU closed to it's top limits temperature as normal. I have an Intel I7 gen3 CPU o an HP notebook and it runs in games at 90C. But I may find that acceptable for a notebook while it is difficult to use a different cooling method anyway. Yet I do not plan to use in my side full power of Opterons, and hence even if this is custom BIOS for OC discussion, I am closed to give up eitirely for that, as I was not able to find real advangages in such case of OC, but also I do not want to worry all the time about cooling of the CPU's. The Noctua NH-U9DO A3 are not to be found anymore nor on Ebay. Yet I tested an AM3+ Noctua coller adapted to G34 and it was easy reach 55C while looks abovios that Opterons are just heating very much and very fast. This is no wonder that servers fans are running at 16 000 rpms to cool down such CPU's.

Was easy for me to adapt to G34 to mounting system for cooler. I used for this an aluminium L profile on AM4 original mount bases, but due to cooler pipe plugs entire cooler needs to be shifted sideway opposite about 5 mm as screw and spring will hit the central pipe plug. See example in photos on dual mo., but I did put original screws as I have it on quad mo. Note that same screws and springs needs to be on both sides as is mandatorry and it is very very important to have same exact push force in both sides in order to keep right balance and so cooler straight on surface of the CPU, even if enough thermal paste is in place. Yet will be no problem as cooler is wide enough to cover all CPU surface. Based my tests in full load, it was no problem with that, while the liquid cooling was indeed very nice and constant. The liquid coolers connected in a series will have enough power even with only one pump running to flow the liquid, but pumps for this brand are very silent anyway as I could see. This looks indeed a real solution in place of insane fans speed and noice, but in my side I am still testing for opimal way to cool the liquid, as I want an alternative in the office for noisy and windy fans on radiators from factory solution.

BTW. Regarding specific CPU going high temperaturs, if you are not 100% sure that CPU have indeed a physical problem, I will advice to ensure that cooler is straigh enough on the CPU surface. I had one cooler 1mm up one sidedue to a missing washer and CPU was running double temperatura from the others.
 

Attachments

  • mounted.jpg
    mounted.jpg
    349.8 KB · Views: 0
  • mount_custom_extensions.jpg
    mount_custom_extensions.jpg
    396.1 KB · Views: 0
From my side I see just not efficient enough air and water coolers for this in what was top G34 Opterons CPU's, as it looks it reach closed to limit temperature very fast, like in seconds or just minutes, in any case of coolers that I tested, for a considerable processing load as demanded by specific applications. I do not see acceptable to run a desktop CPU closed to it's top limits temperature as normal. I have an Intel I7 gen3 CPU o an HP notebook and it runs in games at 90C. But I may find that acceptable for a notebook while it is difficult to use a different cooling method anyway. Yet I do not plan to use in my side full power of Opterons, and hence even if this is custom BIOS for OC discussion, I am closed to give up eitirely for that, as I was not able to find real advangages in such case of OC, but also I do not want to worry all the time about cooling of the CPU's. The Noctua NH-U9DO A3 are not to be found anymore nor on Ebay. Yet I tested an AM3+ Noctua coller adapted to G34 and it was easy reach 55C while looks abovios that Opterons are just heating very much and very fast. This is no wonder that servers fans are running at 16 000 rpms to cool down such CPU's.

Was easy for me to adapt to G34 to mounting system for cooler. I used for this an aluminium L profile on AM4 original mount bases, but due to cooler pipe plugs entire cooler needs to be shifted sideway opposite about 5 mm as screw and spring will hit the central pipe plug. See example in photos on dual mo., but I did put original screws as I have it on quad mo. Note that same screws and springs needs to be on both sides as is mandatorry and it is very very important to have same exact push force in both sides in order to keep right balance and so cooler straight on surface of the CPU, even if enough thermal paste is in place. Yet will be no problem as cooler is wide enough to cover all CPU surface. Based my tests in full load, it was no problem with that, while the liquid cooling was indeed very nice and constant. The liquid coolers connected in a series will have enough power even with only one pump running to flow the liquid, but pumps for this brand are very silent anyway as I could see. This looks indeed a real solution in place of insane fans speed and noice, but in my side I am still testing for opimal way to cool the liquid, as I want an alternative in the office for noisy and windy fans on radiators from factory solution.

BTW. Regarding specific CPU going high temperaturs, if you are not 100% sure that CPU have indeed a physical problem, I will advice to ensure that cooler is straigh enough on the CPU surface. I had one cooler 1mm up one sidedue to a missing washer and CPU was running double temperatura from the others.

i have some 63xx ES chips coming to me in a few days. i will report the differences in temps between these and the 6328s that i'm currently running. When i was running 61xx ES at full bore, temps stabilized at around 55C.
 
i have some 63xx ES chips coming to me in a few days. i will report the differences in temps between these and the 6328s that i'm currently running. When i was running 61xx ES at full bore, temps stabilized at around 55C.
My understanding is that you already run the best air coolers, as Noctua's. I am currions if higher TDP will still be handled by your coolers.
With cheap Apline 64 GT I can't run full load more than a single minute as it goes close to 65C. Yet not supposed to anyway, as these are 70W coolers while Opteron 6386 SE is 140W.
https://www.arctic.de/en/a64gt
This works for me as idle 25C and average load under 55C. This is for 400W AC power average load for all quad.
While using DeepCool Maelstrom liquid I can see temperature going 10C+ above idle in full load, to about 28-30C maximum. Then it depends by how hot liquid goes without been cooled.
 
