1usmus does it again! This time WST 1.0K - check this out

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,931
Would be nice if it was actually available now, Luckily September is not far away.
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,711
does per CCX overclocking disable power save features like idle clockdown?
 

mda

2[H]4U
Joined
Mar 23, 2011
Messages
2,111
Wonder if there's going to be a way to plug these values in the BIOS when the numbers have been locked down.
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
22,230
Wonder if there's going to be a way to plug these values in the BIOS when the numbers have been locked down.
as long as your bios has the options available you should be able to. the apps more of a dummy proof way to overclock for people that don't want to mess around in the bios.
 

kirbyrj

Fully [H]
Joined
Feb 1, 2005
Messages
29,091
as long as your bios has the options available you should be able to. the apps more of a dummy proof way to overclock for people that don't want to mess around in the bios.

Not to mention the stress testing of the settings built into the app (not sure if that was from this video or somewhere else I was reading about it).

The way I see it, there's really no reason for anyone NOT to use this app. It's a personalized binning of your chip for more performance AND lower power consumption.
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
22,230
Not to mention the stress testing of the settings built into the app (not sure if that was from this video or somewhere else I was reading about it).

The way I see it, there's really no reason for anyone NOT to use this app. It's a personalized binning of your chip for more performance AND lower power consumption.

yeah you just have to download r20(not the microsoft store version), not sure what it uses or does with the higher AVX workload setting, if it uses something else though. it'll auto stress test it using that and graphs everything out for you as well. the app looks pretty insane for being free, i'll definitely try it with my 3600 just to see if it's able to get better numbers than i was able to get when i did have it overclocked.

but there is one caveat to it though, just because you can hit X numbers overclocking in windows, doesn't guarantee that those exact settings will work trying to come out of a cold boot so that's something to take into consideration if you want to transfer those settings into your bios.
 

kirbyrj

Fully [H]
Joined
Feb 1, 2005
Messages
29,091
yeah you just have to download r20(not the microsoft store version), not sure what it uses or does with the higher AVX workload setting, if it uses something else though. it'll auto stress test it using that and graphs everything out for you as well. the app looks pretty insane for being free, i'll definitely try it with my 3600 just to see if it's able to get better numbers than i was able to get when i did have it overclocked.

I read it had Prime95 built in. Not sure where I saw that. It showed up in my Google news feed on my phone. I'm pretty sure it wasn't LTT.
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
22,230
I read it had Prime95 built in. Not sure where I saw that. It showed up in my Google news feed on my phone. I'm pretty sure it wasn't LTT.

ok then that would explain the higher AVX workload, wasn't sure if it was p95 or aida64.
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,711
I would guess most will only see about 2-3% performance increase on CPUs with 2-3 CCXs. Still better than nothing, but for me I will keep PBO enabled instead so my idle power usage doesn't skyrocket.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
I read it had Prime95 built in. Not sure where I saw that. It showed up in my Google news feed on my phone. I'm pretty sure it wasn't LTT.

The intro video didn't state that, just that for intents and purposes it was similar to it. That said, I don't see how it's possible to do any of this if it's under a real P95 load. P95 loads the chip so hard there's nothing to squeeze out of it. Yea, I'm a lil skeptical atm.
 

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
309
I would guess most will only see about 2-3% performance increase on CPUs with 2-3 CCXs. Still better than nothing, but for me I will keep PBO enabled instead so my idle power usage doesn't skyrocket.

Just setting my LLC to 3 as the video suggests raised my idle temps over 15C compared to LLC=8 where I have it on my MSI board. I set the EDC bug trick in PBO and my idle temps are around 36-38C and when Windows loaded with just the LLC=3 my fans kicked up in speed, fan curve, as the idle temps hit in the 53-56 range. My all core bench in R20 results also dropped and my ST score dropped a point or 2. So I am not sure how much the software would help since I already score in the high 7500 to low 7600 range on R20 MT the way I have it setup now.
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,711
Just setting my LLC to 3 as the video suggests raised my idle temps over 15C compared to LLC=8 where I have it on my MSI board. I set the EDC bug trick in PBO and my idle temps are around 36-38C and when Windows loaded with just the LLC=3 my fans kicked up in speed, fan curve, as the idle temps hit in the 53-56 range. My all core bench in R20 results also dropped and my ST score dropped a point or 2. So I am not sure how much the software would help since I already score in the high 7500 to low 7600 range on R20 MT the way I have it setup now.
with a 3900x or better, the tool might help save time vs doing this process manually. However, with my 3800x, I spent a little time trying out various voltages and CCX frequencies and testing cinebench. Didn't take all that long with just two CCXs. That being said, I aim for 100% stability, and once I found what I thought to be stable settings, I failed AIDA stress test after 8 hours. Plenty stable for gaming I'm sure, but I demand 24hr AIDA stability in case I do something intensive like a video render project I can't afford to have fail halfway through.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
Just setting my LLC to 3 as the video suggests raised my idle temps over 15C compared to LLC=8 where I have it on my MSI board. I set the EDC bug trick in PBO and my idle temps are around 36-38C and when Windows loaded with just the LLC=3 my fans kicked up in speed, fan curve, as the idle temps hit in the 53-56 range. My all core bench in R20 results also dropped and my ST score dropped a point or 2. So I am not sure how much the software would help since I already score in the high 7500 to low 7600 range on R20 MT the way I have it setup now.

