Overclock.net banner
Cancel
Create thread New Forums

[Official] Zen 4 X3D Owner's Club (7800x3D / 7900x3D / 7950x3D)

Tags
3d v-cache 7800x3d 7900x3d 7950x3d amd
5 reading
Jump to Latest
5,121 - 5,140 of 9,076 Posts
#5,121 · (Edited)
LtMatt said:
I don't think so, but it's been so long since I used a BIOS like 1415 or the betas that were around at the time I've not really kept track.
Click to expand...
man i lose 10fps in WDL, around 10 cpu average in MW3 benchmark, 8 in SOTTR

Warzone seems to work the same though but WDL in game OMG like nearly 20 fps 1%.

what's so baffling is that CB23 is tiny better and aida64 is .5ns better with 2200fclk which is stable now, .4ns with 2167fclk.

i went back to 1415 after all day yesterday messing with 1709 and 1801

tried with new chipsets drivers and without

even TimeSpy Extreme performance about is +1% worse
 
#5,127 ·
HappySheep said:
removed chipset drivers and got half of the performance back in WDL 🤯
Click to expand...
Does it take forever to uninstall the chipset driver? Mine seems stuck at "Checking your Pc's hardware for driver compatibility".

Edit: nvm i got thru right after i post. It does take awhile.

Edit: Btw, uninstall gpu n chipset driver cause mismatch. Pretty sure i have fixed mine w lot of testing, cold boot, shutdown etc..Has anyone experienced the same thing?
 
#5,128 ·
eighty20 said:
Does it take forever to uninstall the chipset driver? Mine seems stuck at "Checking your Pc's hardware for driver compatibility".

Edit: nvm i got thru right after i post. It does take awhile.

Edit: Btw, uninstall gpu n chipset driver cause mismatch. Pretty sure i have fixed mine w lot of testing, cold boot, shutdown etc..Has anyone experienced the same thing?
Click to expand...
tbh I think new bioses have nerfed 1:1 memory compared to 1415 because nothing i change or do gets me my performance back, even TimeSpy Extreme performs worse on 1709 for me
 
#5,131 · (Edited)
eighty20 said:
Maybe try 1801.
Click to expand...
eighty20 said:
Maybe try 1801.
Click to expand...
i have but same results

after reinstall 1709 again now mw3 benchmark seems to have regained most of fps but still worse, WDL still about 2-3% worse, SOTTR still 2% worse and TimeSpy Extreme still about 1% worse....

2200fclk seems stable now even with 1.19soc, what kind of sacrifices did Asus make hmmmmm......



EDIT: Fixed it by changing CPU speculative from Auto to Less then saved it then went back and changed it to Auto again. All my fps is back now...

2nd Fix: Reflashed bios
-used 7zip for bios extract instead of windows one
-CMOS 30secs
-Restore default optimized bios settings SAVE
-EZflash
--Restore default optimized bios settings SAVE
-Load CMO and enable Nitro settings

RESULT: better fps then with 1415 bios
 
#5,134 ·
Sorry for the noob question but if my 7950X3D runs at -25 without any reboots at idle, should I just use it like that or are there other tests I need to run? Like, will boost speeds be affected negatively by not having enough voltage even if the system is stable.
 
#5,135 ·
Chodiddy said:
Doesn't the 7950X3D only use 120W at most?
Click to expand...
Should be 150+w problem is the gpu. If I bench after a gaming session my scores decrease a tad, when the temps cool down again it starts performing higher again. In the old good times of the i9900k the performance were set in stone no matter the temps as long as the cpu was stable. I recently come up with a solution of my particular build to improve the airflow to the radiators and I am scoring higher lol
 
#5,136 ·
Chodiddy said:
Sorry for the noob question but if my 7950X3D runs at -25 without any reboots at idle, should I just use it like that or are there other tests I need to run? Like, will boost speeds be affected negatively by not having enough voltage even if the system is stable.
Click to expand...
You are extremely lucky to have a 7950X3D that runs -25 all core at idle without randomly restarting under light load. I've tried many and most cannot even do -10 all core.

