[BHS1975]

So with my cpu and a b550-F it looks like upping VSOC to 1.15 gets rid of the few WHEAs I had at 1900 FLCK. This is with LLC 1.
I put the other voltages at auto just to see where they would go. Have not tried lowering any of them.
Auto VSOC was at 1.1v, but I still had WHEAs at that level.

View attachment 2560041

The thing is this didn't seem to help improve my 1800 FCLK results in any of the handful of benchmarks I ran. Need more testing, but I will probably settle back on 1800 FLCK for daily use as I can keep voltages and therefore SoC power much lower then.

Yeah just run it at 1866 and tighten up your ram as much as you can and keep ram under 45C. I can run run 1v SOC and vddp etc in the 800s mV.

 

[ice445]

So I just got mine, batch # is 2208PGS. So far I'm really impressed. I came from a 5600X so I wasn't expecting anything massive, but in CS:GO, a game I play often, I went from having great FPS with the occasional dip into the 200's, to literally 399FPS constant. It actually sits on the cap the entire time during a competitive 5V5 match. Insane.

Temps wise I'm even more surprised, I'm using my old DeepCool Gammaxx 400V2 while I wait for my Arctic Freezer II 240 to show up, and I'm seeing around 70C in most games like CS, with my Cinebench R23 all core max temp 84C. Hard to believe for a $30 cooler to be honest. This is in a Corsair 4000D airflow for reference. How much variance are you guys seeing with thermals? I wonder if they changed something or I just got really lucky with binning...

Running this on a B550-F Gaming with 4 sticks/ 32GB of Micron E die ram (3600CL16). Just threw it on DOCP while I was testing and I'll probably just leave it honestly. Cinebench score was surprisingly good for the lower clock speeds, it actually sits at a higher frequency on all core than my 5600X did. Held steady at 4.3Ghz, whereas my 5600X liked to drop to around 4.1 despite much lower thermals (I did not run PBO so I suspect this is because the 5800X3D has a generous power limit and higher efficiency).

 

[OCmember]

That's the second 2208PGS that I've seen. I have one also and the IF seems to be really strong. 1.0000v (minimum value) for 1900 stable

 

[tcclaviger]

Is this normal? My last cpu was a Intel 8700k. I am not familiar on how the new ones are supposed to act.

EDIT. Am I correct in thinking the c state control will fix this behavior?

Not sure, Mine idles at the 36 multiplier, 5950x also idles at 38 multiplier, not sure why. Both draw 0 AMPs at idle and instantly boost up as soon as anything at all is occurring so it doesn't matter much.

I'm also on 2208GPS, yes it's very capable of running low vSOC, CCD and IOD values at or below 1900/3800. It's just less sensitive to heat than the normal Zen 3 chips for all core loads. Similar to the XT line of Zen 2, it pursues all core boost more aggressively than X chips do, in moderate cooling it'll hit the voltage/freq limit before the temperature scaling starts or power limits kick in and only when it keeps climbing after that first couple of speed step reductions will it hit thermal limit and start pulling more.

Like XT chips, it manifests as more consistent speed and more time at max attainable speed, both critical for consistent smooth gaming. It's almost the inverse of X skus which are constantly pursuing downclocking, X3D seems to be constantly looking for any reason to boost to 44.5.

 

[MrHoof]

Balanced mode with power slider on powersaving can drop it to 3ghz but like u said barely a diffrence to 3.6ghz in power usage.

Also i seem to have the oldest batch in here with 2207PGS

 

[OCmember]

Also i seem to have the oldest batch in here with 2207PGS

How does the IF perform in terms of voltage & speed? My 2208 is very nice. I hope I didn't screw anything up by pushing too much vsoc (1.0875v) with 1900. It's currently 1.0000v with DF-State disbled, APBDIS 1, SOC P-States 0, DPM values @ 2 From what I gathered that puts a lot of stress on the IF, correct me if I'm wrong.

really excited with this chip as my old 5800x could only do 1866

 

[MrHoof]

Currently running 1900fclk with 1.0125v SOC 0.95v IOD 0.9 CCD WHEA free.
DF-State , APBDIS, SOC P-States, DPM values all at default.
Zentimings reads 1.0063V and 1v after Vdroop for soc at auto llc.
Wasnt much tuning needed Ram settings i could almost 1/1 copy from my 5800x only tWTRS wants 4 compared to 3 before else I get errors fast.

 

[tcclaviger]

Nice. Seems B2 does really help with signalling integrity.

I impulse bought a 6950xt OC Formula today....and canceled it lol. Going to hold out for 7xxx GPUs. As much as I want to play with RDNA2+X3D, it just doesn't make sense to buy a 6950xt when I already have a 3080ti and 7000s are coming later this year.

 

[MrHoof]

Ye looks like its basicly a 6900xt with a bigger cooler a diffrent less locked bios and 3rd 8pin not impressed.

 

[EastCoast]

Ye looks like its basicly a 6900xt with a bigger cooler a diffrent less locked bios and 3rd 8pin not impressed.

Nope
AMD already has specific drivers for it. They fine tuned Sam for the 6950XT in a few games (called AMD Software Preview Driver May 2022 Release). Further more it's using an improved Samsung memory ic's. That are guaranteed to overclock between 2350mhz to 2450mhz depending on cooling. That's 250mhz to 350mhz higher then most water-cooled 6900xt xtxh variants.

AMD is sandbagging this ktxh variant. As time goes on we will find out what else they've done. But at a guess it is looking like a revised Uarch. Because the gpu clock speeds are the same as the 6900xt xrxh but provides much better performance then just a 16Gbps to 18Gbps vram improvement. But I digress.

Oh. and the 6950xt seems to work best with a 5800x3d. Go figure.

 

[MrHoof]

Aslong this bug is in the 22. branch I wouldnt install it on my system cause its not flicker but frame drops that cause blackscreens on bad freesync setups.
I always run a Stream on the 2nd monitor while gaming so even if it improves fps by x% its not worth the fps drops while using hardware acceleration, I dont get blackscreens on freesync premium pro monitor but noticeable frame drops compared to 21.11.3.

• Display may flicker black during video playback plus gameplay on some AMD Graphics Products such as the Radeon™ RX 6700 XT(or every 6xxx card).

Still waiting for benchmarks with the 5800x3D but i doubt it does Magic but just reducing the cpu bottlekneck.

But enough offtopic

 

[CCoR]

Nice. Seems B2 does really help with signalling integrity.

I impulse bought a 6950xt OC Formula today....and canceled it lol. Going to hold out for 7xxx GPUs. As much as I want to play with RDNA2+X3D, it just doesn't make sense to buy a 6950xt when I already have a 3080ti and 7000s are coming later this year.

Whats B2 stand for?

 

[RedF]

Whats B2 stand for?

Revison/Stepping 2

 

[Taraquin]

even though ram tuning often doesn't improve avg fps by much on 5800X3D, it can work wonders for min fps

 

[ilmazzo]

If average fps is an .... average of min/max fps in a given time period how could not this improve the average number while increasing the min values?

 

[tcclaviger]

Smaller standard devian from the mean. It spends less time at 99% and less time at 1% so averages remain the same but frame pacing is more consistent providing improved consistency. Pretty much exactly what is observed by most people as smooth game play but doesn't show up well in metrics on charts.

EDIT: I do appreciate everyones input in online reviews, I'm not maliciously attacking jufus but...

I wish jufus would post SotTR 1080 lowest with SMT off as a measuring stick for comparison to determine the efficacy of his OCing. I doubt his x3d setup would crack 370fps.

It's another data point but, IMHO Jufus is, at best, mediocre when tuning Zen chips, advocating for static OCs as the best method, so his AMD CPU results should be viewed as the equivalent of low effort CPU OCing.

In his x3d vs AL video he was allegedly running at 4.7ghz on x3d but covered all info except latency on his Aida ram screenshot. 57.4 ns suggest loose timings for a 4.7 OC or massive clock stretching. 3800C14 for me, returns 56.2ns at 4610 all core speeds. 4112 C15 is 53.2ns. I mention this as, if this is his max tuned setup, it creates doubt as to his actual CPU OC efficacy. Notice he showed zero productivity tests as a result of his OC. R23 should have been 16k. You think he scores 16k....???

He claims cooling issues, but during gaming I never see my CPU over 42c CCD1. Add 20c to eliminate chiller and...62c, not exactly hot. Suggests he's brute forcing stability with too much voltage. I know from testing that is the wrong way to OC x3d and leads to stock level performance as FIT throws the brakes on hard.

Compared to a self proclaimed golden sample 12900k running max tuned DDR 5 7000+ with 32 CL and 4133 C15 on a different 12900k that has great DDR4 IMC.

 

[hotlookman]

What is 2208 number u all talking about ?

 

[RedF]

Produktion Week 2022. My X3D is 2207.

 

[domdtxdissar]

Smaller standard devian from the mean. It spends less time at 99% and less time at 1% so averages remain the same but frame pacing is more consistent providing improved consistency. Pretty much exactly what is observed by most people as smooth game play but doesn't show up well in metrics on charts.

EDIT: I do appreciate everyones input in online reviews, I'm not maliciously attacking jufus but...

I wish jufus would post SotTR 1080 lowest with SMT off as a measuring stick for comparison to determine the efficacy of his OCing. I doubt his x3d setup would crack 370fps.

It's another data point but, IMHO Jufus is, at best, mediocre when tuning Zen chips, advocating for static OCs as the best method, so his AMD CPU results should be viewed as the equivalent of low effort CPU OCing.
View attachment 2560237
In his x3d vs AL video he was allegedly running at 4.7ghz on x3d but covered all info except latency on his Aida ram screenshot. 57.4 ns suggest loose timings for a 4.7 OC or massive clock stretching. 3800C14 for me, returns 56.2ns at 4610 all core speeds. 4112 C15 is 53.2ns. I mention this as, if this is his max tuned setup, it creates doubt as to his actual CPU OC efficacy. Notice he showed zero productivity tests as a result of his OC. R23 should have been 16k. You think he scores 16k....???

He claims cooling issues, but during gaming I never see my CPU over 42c CCD1. Add 20c to eliminate chiller and...62c, not exactly hot. Suggests he's brute forcing stability with too much voltage. I know from testing that is the wrong way to OC x3d and leads to stock level performance as FIT throws the brakes on hard.

Compared to a self proclaimed golden sample 12900k running max tuned DDR 5 7000+ with 32 CL and 4133 C15 on a different 12900k that has great DDR4 IMC.

He dont know how to tune Zen3 with or without v-cache.

All his numbers 5800x3d numbers seems too low. His "max clocked" 5800x3d @ "4.7ghz" is scoring lower than my 5800x3d @ stock 4550mhz in gamebenchmarks and why does he hide everything in aida here ? Dont make sense to me.

(i'm getting down to 55.4ns @ 1900:3800 with cpu at 4550mhz without any L2 prefetch shenanigans btw)

Guess i could do some more benchmarks in riftbreaker for a easy comparison but what's the point..

 

[EastCoast]

@tcclaviger @domdtxdissar
I stopped watching him a long time ago. And his videos are not well received at other fourms when posted by randoms. He gets a lot of criticism. So you are not the only one.