• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

Will you be purchasing the 7800X3D on the 6th?

  • Total voters
    191
  • Poll closed .
HW Info confirms is A die (2x32gb)

And I think I have found what caused the massive instability with Buildzoids easy DDR5 settings as a new video was posted specifically for dual sided 2x32gb

Basically tRAS was way off

I can't even set the 40 he recommends for dual sided (yet) but 46 is working aok

I think for single sided he had it set at 28 but said needed 36 to be stable for dual sided - and I haven't even managed that yet I'm at 40 (still an improvement over stock)
 
Last edited:
When adjusting PBO under volts on a per core basis as they relative ?

As in the more tweaked (still stable) the better cores are set to the higher the worse core needs to be set

As in not as low as possible on some cores as could be and the worse core could be fine at -9 but then when the other cores are set to their lowest possible -9 is no longer stable and it needs -5 or higher

I'm finding all Prime95 Small much harder to get stable than core cycler which seems easier to pass even on 3m runs as suggested earlier in thread
 
I have a quick question. I recently bought a 7950X3D for tinkering fun until the 9950x3d gets released. I have two pc's already that were installed using a 9800x3d. If I were to slot 7950x3d in would I have to completely reinstall windows again with it updates as it is generally recommended?

I know it is a good idea to but I am just looking for any excuse to get out of it as I hate it :P
 
Last edited:
GoRedwings19 said:
I have a quick question. I recently bought a 7950X3D for tinkering fun until the 9950x3d gets released. I have two pc's already that were installed using a 9800x3d. If I were to slot 7950x3d in would I have to completely reinstall windows again with it updates as it is generally recommended?

I know it is a good idea to but I am just looking for any excuse to get out of it as I hate it :P
Click to expand...
I think the format and clean install was if you where going from a non X3D chip to a 3D chip with 2 x CCDs as the chipset drivers would not install the X3D services needed to park the extra cores.

Someone else maybe able to confirm if I am right or wrong! lol
 
ED209 said:
I think the format and clean install was if you where going from a non X3D chip to a 3D chip with 2 x CCDs as the chipset drivers would not install the X3D services needed to park the extra cores.

Someone else maybe able to confirm if I am right or wrong! lol
Click to expand...
Right or wrong I like your response as to get out of it. :P
 
Struggling to get anything over fclk 2000 stable on my 7800x3d

Even at 2036 - it will run Prime95 overnight for like 12 hours (RAM is at 6000 cl28) and games fine but then hang in bios

I had it before even worse about 2 weeks back when I tried 2133 it was it was so bad in bios it got to the point it woulen write qnt settings and I had to do a jumper reset

I tried low vsoc down to 1.18 and more sensible volts at 1.26

Voltage is set to manual override vs amd over locking as don't know the difference

MSI Tomahawk latest bios

Don't know why bios starts to so unstable at above 2000 fclk

Current stable is fclk 2000, 6000 cl28 tight Bzoid timings so probably good enough anyway
 
Last edited:
Buckster said:
Struggling to get anything over fclk 2000 stable on my 7800x3d

Even at 2036 - it will run Prime95 overnight for like 12 hours (RAM is at 6000 cl28) and games fine but then hang in bios

I had it before even worse about 2 weeks back when I tried 2133 it was it was so bad in bios it got to the point it woulen write qnt settings and I had to do a jumper reset

I tried low vsoc down to 1.18 and more sensible volts at 1.26

Voltage is set to manual override vs amd over locking as don't know the difference

MSI Tomahawk latest bios

Don't know why bios starts to so unstable at above 2000 fclk

Current stable is fclk 2000, 6000 cl28 tight Bzoid timings so probably good enough anyway
Click to expand...

The easiest way to test FCLK is save you current bios config, reset your bios, just set VSOC to 1.25v and FCLK to 2200, if the system posts then great, if it doesnt then try 2167, again see if the system boots, anything that boots fine, boot into safe mode (sign out of windows and hold the shift key whilst selecting reboot) and test with Linpack extreme 1.18, 10gb RAM, 15 passes, you dont want the GFLOPS to be any further than 2 or 3 apart, anymore shows unstable FCLK, maybe needs more voltage, or maybe needs less FCLK, if that passes theres something else causing instability.

With Buildzoids easy DDR5 timings, ive found I cant run the SCLs at 4 unless im running my default timings for the first 4 timings, anything tighter and I need the SCLs to be at 5 or 6, so on my wifes system im running 28-36-36-30, it needs 1.45v VDD and VDDQ, 1.4v VDDIO and 1.20v SOC, the rest is set to buildzoids timings with SCLs at 5, so if youre running 6000 CL28 then unless you have one of the new G.Skill kits, youre not running your kits first 4 default timings and may need more voltage.
 
Last edited:
Jamin280672 said:
The easiest way to test FCLK is save you current bios config, reset your bios, just set VSOC to 1.25v and FCLK to 2200, if the system posts then great, if it doesnt then try 2167, again see if the system boots, anything that boots fine, boot into safe mode (sign out of windows and hold the shift key whilst selecting reboot) and test with Linpack extreme 1.18, 10gb RAM, 15 passes, you dont want the GFLOPS to be any further than 2 or 3 apart, anymore shows unstable FCLK, maybe needs more voltage, or maybe needs less FCLK, if that passes theres something else causing instability.