My understanding is that you already run the best air coolers, as Noctua's. I am currions if higher TDP will still be handled by your coolers.
With cheap Apline 64 GT I can't run full load more than a single minute as it goes close to 65C. Yet not supposed to anyway, as these are 70W coolers while Opteron 6386 SE is 140W.
https://www.arctic.de/en/a64gt
This works for me as idle 25C and average load under 55C. This is for 400W AC power average load for all quad.
While using DeepCool Maelstrom liquid I can see temperature going 10C+ above idle in full load, to about 28-30C maximum. Then it depends by how hot liquid goes without been cooled.

For me the Noctuas worked much better than expected. Back when i was running confirming stability by running IntelBurnTest for hours and hours on end, with all 48 cores of the 61xx ES at 3.0 GHz, IIRC power draw at the wall was around just under 1000 watts. Temps stabilized at around 55-56. Just under the 60C limit. So i knew i was okay. Of course i did a few things to help the coolers work better. i updated the intake fans to the latest Noctua PWM fans (they run at 2k rpm and better flow than the ones that came). On my SC748 case, i removed the drive carrier and installed a 120mm fan. At the rear i also added a SuperMicro GPU exhaust fan to pull more air out. I also added a fan controller to run all the fans at optimum speeds (had to run the internal fans at different speeds, because i found that they affected the intake airflow to the Noctua coolers, which kept them from being consistent).

i would love to try those Maelstroms. We'll see how it goes with the 63xx ES chips.

i would be happy if i can run the 63xx at 4 GHz.
 
i have some 63xx ES chips coming to me in a few days. i will report the differences in temps between these and the 6328s that i'm currently running. When i was running 61xx ES at full bore, temps stabilized at around 55C.
I ran hyper 212s on my 6174's and when I upgraded AND overclocker my 6378's, I didn't notice any temperature difference.
 
For me the Noctuas worked much better than expected. Back when i was running confirming stability by running IntelBurnTest for hours and hours on end, with all 48 cores of the 61xx ES at 3.0 GHz, IIRC power draw at the wall was around just under 1000 watts. Temps stabilized at around 55-56. Just under the 60C limit. So i knew i was okay. Of course i did a few things to help the coolers work better. i updated the intake fans to the latest Noctua PWM fans (they run at 2k rpm and better flow than the ones that came). On my SC748 case, i removed the drive carrier and installed a 120mm fan. At the rear i also added a SuperMicro GPU exhaust fan to pull more air out. I also added a fan controller to run all the fans at optimum speeds (had to run the internal fans at different speeds, because i found that they affected the intake airflow to the Noctua coolers, which kept them from being consistent).

i would love to try those Maelstroms. We'll see how it goes with the 63xx ES chips.

i would be happy if i can run the 63xx at 4 GHz.
That's just close to temp limit. I won't be so confortabil to run anything while closed to temp limit. This is why I still mess with Maelstrom liquid coolers. Yet in my case I still review if necessary to use it as I am not necessary run full load on CPU's.
I wonder if 6300's will handle the same. Full load at least 950W non redundant on AC. On idle power consumption is about 220W with power saving featues activated on BIOS. Not bad for quad dyno Opterons as an Athon X6 ride almost that power alone, but of course idle is more or less important. However I am using it as workstation, so it may still matter.
I barely think 6300's will run at 4Ghz stable, but I will be curious. Yet looks I am giving up in my side anyway as I am satisfined so far with CPU's performance.
 
I ran hyper 212s on my 6174's and when I upgraded AND overclocker my 6378's, I didn't notice any temperature difference.
6300 are newer with same or lower TDP I think. Then technology for transistors is lower size on newer ones 6300's, but even is with lower techology, current ones, at least in AMD side, have now TDP even up to 400W.
I barely thing that old, or even actual air coolers, will face that. Of course except servers air coolers which run on 16 000 rmps or above... .
It looks we are still doomed on workstation's CPU's to have either very big either insane noizy air coolears. Then liquid coolears need much more imprevements.
 
6300 are newer with same or lower TDP I think. Then technology for transistors is lower size on newer ones 6300's, but even is with lower techology, current ones, at least in AMD side, have now TDP even up to 400W.
I barely thing that old, or even actual air coolers, will face that. Of course except servers air coolers which run on 16 000 rmps or above... .
It looks we are still doomed on workstation's CPU's to have either very big either insane noizy air coolears. Then liquid coolears need much more imprevements.
i just recieved my 63xx Extra Spicy chips today. Will be putting them in over this coming weekend. In the meantime my 6328s have been running just fine (ie completely stable at 211 ref clock) - 3.7 GHz all core turbo, and 4.0 GHz turbo. Will report back later.
 
Back
Top