Question Zen 4 builders thread

[Markfw]

I know its early, but as someone pointed out, I better get my ram now before the scalpers are all over it.

To go with a 7950x, what do you think of this ram:

https://www.newegg.com/g-skill-32gb-288-pin-ddr5-sdram/p/N82E16820374369?Item=N82E16820374369

(edited for my best recommendation, 12/5/2023)

 

[biostud]

Crap, it happened again, about 3,5 hours after the first reboot. This time i was watching utube video and shortly before updated GPU driver, as i wanted to install Modern Warfare III beta. Actually considered to reboot anyway, but wanted to finish watching the video.
No bugcheck again.

Last time i had random reboots it was because of faulty memory. But there were bluescreens back then. Could it still be memory related? Or since the screen goes black, could it be GPU?

Specs+bios version?

Any overclock?

 

[Det0x]

Some new options in AMD AGESA 1.0.0.8

CPU speculative store modes: (cannot find any meaningful performance difference between "less" and "more"


AVX512 enabled/disabled:

 

[Det0x]

New agesa also give the opportunity to run faster FCLK on some CPUs:

Have done a performance/FCLK comparison between amd agesa 1.0.8.0 and 1.0.7.0 on my air cooled living room computer as its a prime candidate to gain much for this new bios generation.
The 7950X3D in this machine have previous been limited to only 2033mhz FCLK stable, so it could stand to gain some frequency there 👍 (previously would not boot 2100mhz FCLK)

All my tests have been performed with the following settings:

• -15 allcore PBO CO
• 6000MT/s EXPO memory timings
• Auto VSOC(~1.25v)
• Noctua NH-D15 air cooler

AGESA 1.0.7.0

2033mhz FCLK = no performance throttling -> Linpack xtreme = 771.3 GFLOPS average and 245.3mb/s in karhu ramtest


2066mhz FCLK = clear performance throttling -> Linpack xtreme = 745.3 GFLOPS average and 221.5mb/s in karhu ramtest


AGESA 1.0.8.0

2033mhz FCLK = no performance throttling -> Linpack xtreme = 771.3 GFLOPS average and 245.4mb/s in karhu ramtest


2066mhz FCLK = no performance throttling -> Linpack xtreme = 774.3 GFLOPS average and 245.4mb/s in karhu ramtest


2133mhz FCLK = no performance throttling -> Linpack xtreme = 772.4 GFLOPS average and 246.7mb/s in karhu ramtest
(room getting warmer and alittle rng)


2166mhz FCLK = some performance throttling -> Linpack xtreme = 766.8 GFLOPS average and 241mb/s in karhu ramtest


2166mhz FCLK = small performance throttling -> Linpack xtreme = 772 GFLOPS average and 241.6mb/s in karhu ramtest
(dropping vsoc down to 1.2v lessened the FCLK throttling a tad)


So i can draw two conclusions from this testing:

1. The performance is right there where it should be, pretty much identical between the two agesa's when running same settings
2. This 7950x3d in particular have gained around 100mhz FCLK capability, making it a prime candidate for my main rig, since it have a very good/high boosting CCD2 (40.6k in cinebench r23)

 

[Det0x]

Little followup, have flashed my other system with currently a 7800X3D installed... Strangely the new bios made zero FCLK difference on anything on this system
I'm still capped to 2166mhz FCLK max just like before. (2200FCLK and 8200MT/s wont boot)


And since that i have installed the "living room 7950x3d" back in my main system... When running high memory speeds in 2:1 mode it seems to be limited to 2133mhz FCLK in this system..
But its better than before with new agesa anyway, since it could only run 2066mhz FCLK stable the last time i tried on 1.0.0.7

 

[Timmah!]

Specs+bios version?

Any overclock?

7950x, 64GB Trident 6000MHz, 2x4090, Seasonic 1600W Titanium PSU, still BIOS 805.

There is overclock, combination of PBO and dynamic OC switcher to 5,25/5,175 GHz during all-core loads like cinebench. I can post almost exact step by step, what i did.
Has been rock-solid for last 4,5 months at everything i threw at it, browsing, rendering, 3D modeling, ACAD, photoshop, gaming..., but admittedly i never thoroughly tested it for stability. So if its the OC, its interesting it started to be unstable just out of blue. Is that normal occurence?
If it happens again, i will revert to optimized defaults to see if its the OC.

BTW played 4 matches of MW3 beta since that second crash, for about 40 minutes, and everything was fine. So i presume GPU failing is not the reason (thankfully).

 

[igor_kavinski]

So i presume GPU failing is not the reason (thankfully).

Inspect the GPU connector for signs of charring/melting.

 

[Ajay]

7950x, 64GB Trident 6000MHz, 2x4090, Seasonic 1600W Titanium PSU, still BIOS 805.

There is overclock, combination of PBO and dynamic OC switcher to 5,25/5,175 GHz during all-core loads like cinebench. I can post almost exact step by step, what i did.
Has been rock-solid for last 4,5 months at everything i threw at it, browsing, rendering, 3D modeling, ACAD, photoshop, gaming..., but admittedly i never thoroughly tested it for stability. So if its the OC, its interesting it started to be unstable just out of blue. Is that normal occurence?
If it happens again, i will revert to optimized defaults to see if its the OC.

BTW played 4 matches of MW3 beta since that second crash, for about 40 minutes, and everything was fine. So i presume GPU failing is not the reason (thankfully).

So, easy tests to do. Run a stress test (mem/CPU separately and then together - see where it fails (if it does). If it fails, remove OC on RAM or CPU respectively and retest with stress test. Another option is to boot up with MemTest and just check the memory for errors. It's a PITA, but that's the way it goes. If you have a multi-meter, you can check the PSU voltages under load - I've had them fail and they cause all sorts of odd errors.

 

[biostud]

7950x, 64GB Trident 6000MHz, 2x4090, Seasonic 1600W Titanium PSU, still BIOS 805.

There is overclock, combination of PBO and dynamic OC switcher to 5,25/5,175 GHz during all-core loads like cinebench. I can post almost exact step by step, what i did.
Has been rock-solid for last 4,5 months at everything i threw at it, browsing, rendering, 3D modeling, ACAD, photoshop, gaming..., but admittedly i never thoroughly tested it for stability. So if its the OC, its interesting it started to be unstable just out of blue. Is that normal occurence?
If it happens again, i will revert to optimized defaults to see if its the OC.

BTW played 4 matches of MW3 beta since that second crash, for about 40 minutes, and everything was fine. So i presume GPU failing is not the reason (thankfully).

Which BIOS/AGESA version are you currently running? Are you using memory context restore?

 

[JoeRambo]

AVX512 enabled/disabled:

Awesome, at long last. We have one workload that is not using AVX512 at all, but getting actively hurt by increased context switch costs. Looking forward to Gigabyte releasing it for our B650M's

 

[mmaenpaa]

7950x, 64GB Trident 6000MHz, 2x4090, Seasonic 1600W Titanium PSU, still BIOS 805.

There is overclock, combination of PBO and dynamic OC switcher to 5,25/5,175 GHz during all-core loads like cinebench. I can post almost exact step by step, what i did.
Has been rock-solid for last 4,5 months at everything i threw at it, browsing, rendering, 3D modeling, ACAD, photoshop, gaming..., but admittedly i never thoroughly tested it for stability. So if its the OC, its interesting it started to be unstable just out of blue. Is that normal occurence?
If it happens again, i will revert to optimized defaults to see if its the OC.

BTW played 4 matches of MW3 beta since that second crash, for about 40 minutes, and everything was fine. So i presume GPU failing is not the reason (thankfully).

I would test memory first. Memory also can "age". Easiest is Karhu ramtest, but is not free. I believe memory testing while running windows shows errors better than standalone bootable memory test (like Memtest & Memtest86+).

I had one EXPO kit which would show errors (in Karhu) while running at expo settings (6000MHz).
Changing first CL30->CL36 solved that problem.

 

[Timmah!]

Which BIOS/AGESA version are you currently running? Are you using memory context restore?

BIOS 805, dont know its AGESA number. Its one of the early ones, from last year, but used to be considered better and more stable than bunch of bioses, that came later, those 1000+ ones. Probably have changed since then, but since it worked for me without issue, did not see the reason to meddle with it.

I would test memory first. Memory also can "age". Easiest is Karhu ramtest, but is not free. I believe memory testing while running windows shows errors better than standalone bootable memory test (like Memtest & Memtest86+).

I had one EXPO kit which would show errors (in Karhu) while running at expo settings (6000MHz).
Changing first CL30->CL36 solved that problem.

Thanks. So what you are saying its not un-common for it to run couple of months just fine at certain settings (basically OC ones, since i run at 6000 MHz) and then "age" and needing slight setting downgrade, as it wont be stable at those original settings anymore?
Would RAM errors manifest themselves as those hard reboots, i mean without BSODs or event bugchecks? As i said before, the only time i ever had faulty ram on my previous rig, and diagnosed with MemTest eventually, it was similar, random crashes while say watching hockey stream - but then i was getting BSODs.

Just turned on the computer today. We shall see what happens next.

 

[mmaenpaa]

BIOS 805, dont know its AGESA number. Its one of the early ones, from last year, but used to be considered better and more stable than bunch of bioses, that came later, those 1000+ ones. Probably have changed since then, but since it worked for me without issue, did not see the reason to meddle with it.

Thanks. So what you are saying its not un-common for it to run couple of months just fine at certain settings (basically OC ones, since i run at 6000 MHz) and then "age" and needing slight setting downgrade, as it wont be stable at those original settings anymore?
Would RAM errors manifest themselves as those hard reboots, i mean without BSODs or event bugchecks? As i said before, the only time i ever had faulty ram on my previous rig, and diagnosed with MemTest eventually, it was similar, random crashes while say watching hockey stream - but then i was getting BSODs.

Just turned on the computer today. We shall see what happens next.

Usually if reboot is "hard" ie. it does not leave any memory dumps it is hardware fault not OS/driver (as OS does not have time to dump memory for diagnostics).

You can check that with for example WhoCrashed tool: https://www.resplendence.com/downloads

Ram testing is like in networking if there are network slowdowns / problems.

It’s not DNS
There’s no way it’s DNS
It was DNS

🤓

 

[biostud]

BIOS 805, dont know its AGESA number. Its one of the early ones, from last year, but used to be considered better and more stable than bunch of bioses, that came later, those 1000+ ones. Probably have changed since then, but since it worked for me without issue, did not see the reason to meddle with it.

Thanks. So what you are saying its not un-common for it to run couple of months just fine at certain settings (basically OC ones, since i run at 6000 MHz) and then "age" and needing slight setting downgrade, as it wont be stable at those original settings anymore?
Would RAM errors manifest themselves as those hard reboots, i mean without BSODs or event bugchecks? As i said before, the only time i ever had faulty ram on my previous rig, and diagnosed with MemTest eventually, it was similar, random crashes while say watching hockey stream - but then i was getting BSODs.

Just turned on the computer today. We shall see what happens next.

Personally I would update the bios to the latest non beta, to see if it would help. The 805 (I assume it is an ASUS board) is from before SoC voltage protection, it might be a far shot but could it be related.

 

[Timmah!]

Usually if reboot is "hard" ie. it does not leave any memory dumps it is hardware fault not OS/driver (as OS does not have time to dump memory for diagnostics).

You can check that with for example WhoCrashed tool: https://www.resplendence.com/downloads

Ram testing is like in networking if there are network slowdowns / problems.

It’s not DNS
There’s no way it’s DNS
It was DNS

🤓

Thanks. Well, thats what i fear, that it might be hardware failure of some sort. But that could mean faulty RAM as well, right?

Will WhoCrashed help in my situation, i mean without memory dump and and bugcheck = 0?

Anyway, so far so good today (knocking on the wood). Played 2 hours of MW3 beta without any issues.

 

[igor_kavinski]

Anyway, so far so good today (knocking on the wood).

Ensure that the SoC voltage stays below 1.3V

Unsure if your version of AGESA does that.

 

[Timmah!]