With Buildzoids easy DDR5 timings, ive found I cant run the SCLs at 4 unless im running my default timings for the first 4 timings, anything tighter and I need the SCLs to be at 5 or 6, so on my wifes system im running 28-36-36-30, it needs 1.45v VDD and VDDQ, 1.4v VDDIO and 1.20v SOC, the rest is set to buildzoids timings with SCLs at 5, so if youre running 6000 CL28 then unless you have one of the new G.Skill kits, youre not running your kits first 4 default timings and may need more voltage.
Click to expand...

thank you I'll give that a try

have been doing some more PBO -ve optimisations and it is very odd, per core -ve seems to be very dependant on the -ve curve on the worse core

get to a stable point (and no further) with say that worse core at -4, ie you get to a point where -ve point is optimised for each core and pushing any of the other cores past that stable point and prime95 all cores fails one or more cores and over time even causes a complete reboot (after several hours). Make that worse core then -3, and suddenly you can go -2, -3 or even more further past that point on most of the other cores very odd

I'm interested to see how far it goes -I think I posted this previously but whilst I don't under stand this inter-core voltage relationship, working on that worse core and allowing for more negatives on the other cores at a guess is better for gaming but worse for a completely all core load like video rendering

the inter-core voltage could possibly I guess be just a relative voltage, as in if the worse core is set to a "higher" -ve curve (as in -3 vs -4) then maybe the average voltage across the other cores is slightly higher thus allowing better -ves on the other cores so you are not actually gaining anything at all -not sure

small prime sits at just over 4.5 gig when all core running, pulls about 74W and after 10 hours CPU is 56C which is not bad for cheap air cooling !
 
Last edited:
Buckster said:
thank you I'll give that a try

have been doing some more PBO -ve optimisations and it is very odd, per core -ve seems to be very dependant on the -ve curve on the worse core

get to a stable point (and no further) with say that worse core at -4, ie you get to a point where -ve point is optimised for each core and pushing any of the other cores past that stable point and prime95 all cores fails one or more cores and over time even causes a complete reboot (after several hours). Make that worse vecore then -3, and suddenly you can go -2, -3 or even more further past that point on most of the other cores very odd

I'm interested to see how far it goes -I think I posted this previously but whilst I don't under stand this inter-core voltage relationship, working on that worse core and allowing for more negatives on the other cores at a guess is better for gaming but worse for a completely all core load like video rendering

the inter-core voltage could possibly I guess be just a relative voltage, as in if the worse core is set to a "higher" -ve curve (as in -3 vs -4) then maybe the average voltage across the other cores is slightly higher thus allowing better -ves on the other cores so you are not actually gaining anything at all -not sure

small prime sits at just over 4.5 gig when all core running, pulls about 74W and after 10 hours CPU is 56C which is not bad for cheap air cooling !
Click to expand...

You have to find a balance, its a PITA, RAM overclocks eat into PBO, so I just upgraded a couple of weeks ago to a 48gb kit from a 32gb kit, running it at 8000MT/s CL34, the same as my 32gb kit, but ive had to start all over with my CO, im not getting anywhere near the amount of -curve that I had with my 32gb kit.

It my understanding there is only 1 algorithm for the CO to each CCD, so as far as im aware, regardless of if you set 1 core to -20 and one core to -10, its averages it out, so effectively it would try -15 to every core.

Also finally, I had to go through 4-5 7950X3D's until I got a decent one, I ordered some off the rainforest and sent them back, "performance or quality", ordered one off OCuk and got a good one straight away, maybe the rainforest arnt storing them correctly, but it very very pot luck if you get a good chip with a decent IMC in it, 9000 series isnt any better as it uses the same IOD from 7000 series.
 
Last edited:
So anyone can test this theroy really, if you have your best core set way lower than your other cores, say -5 for your best core and -15 for the others, see if you lower your others to -10, if your best core will also do -10 or -8 now.
 
Last edited:
After an epic fail (prob just unlucky) I've decided to keep as is now

After being 10 hours Prime 95 stable, core cycler stable and games no problem

I dropped each of the cores by 1, except for worse core, and it was instable enough that Prime95 caused a hard reboot after 5 seconds

this trashed my Windows 11 install which was not recoverable and I spent a few hours in the evening re-installing Windows 11 !
 
So I have finally got round to installing the 7950x3d I picked up for tinkering fun. Any overclocking tips? Is it similar to the 7800x3d or less wiggle room for overclocking. I see the core boost over ride. Is this like the 7800x3d where you can only get +50 mhz or is it more like the 9800x3d where when you select +200 you can get a freebie 200 mhz boost.
 
GoRedwings19 said:
So I have finally got round to installing the 7950x3d I picked up for tinkering fun. Any overclocking tips? Is it similar to the 7800x3d or less wiggle room for overclocking. I see the core boost over ride. Is this like the 7800x3d where you can only get +50 mhz or is it more like the 9800x3d where when you select +200 you can get a freebie 200 mhz boost.
Click to expand...

Yeah......no, if you cant be bothered just stick it in and enable PBO, I can do +150mhz on mine, but it took a lot of work to find limits and curves, a lot of homework and studying, and with the RAM overclock, took 2 weeks to get fully stable, verified with Core Cycler, Y-Cruncher, Prime95 Blend SSE, AVX2 and AVX512, Memtest Pro, Karhu, and AIDA64 Stress test, all 12hour ish overnight runs, TM5 for initial ram tests as it picks up errors really fast.

nEkzmdb.jpeg


4UbmuF5.jpeg
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom