Overclock.net banner
341 - 360 of 603 Posts

·
Registered
Joined
·
3 Posts
Canadian distributors be like that. I ordered a launch day 5900X at 10:29 in the morning at got my CPU 2 weeks after my friend who ordered at like 5PM. Canada Computers is just like that unfortunately.

My chip from Dec 23rd is pretty poggers. im still testing but it just ran 50MHz boost -30 all core CO. So far no instability issues. Hoping to turn up that boost clock and see where my cores decided to poo out at.

Currently boosts to 4.5GHz all core (hits thermal limit), and 4.9GHz single core (occasionally goes 5ghz in random desktop stuff, but that isnt really doing much for performance).
View attachment 2473867
Looks like you're hitting the power limit. i can get my 5900x to 4.6 sustained all core with pbo and curve optimizer. Try - PPT 210 TDC 130 EDC 185. open ryzen master when benching and see what which one is at 100% You want EDC to be the only one at 100% but dont go higher on the EDC as the cpu will always take more, but you'll be past the point of diminishing returns past 180.
 

·
Registered
Joined
·
27 Posts
On my 5900X, the "best" core doesn't wanna do any negative offset, while all the others (except 2nd best) do as much as -30 without any errors and linear performance increase. Anyone having similar experience?

PPT: 165W
TDC: 125A
EDC: 130A

scalar: auto
freq boost: +50 MHz

2512373
 

·
Registered
Joined
·
553 Posts
On my 5900X, the "best" core doesn't wanna do any negative offset, while all the others (except 2nd best) do as much as -30 without any errors and linear performance increase. Anyone having similar experience?

PPT: 165W
TDC: 125A
EDC: 130A

scalar: auto
freq boost: +50 MHz

View attachment 2512373
My best core needs +10 offset!
 
  • Rep+
Reactions: BarrettDotFifty

·
Registered
Joined
·
27 Posts
After some time playing with the curve I found out that the second CCD is much more efficient, but simply doesnt clock as high. The second CCD works fine with -25 curve on all cores and some could do even better I think. The CCD1 can even run cinebench with -30, but thats about it its not stable with it at all.

On CCD0 its quite different, there is maybe one core that will do -20 stable, others seems like they will need -5,-10,-15. But they all clock higher than all cores on CCD1, so it makes sense they need the voltage more. I cant even get to windows when I set the worst core on this ccd to more than -5.

Both CCD are clearly very different, the first is for high performance bursts and can keep the advertised 4.8ghz boost clock on single thread and the second is nowhere near that even when I am maxing just one thread but its overall more efficient.

In the SSE test with single thread my cores keep clocks around this.
CCD0 (single core) 4,7-4,8ghz
CCD1 (single core) 4,6-4,7ghz

These values are "effective clock" from hwinfo during the test, so not just a blip but really sustained speed. The best 2 cores are usually slightly above 4.8ghz, then two hit 4.77 and then it falls off.
I had 3600x before this and I could hit 4.4ghz on some cores, but that was usually just for a blink of eye and no more. I am getting similar jumps on this but it can go up to 5ghz in similar situations, so thats pretty nice improvement.

Btw I am using B550 tomahawk with PBO enabled with motherboard limits and +50mhz max boost. I am currently using Noctua NH-U12S. Tomorow I will get NH-D15S, so I hope it will get only better.
 

·
Registered
Joined
·
150 Posts
OK seems memory did have an effect, or some other config I changed. I now ran 14 hours of blended prim 95 on all core CO -30 with PBO enabled and I got no errors.
Only had to reduce core 7 to -15( or 20 not sure anymore) and 9 to 25. but everything else is on auto, so still plenty of stuff that could cause issues, least I'm not looking an rma.
 

·
Registered
Joined
·
27 Posts
OK seems memory did have an effect, or some other config I changed. I now ran 14 hours of blended prim 95 on all core CO -30 with PBO enabled and I got no errors.
Only had to reduce core 7 to -15( or 20 not sure anymore) and 9 to 25. but everything else is on auto, so still plenty of stuff that could cause issues, least I'm not looking an rma.
I dont see why you would RMA the chip when it can do -30 ?! :) You should not do that, my CCD1 can run at -30 as well on all core tests where all cores run at slower speed. But when I test the cores one by one with just one thread some of them are unstable even at -15 or -20, because they boost to max clock they can. You should use the core cycler and not all core tests, because your clock speed will be very much limited by temperature of the chip with all core prime95. If you test just one core at time it will boost higher because you dont hit all pbo limits and then it might show which core is actually unstable. I am running the tests for 3 days already and still lowering limits for some cores.
 

·
Registered
Joined
·
150 Posts
That was in relation to a previous comment. My memory was stable but whatever I changed trying to overlock CPU changed that and I was getting error even on pbo disabled. After spending weeks reducing CO. Now starting over with default settings, I was able to at least run prime 95 (for now 30+ hours) without error. As you say, that's not the whole story most likely, but at least I have something that looks stable for now at much higher settings.
 

·
Premium Member
Joined
·
6,366 Posts
I fine-tuned my Curve Optimiser settings.

This passes Core Cycler 4-1344 FFTs. I'm thinking my CPU is a decent sample. :)

2512784




2512785


2512786


2512787
 
  • Rep+
Reactions: kratosatlante

·
Registered
Joined
·
715 Posts

·
Premium Member
Joined
·
6,366 Posts
a nice bonus in the form of seemingly the best performant Core 0 ?
A perfect CPU will do all 30s but your best cores need lower offsets.
 

·
Registered
Joined
·
3,393 Posts

·
Registered
Joined
·
62 Posts
I see a lot of users using Curve optmizer to undervolt the cores and get more boosts, but why are you adding +volts offset to vcore? Does this mean the threads which due to the undervolt from curve optimizer get their "voltage back" from the offset set on the vcore meaning they can boost higher?

I tried a positive offset on my vcore and the only result was higher temps, no better single thread performance, in fact... i tested a negative 0.50mv offset on the vcore and the result was lower temps and the same single thread + multi thread performance in cinebench r20.

So can anyone please explain to me why you are doing positive offsets for vcore, what is the supposed benefit of doing this? Also with a positive vcore voltage offset my max is over 1.5v , is this safe long term for a 5950x?

EDIT: And does it mean that if you give vcore a positive offset that your potential to go lower on the curve optimizer values for cores that may not have been stable before at default vcore with a negative value? Example my core 2 won't go any kower than -5, with a +.0.250mv vcore offset could i potentially reach -10 or more with this core in the curve optimizer, is that the idea?
 

·
Registered
Joined
·
554 Posts
Its been late to asked for a post before 4 months but I want to know if you were having WHEA errors back then with agesa 1.1.0.0 and which version windows you were running.With latest windows IF has problems with OC and I have many errors with this agesa.Im running x370 and AMD block the beta bios do it's impossible to get any update on the future.
Sorry dude I just Signed in Very late response, ever bios has been different for the most part or they take 4 steps back, not really sure anyone AIB knows what is going on I haven't really tried pushing my FLCK anymore and pretty much gave up OCing the system since... But no i wasn't per that post to answer you
 

·
Registered
Joined
·
142 Posts
I decided to revisit older BIOS after comparing some benchmark results with the latest AGESA 1.2.0.2 against older UEFIs. I flashed the A82 for my board that uses AGESA 1.1.0.0, re-tuned Curve Optimizer and as a result I am seeing higher single core boosts and benchmark numbers as well. All core load performance not much different but there's a clear difference in low core count boost frequency and performance.

E.g. CPU-Z single core over 700 points and frequency >5GHz is only reached with older AGESA. With 1.2.0.2 I can barely get 680 points and reach 4900MHz... Another thing I see with older AGESA is that with simple loads (boost tester) setting PBO frequency override actually increases the Effective Frequency while with the newer AGESA the CPU never really gets closer to the frequency + override (e.g. 5050 boost + override 100MHz -> 5150MHz). AGESA 1.2.0.2 caps the max frequency for my CPU at 5082MHz no matter what boost override I set if above 50MHz...

PS. I just finished testing CoreCycler and the CPU is actually stable with higher boosts. Anybody knows how to get boost back with newer AGESAs?

Anybody seeing this behavior?
 

·
Registered
Joined
·
1,529 Posts
Single core performance is controlled by the "fastest" core(s). Negative CO offset can boost that, but the fastest core on my 5900X runs into errors at -10 already (maybe even earlier if I tested more thoroughly).

All-core performance is controlled by the "slowest" core(s). Negative CO offset to bring these closer to the faster cores enabled all cores to run faster. On my 5900X the slowest cores at -30 still cannot match the fastest ones, though.

Low core performance is a special case, even when only 2 cores are active. The slowest active core dictates the global frequency limit, aka pull the fastest active core(s) down to its clock-rate. Running the 2 fastest cores concurrently (2+3 on my 5900X) results in higher clock-rate on both active cores than running the fastest + slowest cores concurrently (2+10 on my 5900X).

CPPC both helps and hinders to get the fastest possible cores working concurrently. It helps, because Windows schedules threads to run on the fastest cores first. It hinders, because Windows schedules background threads to run on the slowest core, which then may become too active sometimes and thus slow down faster cores again.
 

·
Registered
Joined
·
609 Posts
With the 5950x back in stock here and even a small sale on just now, just joined the club. About 4 months after getting my watercooling setup sorted and we're going to have to do a drain and refill :eek:

Upgrading from a 3900xt, excited to see AMD improving single core and what seems to be pretty decent overclocking and tweaking on the 59xx range.

I'll need to do my homework on what the curve optimizer is compared to how things worked on the 3900xt. I have just watched this video and it seems like the CTR software might be quite good with these new chips

Never really felt like it was needed on last gen, other than some stability testing for manual OC.
 

·
Registered
Joined
·
135 Posts
Hello people!

I'm trying my luck here, because I found none anywhere else. I'm trying to figure out, how, in the name of ZEUS, I am, somehow, 2000 to 3000 points away from some people's CPU Score, on Timespy, with my 5950X.

Let me elaborate a bit: I'm currently using the exact same settings (or so, I've checked as much as I could), as several other users, and currently am sitting on a 16456 pts CPU score. Meanwhile, those users are sitting on 18500-19000 pts CPU scores. After a lot of looking at scores, a lot of people seems to be in my situation, where, despite everything being near-identical, their CPU score is much much lower.

So, here are the things I did to improve the score, and the general settings around it:
  • Disabled SMT (boosted the score from 15000 to 16000 points), given that Timespy uses very few threads.
  • All core OC of 4.9GHz on the first die, 4.8GHz on the second (don't worry, those are benching runs only, and, there doesn't seem to be any clock stretching)
  • Memory at 3866MHz (Flck at 1933MHz) 14-14-14-14-26-42 (4 sticks of Samsung B die)
  • Running on an Asus X570 Dark Hero, with an Evga 1600 T2 power supply (so, power delivery ain't really a problem)
  • Windows power management on "performance mode"
  • This is cooled by an Optimus AMD Foundation cpu block with two 60mm thick 360mm radiators... So the CPU did not even surpass 65 degrees in any of the several runs I did.
Now, I am a bit at a loss for a reason for the low CPU score. I mean, with nearly identical setups getting over 2000 points more, with the same settings, I must be missing something, somewhere, but what?

Thanks for any help.
 

·
Robotic Chemist
Joined
·
4,110 Posts
That is very odd. I get 17733 Timespy CPU score with only a PBO+curve optimizer overclock (SMT disabled) running 1800 fclk, and it is pretty hot in here (water just over 28°C). I have the same motherboard and I even have a -0.05V offset on the CPU voltage.

Maybe compare scores with an 1800 fclk?

I assume you have this power management stuff disabled too?

Advanced\AMD CBS\NBIO Common Options\SMU Common Options
APBDIS: 1
DF Cstates: Disabled
Fixed SOC Pstate: P0
CPPC: Enabled
CPPC Preferred Cores: Enabled

Edit: I don't think it has anything to do with APBDIS, setting it back to auto didn't reduce my scores at all. Very confusing!
 
341 - 360 of 603 Posts
Top