Ensure that the SoC voltage stays below 1.3V

Unsure if your version of AGESA does that.

Thanks. I remember there was an issue with that few months back ago, but actually, the 805 BIOS, unlike the later ones before the fix has been applied, did not have that problem. I did check back then and per HWinfo, my SoC voltage never went above 1,252 or 1,255V.

One more question, is it possible that some form of power outage would have triggered those reboots yesterday, but it would not cause my alarm clock /iphone dock to reset the clock? Such thing usually occurs during power-outages, but the kind when they last at least for few seconds and lights go off as well. And computer in such cases basically shuts off abruptly and needs to be manually turned on, not just reboots itself.

Or since the clock did not reset, it was definitely not down to electric grid / external power delivery problem.

 

[igor_kavinski]

Could have been a brown out but you can usually see the room lights dim when that happens.

 

[In2Photos]

Thanks. I remember there was an issue with that few months back ago, but actually, the 805 BIOS, unlike the later ones before the fix has been applied, did not have that problem. I did check back then and per HWinfo, my SoC voltage never went above 1,252 or 1,255V.

One more question, is it possible that some form of power outage would have triggered those reboots yesterday, but it would not cause my alarm clock /iphone dock to reset the clock? Such thing usually occurs during power-outages, but the kind when they last at least for few seconds and lights go off as well. And computer in such cases basically shuts off abruptly and needs to be manually turned on, not just reboots itself.

Or since the clock did not reset, it was definitely not down to electric grid / external power delivery problem.

It's possible to have a fault on the electric system that lowers the voltage enough to cause a PC to shutdown without you losing power. As was mentioned before you should have a UPS on your PC. It would prevent a shutdown during these faults and it would log the incident as well. Not saying this is what happened but it is possible.

 

[IEC]

Have a 96GB kit of DDR-6400 CL32 RAM on the way. Corsair PN#CMK96GX5M2B6400C32.

From few posts available online, it should be Hynix 24-bit M-die, which behaves similarly to A-die. I'll see how it runs on the X670E this weekend, time allowing.

 

[Markfw]

7950x number 7 done, soon to get an OS ! Things different about this vs other setups I have. Mini ATX motherboard with only ONE 8 pin CPU power plug. Smallest PSU (750 watt) Cheapest motherboard ($100) B650 !

 

[CakeMonster]

Have a 96GB kit of DDR-6400 CL32 RAM on the way. Corsair PN#CMK96GX5M2B6400C32.

From few posts available online, it should be Hynix 24-bit M-die, which behaves similarly to A-die. I'll see how it runs on the X670E this weekend, time allowing.

Keep us updated. Very curious to see how it runs, and OC if you're trying that. I've been mulling going to 96GB because of hobbyist AI projects, but so far I'm waiting for Z5 and what memory speeds it will support, whether there will be 128GB 2 stick kits in not too long, and hoping memory prices won't increase...

 

[Markfw]

7950x number 8 done. I have not booted it yet, taking a break. Taichi motherboard, 7800 ram and a 4090. I have a 1600 watt PSU on it since I was also going to put a Titan V in there, but it won't fit ! The 4090 is 3.5 spaces wide.

 

[IEC]

Keep us updated. Very curious to see how it runs, and OC if you're trying that. I've been mulling going to 96GB because of hobbyist AI projects, but so far I'm waiting for Z5 and what memory speeds it will support, whether there will be 128GB 2 stick kits in not too long, and hoping memory prices won't increase...

Update: Shipping delay means I won't get it until this weekend. So I'll plan on fiddling with it then. I have a ASUS X670E rig and a ASRock B650 microATX rig (2 DIMMs) to test it out in.

 

[Ajay]

and hoping memory prices won't increase...

They will be going up - but so far it looks like it will happen slowly. So long as *all* the ddr5 mem companies keep output down (not so great for us, but we do need them to stay in business).

 

[IEC]

96GB @ DDR5-6000 on AM5, first attempt. Plenty of room to improve.
6000 FCLK=MEMCLK, vSOC 1.2V, MC 1.34V, VDD, VDDQ 1.4V



Timings adapted from this thread