That said you should definitely check stability using core cycler, y cruncher etc.
 
#5,137 ·
LtMatt said:
You are extremely lucky to have a 7950X3D that runs -25 all core at idle without randomly restarting under light load. I've tried many and most cannot even do -10 all core.

That said you should definitely check stability using core cycler, y cruncher etc.
Click to expand...
Yeah, I was really expecting it to restart at -25 but it never did. I just have it at -20 tho just in case. I also tried disabling CCD1 and using the CPU like a 7800X3D. In that mode, I can do -30 with no restarts but that is probably more common for a solo X3D CPU.
 
#5,138 ·
Chodiddy said:
Sorry for the noob question but if my 7950X3D runs at -25 without any reboots at idle, should I just use it like that or are there other tests I need to run? Like, will boost speeds be affected negatively by not having enough voltage even if the system is stable.
Click to expand...
Chodiddy said:
Yeah, I was really expecting it to restart at -25 but it never did. I just have it at -20 tho just in case. I also tried disabling CCD1 and using the CPU like a 7800X3D. In that mode, I can do -30 with no restarts but that is probably more common for a solo X3D CPU.
Click to expand...
You should test stability with :
  • Corecycler « y-cruncher Kizuna » test, at least 4 iterations.
  • y-cruncher « 1-7-0 » at least 10 iterations.
  • Aida64 « FPU Julia » and « SHA3 » tests.
  • your favorite games for hours.
 
  • Rep+
Reactions: BlueShark and LtMatt
Save
Reply Quote Rep+ Rep+
#5,139 ·
I am looking for a golden sample 7950X3D.
Thanks !
 
  • Rep+
  • Love
Reactions: BlueShark and LtMatt
Save
Reply Quote Rep+ Rep+
#5,140 · (Edited)
HappySheep said:
EDIT: Fixed it by changing CPU speculative from Auto to Less then saved it then went back and changed it to Auto again. All my fps is back now...
Click to expand...
I know what that "CPU Speculative Mode" setting does if anyone is interested. In short, it changes how often local cache changes are reflected to other cores and ccds caches.

The instruction pipeline in each core works by executing instructions in advance (that is: speculatively), when the time finally comes to confirm the execution of an instruction the caches are checked to see if anything changed somewhere else and if so the impacted caches are updated with the latest data and the result of the instruction is recalculated, if nothing changed in the scope of the current instruction instead the pipeline commits the previously anticipated result and passes to the next instruction.

When a core changes the content of a memory location, this change can be immediately sent to the caches of all the other cores and ccds so that they can consider the latest data in their speculations (that is the "More Speculative" setting), or wait that the cache on the other cores is just available to accept the update (that is the "Balanced" setting), or wait that the speculations on the other cores need to actually consider that change (that is the "Less Speculative" setting). But sending these cache updates to the other cores and ccds too often wastes IF bandwidth and may also fill the remote caches with data that will not even be needed. And sending these updates too late will require the pipeline of the remote cores to stall waiting for their cache to be updated, if anything changed and their speculatively anticipated result needs to be recalculated with the latest data.

So, for videogames and normal usage, the "Balanced" setting produces the best results (I believe Auto sets it to Balanced). However for particular applications where many threads, especially on different CCDs, continuously read and write data in the same small memory space then the "More Speculative" setting may be some few percent points better (as it will send cache updates pre-emptively, but will waste IF bandwidth). And I can't think of any case in which the "Less Speculative" setting would make sense, as the CPU speculations would suffer a lot, unless you need even the last bit of IF bus availability (like maybe in the AIDA64 bandwidth benchmark maybe ?).
 
5,121 - 5,140 of 9,076 Posts
About this Discussion
9.1K Replies
540 Participants
Last post:
MarkPost
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling