ClockTuner for Ryzen Simplifies "Zen 2" Overclocking, Squeezes Out Double-digit Percent Performance

[btarunr]

ClockTuner for Ryzen (CTR) by Yuri "1usmus" Bubliy, is an evolution of the DRAM Calculator for Ryzen utility. The utility goes beyond the functionality of the DRAM Calculator - which finds the most precise memory settings for Ryzen processors - and does your homework for Ryzen CPU overclocking. Optimized for processors based on the "Zen 2" microarchitecture, CTR has been designed both for Socket AM4 and sTRX4 (Threadripper) processors, and Linus Tech Tips in its announcement video of CTR demonstrated the tool's prowess in squeezing out a neat 10% performance gain for their Threadripper 3960X processor. Besides CPU and memory settings, the tool performs stability testing and benchmarking. 1usmus expects to release CTR 1.0 in September 2020.



View at TechPowerUp Main Site

 

[Space Lynx]

the linustechtips video on it didn't show much benefit really, they were not able to replicate the 10% gain. at end of day Ryzen overclocks itself well enough, just leave everything default in BIOS. if you want fps gains, then you want to get good bdie ram and overclock it with dram calculator made by the same guy here 1usmus

but since all of us here do enjoy tweaking from time to time, I have to admit I will probably give this a shot at some point if it ends up supporting ryzen 4800x, seems neat anyway

 

[biffzinker]

I’m curious enough to try tuning my 3800X. It would be one way to find out if it’s a gold bin.

 

[Tomorrow]

Well i hope this turns out better than DRAM Calc as that may work fine for B-Die but offers completly unusable numbers for RAM using Hynix chips (DJR, CJR, JJR)

 

[Xzibit]

3700X score is extremely low. Most of the reviews had it at around 4900s on launch BIOSes

 

[Deleted member 185088]

Why can't Ryzen maintain it's boost speeds, I believe for Intel MOBO manufacturers can enforce unlimited boost time this offering better performance by "default".

 

[thesmokingman]

Did you see the asterik during the video around the 7min mark? "*except for single core boost." I'm going to assume single core boosting is unaffected by this tool which is not a bad thing imo.

Why can't Ryzen maintain it's boost speeds, I believe for Intel MOBO manufacturers can enforce unlimited boost time this offering better performance by "default".

That's NOT the point of this tool. Boost, if you want to waste wattage 24/7 just run an all core setting. However this is like the opposite extreme, kind of like hypermiling this is to increase perf while reducing wattage where possible.

 

[biffzinker]

Why can't Ryzen maintain it's boost speeds, I believe for Intel MOBO manufacturers can enforce unlimited boost time this offering better performance by "default".

Power draw at the higher voltage to maintain the boosted clockspeed is the issue. Electromigration would make quick work of degrading the 7nm silicon die to point the clockspeeds would be unstable to the cores going kaput. Intel is still riding it out on 14nm+++ which can handle the higher all core clockspeed without sudden degradation.

 

[zo0lykas]

 

[TheLostSwede]

Well i hope this turns out better than DRAM Calc as that may work fine for B-Die but offers completly unusable numbers for RAM using Hynix chips (DJR, CJR, JJR)

Or maybe you're the one that's clueless and doesn't know how to use it? It worked perfectly fine for me with CJR modules. I'm up 200MHz with lower latency than my modules were sold as and that's with four sticks. So please keep your opinions to yourself.

 

[LutinChris]

Power draw at the higher voltage to maintain the boosted clockspeed is the issue. Electromigration would make quick work of degrading the 7nm silicon die to point the clockspeeds would be unstable to the cores going kaput. Intel is still riding it out on 14nm+++ which can handle the higher all core clockspeed without sudden degradation.

Please enough with fanboy(ism) attitude. I'm on xeon Intel CPU (kaby lake) in my laptop (hp zbook 17 G4). Intel marketed my xeon cpu as a 3.1GHz with a boost at 4.2GHZ (one Core). Checking with Aida64, Intel Extrem tuning and antiThrottle tool from TechpowerUp, I notice the 4.2GHz frequency never last more than 1s and appears very rarely. 90% of the time, in maximum performance mode, all my CPU core works between 3.0GHz & 3.4GHz (3.6 at best). So excuse me, but the hypothetical higher clockspeed advantage achieve by Intel with it's 14nm++++++ sounds more like bullshit and marketing nuisance for me.

Intel cpu are not the best anymore. Amd catch them up and amd cpu are now more Advanced / Cooler / Powerfull / Secure. Competition is good for inovation, so let's see what's happen in next years. but For now Intel is just a follower in x86 cpu market and Amd the leader in all the aspects I've mentionned, including technologically.

 

[phill]

I’m curious enough to try tuning my 3800X. It would be one way to find out if it’s a gold bin.

No my 3900X wasn't fast enough (Continued from the other thread...) I do wish I'd gone for the 3950X but it seems my 3900X is pretty decent so it'll be fine for the moment Maybe a treat for myself at Christmas for the 3950X or 4950X.. When crunching you can never have enough cores But free performance is free performance

Can't wait to see what this tool is like Looks very promising...

 

[Tomorrow]

Or maybe you're the one that's clueless and doesn't know how to use it? It worked perfectly fine for me with CJR modules. I'm up 200MHz with lower latency than my modules were sold as and that's with four sticks. So please keep your opinions to yourself.

Wow how insightful. Maybe get some JJR based sticks (i have 4x8GB) and try to boot at 3733 CL16 when the sticks can only do CL16 at 3200 (that's the XMP profile). At 3733 they can't go below CL18. DRAM Calc for reason keeps suggesting CL16. And yes - i did import my Thaiphoon Burner profile. So it's not that.
Plus too low of a voltage, plus unreasonably low tRFC (only values that b-die can do) etc.

So don't come here saying that i don't know what im talking about. I am running at 3733 CL18 with tightened subtimings (~67ns latency with 3800X). But thanks to calclulator for useless recommendation.

 

[Chomiq]

Wow how insightful. Maybe get some JJR based sticks (i have 4x8GB) and try to boot at 3733 CL16 when the sticks can only do CL16 at 3200 (that's the XMP profile). At 3733 they can't go below CL18. DRAM Calc for reason keeps suggesting CL16. And yes - i did import my Thaiphoon Burner profile. So it's not that.
Plus too low of a voltage, plus unreasonably low tRFC (only values that b-die can do) etc.

So don't come here saying that i don't know what im talking about. I am running at 3733 CL18 with tightened subtimings (~67ns latency with 3800X). But thanks to calclulator for useless recommendation.

Sample base: 1
"Completely unusable"

 

[nangu]

Well i hope this turns out better than DRAM Calc as that may work fine for B-Die but offers completly unusable numbers for RAM using Hynix chips (DJR, CJR, JJR)

The calc offers a baseline to start tunning, it's not a "enter and forget settings" app. My 3600 c18 CJRs are running @3800 c16 in the 2x8Gb configuration, tweaked from the Calc suggested settings as a baseline, so not bad at all for a little free utility.

About this new Clock Tuner utility, It seems that its purpose is to tune a per CCX overclock with an enfasis on lower power consumption at the same, or better, all core performance than stock, which everybody can achieve by tuning in the BIOS. This app makes things easier by setting values and testing without user intervention, and without the hassle that is to roundtrip to BIOS, Windows, stress testing, benchmark, and all over again to test other values, so it's interesting to check it at least. If you want max single core boost performance, this tool is not for you. In fact, it's highly probable you end up with lower single core boost than stock.

On my 3900X I gained 7% all core performance by tweaking per CCX, at the same overall power consumption and temperature, so the gain figures presented by 1usmus are achievable I guess, depending on silicon quality of course.

 

[Tomorrow]

Sample base: 1
"Completely unusable"

That could be said about almost anything. So unless you are Silicon Lottery that you can test hundreds of samples then 99% of problems with any product on the Internet fall into the sample size: 1 category. 2 or 3 best case if the user had RMA returs or bought extra one.

 

[TheLostSwede]

Wow how insightful. Maybe get some JJR based sticks (i have 4x8GB) and try to boot at 3733 CL16 when the sticks can only do CL16 at 3200 (that's the XMP profile). At 3733 they can't go below CL18. DRAM Calc for reason keeps suggesting CL16. And yes - i did import my Thaiphoon Burner profile. So it's not that.
Plus too low of a voltage, plus unreasonably low tRFC (only values that b-die can do) etc.

So don't come here saying that i don't know what im talking about. I am running at 3733 CL18 with tightened subtimings (~67ns latency with 3800X). But thanks to calclulator for useless recommendation.

What does this have to with the software you're trashing? It doesn't even support your modules
And yes, you clearly have no idea what you're talking about.

 

[Rahnak]

the linustechtips video on it didn't show much benefit really, they were not able to replicate the 10% gain. at end of day Ryzen overclocks itself well enough, just leave everything default in BIOS. if you want fps gains, then you want to get good bdie ram and overclock it with dram calculator made by the same guy here 1usmus

but since all of us here do enjoy tweaking from time to time, I have to admit I will probably give this a shot at some point if it ends up supporting ryzen 4800x, seems neat anyway

It still had a small performance uplift. Though I did notice their 3970X was below average. And the ram situation on the bench was weird too. 3600CL14 memory running at 3200CL16.
Still, the 30 watt decrease running cinebench was well worth it, imo.

Also, keep in mind the version LTT used was older. It was still called "Workstation Tool" rather than "ClockTuner for Ryzen" and the date on the bench was 08/06/2020. Which to me it means June 8th, but it's probably August 6th. Some countries and dates..

 

[kayjay010101]

...and the date on the bench was 08/06/2020. Which to me it means June 8th, but it's probably August 6th. Some countries and dates..

Well, DD/MM/YYYY is the right way to format it

 

[ChosenName]

Well i hope this turns out better than DRAM Calc as that may work fine for B-Die but offers completly unusable numbers for RAM using Hynix chips (DJR, CJR, JJR)

Works absolutely fine (tested when CPU folding for over a month) on my RAM which uses Hynix chips. Which might suggest that the numbers are not "completly unusable".

 

[Bobweadababyitsaboy]

Please enough with fanboy(ism) attitude. I'm on xeon Intel CPU (kaby lake) in my laptop (hp zbook 17 G4). Intel marketed my xeon cpu as a 3.1GHz with a boost at 4.2GHZ (one Core). Checking with Aida64, Intel Extrem tuning and antiThrottle tool from TechpowerUp, I notice the 4.2GHz frequency never last more than 1s and appears very rarely. 90% of the time, in maximum performance mode, all my CPU core works between 3.0GHz & 3.4GHz (3.6 at best). So excuse me, but the hypothetical higher clockspeed advantage achieve by Intel with it's 14nm++++++ sounds more like bullshit and marketing nuisance for me.

Intel cpu are not the best anymore. Amd catch them up and amd cpu are now more Advanced / Cooler / Powerfull / Secure. Competition is good for inovation, so let's see what's happen in next years. but For now Intel is just a follower in x86 cpu market and Amd the leader in all the aspects I've mentionned, including technologically.

I think your problem is your laptop cooler. Put the same chip on a better cooler and I bet you could boost for longer. No fanboisim here as I want AMD to kick the shit out of intel. If you’re watching the benchmarks intel still has a slight edge when it comes to single core-performance.

 

[amit_talkin]

Well i hope this turns out better than DRAM Calc as that may work fine for B-Die but offers completly unusable numbers for RAM using Hynix chips (DJR, CJR, JJR)

I agree to this. I have DJR ( 4000 MHz 18-18-18 XMP ). I imported xmp to DRAM Calculator and it suggest me 14-18-20-20 @ 3600. However System even doesnt start with 14 CAS, so I guess this needs to be fixed.

 

[Ja.KooLit]

Cant wait to try this. I am basically settled with my settings but more fine tuning is better.

Linus did say however that as per 1usmus, this works better with a cpu with more ccx.

 

[Max(IT)]

The calc offers a baseline to start tunning, it's not a "enter and forget settings" app. My 3600 c18 CJRs are running @3800 c16 in the 2x8Gb configuration, tweaked from the Calc suggested settings as a baseline, so not bad at all for a little free utility.

About this new Clock Tuner utility, It seems that its purpose is to tune a per CCX overclock with an enfasis on lower power consumption at the same, or better, all core performance than stock, which everybody can achieve by tuning in the BIOS. This app makes things easier by setting values and testing without user intervention, and without the hassle that is to roundtrip to BIOS, Windows, stress testing, benchmark, and all over again to test other values, so it's interesting to check it at least. If you want max single core boost performance, this tool is not for you. In fact, it's highly probable you end up with lower single core boost than stock.

On my 3900X I gained 7% all core performance by tweaking per CCX, at the same overall power consumption and temperature, so the gain figures presented by 1usmus are achievable I guess, depending on silicon quality of course.

what about single core performance ? After you tuned the CPU for all core, I mean...

Well, DD/MM/YYYY is the right way to format it

except for americans ...

Cant wait to try this. I am basically settled with my settings but more fine tuning is better.

Linus did say however that as per 1usmus, this works better with a cpu with more ccx.

I think 3900X, 3950X and Threadripper's will benefit more from this tool.

 

[Deleted member 185088]

Power draw at the higher voltage to maintain the boosted clockspeed is the issue. Electromigration would make quick work of degrading the 7nm silicon die to point the clockspeeds would be unstable to the cores going kaput. Intel is still riding it out on 14nm+++ which can handle the higher all core clockspeed without sudden degradation.

Maybe, but Intel isn't doing it MOBO manufacturers are doing it,

Power draw at the higher voltage to maintain the boosted clockspeed is the issue. Electromigration would make quick work of degrading the 7nm silicon die to point the clockspeeds would be unstable to the cores going kaput. Intel is still riding it out on 14nm+++ which can handle the higher all core clockspeed without sudden degradation.

Maybe, but still running Intel's CPUs beyond their limits isn't technically within specs.
Edit: removed some nonese