Overclock.net banner

241 - 260 of 292 Posts

·
Registered
Joined
·
199 Posts
Actually, I think you're right. With an offset getting unexplained lag in games and Windows. :(
Glad you figured it out. They made CO to replace offset because of the clock stretching everyone was getting on Zen2.

These are tricky chips, I ending up going to a per-ccx manual overclock for my uses.
 

·
Robotic Chemist
Joined
·
3,592 Posts
Actually, I think you're right. With an offset getting unexplained lag in games and Windows. :(
Clock stretching isn't going to cause unexplained lag in games or Windows, is it?

That doesn't make sense to me since clock stretching is a very short time period thing, something that happens on timescales completely below detectable by humans. I don't notice any downside due to my negative offset until I set it a bit higher, when it starts to harm performance in benchmarks.

Glad you figured it out. They made CO to replace offset because of the clock stretching everyone was getting on Zen2.
Is there some objective test I could do to detect issues with clock stretching? I assumed performance in benchmarks or similar would be able to detect clock stretching and that is the only context I see it discussed in. If it doesn't hurt benchmarks what else do I need to check?
 

·
Premium Member
Joined
·
6,013 Posts
Clock stretching isn't going to cause unexplained lag in games or Windows, is it?

That doesn't make sense to me since clock stretching is a very short time period thing, something that happens on timescales completely below detectable by humans. I don't notice any downside due to my negative offset until I set it a bit higher, when it starts to harm performance in benchmarks.



Is there some objective test I could do to detect issues with clock stretching? I assumed performance in benchmarks or similar would be able to detect clock stretching and that is the only context I see it discussed in. If it doesn't hurt benchmarks what else do I need to check?
Might have been I had my offset too low, but with the CPU voltage on Auto, no more Windows and game lag. I mean it was quite bad. :(
 

·
Robotic Chemist
Joined
·
3,592 Posts
Might have been I had my offset too low, but with the CPU voltage on Auto, no more Windows and game lag. I mean it was quite bad. :(
Very interesting, I wonder what it is from. If it was very obvious at least I know mine (probably) isn't doing it. :unsure:

I wish it just crashed, overclocking is so hard when the CPU does this weird auto-downclocking (sleeping?) with voltage being too low.
 

·
Premium Member
Joined
·
6,013 Posts
Very interesting, I wonder what it is from. If it was very obvious at least I know mine (probably) isn't doing it. :unsure:

I wish it just crashed, overclocking is so hard when the CPU does this weird auto-downclocking (sleeping?) with voltage being too low.
I lowered the offset to .0625 and no more Windows or game lag. Multicore R20 is still nearly 100 points faster.
 

·
Robotic Chemist
Joined
·
3,592 Posts
I lowered the offset to .0625 and no more Windows or game lag. Multicore R20 is still nearly 100 points faster.
That is exactly the same offset I found. At first I thought it was -0.050V because -0.075 caused slowdown, but after testing finer steps I think -0.0625V is the optimal (for my CPU?).

Leaving it on Auto seems a little intense, 1.5V seems high even though it is stock. I like how the offset lowers the power/heat for multicore too. :)
 

·
Premium Member
Joined
·
6,013 Posts
That is exactly the same offset I found. At first I thought it was -0.050V because -0.075 caused slowdown, but after testing finer steps I think -0.0625V is the optimal (for my CPU?).

Leaving it on Auto seems a little intense, 1.5V seems high even though it is stock. I like how the offset lowers the power/heat for multicore too. :)
With a .0625 offset OCCT is 15C cooler AVX2 Extreme preset and I gain close to 100 points in R20 multicore. :)

2488192
 

·
Registered
Joined
·
1,477 Posts
Trying negative Vcore offset is one of the first things I am looking at with my 5900X. I am less worried about temperatures, but about power usage. At idle my 5900X system draws 10-15 watts more from the wall than my 9900K system, using the same components and a lesser number of fans.

In WoW Shadowlands the 5900X systems draws 50-60 watts more power than the 9900K system displaying the same scenes at same 60 fps and same graphics settings while using the same GPU. About 45-50 watts is from increased CPU package power compared to the 9900K.

First impression of the 5900K is a bit of a let-down, because the increased power usage does not correspond to increased performance yet.

- I was mostly hoping for around 20% decrease in load time for single-threaded Lua scripts (WoW and Fantasy Grounds), but the load times are practically the same.

- My highest multi-threaded load is using Topaz Gigapixel AI to up-sample images to enormous size (short of the 32k px limit). But due to this kind of image software not being able to make full use of parallelization I only get around 28% more performance out of the 50% more cores, higher IPC and much larger L3 cache (9900K vs. 5900X).

- On the plus side, flipping and searching through large/multiple image heavy PDF files seems to be about 20% faster now. Opening very large ZIP/Deflate compressed TIF files (single-threaded) is faster, too.

We will see if mild overclocking can improve on that first impression, but I expected a bit more out of the box.
 

·
Registered
Joined
·
199 Posts
Trying negative Vcore offset is one of the first things I am looking at with my 5900X. I am less worried about temperatures, but about power usage. At idle my 5900X system draws 10-15 watts more from the wall than my 9900K system, using the same components and a lesser number of fans.
Do not use offset, use curve optimizer if you're going that route. Curve optimizer is pass/fail, offset will cause the chip to clock-stretch.
 

·
Registered
Joined
·
1,477 Posts
I will start with the offset to check if it makes any difference in wattage. The most shocking figure is the extra 50-60 watts (185w vs 245w) compared to the 9900K system, running the very same load. And that's with the 9900K only being allowed C3 via BIOS and the 5900X being allowed even deeper C-states.
 

·
Registered
Joined
·
155 Posts
feels like a milestone but I finally hit 700 single core, Never rebooted a pc so much in my life... Think I gained 5 more gray hairs in my beard and a pissed off wife.. View attachment 2474480
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.
 

·
Registered
Joined
·
1,477 Posts
Using an offset of -0.0625 V decreased my WoW test-load by about 10 watts tops, still 50 watts more than the 9900K system for doing the same thing. Idle power isn't really affected, but that was expected.
 

·
Registered
Joined
·
199 Posts
Using an offset of -0.0625 V decreased my WoW test-load by about 10 watts tops, still 50 watts more than the 9900K system for doing the same thing. Idle power isn't really affected, but that was expected.
I'm still curious why you're using offset and not curve optimizer? Curve optimizer does the same thing, but on a per core basis, which gives better results.
 

·
Registered
Joined
·
123 Posts
Each point in the Curve Optimizer corresponds to ~3-5mV according to some slides shared by AMD. A full -30 CO setting should result in between -0.150V and -0.090V voltage change.
 

·
Robotic Chemist
Joined
·
3,592 Posts
Each point in the Curve Optimizer corresponds to ~3-5mV according to some slides shared by AMD. A full -30 CO setting should result in between -0.150V and -0.090V voltage change.
I don't see any change in the max voltage when adjusting the curve optimizer. The only way to lower the peak voltage is to use an offset.

What am I missing?
 

·
Registered
Joined
·
199 Posts
I don't see any change in the max voltage when adjusting the curve optimizer. The only way to lower the peak voltage is to use an offset.

What am I missing?
Are you getting the same peak clock speed as well? If so, something in the BIOS is not setup right.
 

·
Registered
Joined
·
1,477 Posts
I am using offset, because it is a fast way to test things (wattage). In my MSI UEFI there is another option that (by name) allows to only lower the peak voltage instead of applying an overall offset. This could be useful. But at this point I was interested in power draw.

Overall I might be most interested to increase single-threaded Lua interpreter performance, as this is the biggest disappointment yet while being the one thing that keeps me waiting for over half a minute whenever I start Wow or Fantasy Grounds (VTT).
 

·
Registered
Joined
·
202 Posts
I don't see any change in the max voltage when adjusting the curve optimizer. The only way to lower the peak voltage is to use an offset.
CO doesnt reduce max voltage because it is using primarily temps and power as limiter but it reduces voltage for a given clock speed and should raise avg clock.
 

·
Registered
Joined
·
4,009 Posts
Alright so here's my overclocking experience.

CPU: 5950X
Cooler: 280mm Alphacool XT45 & EK Magnitude Waterblock
Motherboard: Crosshair VIII Impact
Memory: 32GB Dominator Platinum 3600Mhz CL18

I feel like I've got a decent good chip that performs consistently well across the board. I've fiddled around with PBO, as well as the curve optimizer but ended up settling on a simple PBO configuration. With only PBO enabled and a Scalar of 5X, my chip will hit 4.5 to 4.55 Ghz across all cores and hold them in R20/R23. This is with auto on just about all of the other adjustments including voltage. On single core tests, it will hit and stay at 5.0Ghz, Two Cores at least 4.9Ghz and Four cores at 4.75Ghz. This performance seems excessively easy to achieve with just enabling PBO and a scalar adjustment. Temps are usually stable between 82 to 85 degrees C on all core tests, about 65C or less on single core test.

However, further tweaking does not achieve any appreciable increase in performance. When I set the offset to 200Mhz, the system will recognize that the OC scheme goes from PBO to "Auto Overclock" according to Ryzen Master and I'll sometimes lose performance. Small adjutments may result in a overall lower all core boost of 4.3Ghz or 4.4Ghz A negative offset of 20 on all cores seems to match the PBO performance giving me about 4.52Ghz on all cores and again about 5.02 on a single core. At a negative 25 on all cores will result in clear instability across the system. I have not tested the negative 20 offset in all stability scenarios and various single to multi core loads so I'm not positive on stability. A negative 15 should be stable but again, the performance is really close to just a standard PBO. I have not invested the time in per core offsets in the curve optimizer, I'm sure if I do, I could probably eek out a little bit more performance but that leads to my conclusion.

With how well PBO performs out of the box and it's almost guaranteed stability at the temps I'm seeing, I find very little incentive to do further adjustments, curve optimizer, and stability testing to try to match or gain that little bit over what PBO brings to the table. I have even less desire to try my hand at the PBO/Manual Overclock hybrid overclock.

Thoughts?

2488747
2488748
2488750


EDIT: I think the CB bench could have been higher if I were to tighten the timing of my memory up from CL18.
 

·
Registered
Joined
·
127 Posts
Might seem like an obvious question, but what do you guys use to stress the CPU temperature wise?
I tried blender (Classroom/BMW), prime 95 and AIDA and none of those go above 55 degrees, but I know it can get hotter since those all only push up less than 4600.
During 3dmark it went into the 70ties, but there isn't there something better to test that, something that loops and reliably heats it up? Heaven benchmark also left it around the 55 ballpark.
 
241 - 260 of 292 Posts
Top