My skepticism aside, but to be fair changing your llc on a whim w/o using the rest of the settings for the app doesn't equal anything. All that reflects imo is your cooling. Raising or lowering llc or running 1.3v manual doesn't change my idle temps.
 

kirbyrj

Fully [H]
Joined
Feb 1, 2005
Messages
29,091
The intro video didn't state that, just that for intents and purposes it was similar to it. That said, I don't see how it's possible to do any of this if it's under a real P95 load. P95 loads the chip so hard there's nothing to squeeze out of it. Yea, I'm a lil skeptical atm.

I don't know if it was just stressing individual cores, CCXs, or the whole package.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
I don't know if it was just stressing individual cores, CCXs, or the whole package.

Who knows at this point. I just don't think there's much to this yet, just like the power plan. Dram calc I'm half in half out. Dram calc is helpful but if one buys into w/o skeptiscism, it can lead ya into a rabbit hole. Any ways did you know that Cinebench uses AVX? It does but it doesn't use a whole lot of it because as we all know Cinebench is not very hard as a load and a poor indicator of stability. When the app says avxlite, that's what I bet its referring to lol.
 

tool_462

Limp Gawd
Joined
Oct 14, 2008
Messages
150
I read it had Prime95 built in. Not sure where I saw that. It showed up in my Google news feed on my phone. I'm pretty sure it wasn't LTT.

I think I saw that in Guru3D article.

Now I'm here hoping it'll play nice on Threadripper 3970X like the RAM calculator did.
 

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
309
My skepticism aside, but to be fair changing your llc on a whim w/o using the rest of the settings for the app doesn't equal anything. All that reflects imo is your cooling. Raising or lowering llc or running 1.3v manual doesn't change my idle temps.

Currently I have a 280mm AIO so the CPU it was not the cooling. Everything in the BIOS power wise is set to Auto still. So just switching to a different LLC, I imagine, must have boosted my voltage on the CPU which would explain the higher temp spikes. In the past I have never noticed my CPU jump into the high 50C range and close to 60C just loading Windows, which kicked my fans up in speed which is how I noticed this. The chip did not settle down either and performed worse in the quick R20 comparison bench I ran to compare it to the previous setting. The chip also did not boost MT as much as previously. While I assume the software would have helped since it is under volting as well I am not sure I will even test it. My R20 scores in the 7600 range for MT in R20 are already higher by 400+ pts than what they showed in the LTT video with the settings I am using. I was just bored on a day off and decided I wanted to re-boot a few times to test the LLC they said to set for this new software to work and see what, if anything, it may do with my current settings. I really only tried LLC=8, 7, and 6 in my initial tweaking and found 8 to be the best for MT performance while also boosting ST performance from what I had in stock form, with PBO enabled, and with PBO enabled but motherboard PPT, EDC, and TDC settings instead of AMD standard levels. All three would get me from 6900-7100 pts on MT in R20 so I use the EDC bug tweak to get the most out of my chip.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
Currently I have a 280mm AIO so the CPU it was not the cooling. Everything in the BIOS power wise is set to Auto still. So just switching to a different LLC, I imagine, must have boosted my voltage on the CPU which would explain the higher temp spikes. In the past I have never noticed my CPU jump into the high 50C range and close to 60C just loading Windows, which kicked my fans up in speed which is how I noticed this. The chip did not settle down either and performed worse in the quick R20 comparison bench I ran to compare it to the previous setting. The chip also did not boost MT as much as previously. While I assume the software would have helped since it is under volting as well I am not sure I will even test it. My R20 scores in the 7600 range for MT in R20 are already higher by 400+ pts than what they showed in the LTT video with the settings I am using. I was just bored on a day off and decided I wanted to re-boot a few times to test the LLC they said to set for this new software to work and see what, if anything, it may do with my current settings. I really only tried LLC=8, 7, and 6 in my initial tweaking and found 8 to be the best for MT performance while also boosting ST performance from what I had in stock form, with PBO enabled, and with PBO enabled but motherboard PPT, EDC, and TDC settings instead of AMD standard levels. All three would get me from 6900-7100 pts on MT in R20 so I use the EDC bug tweak to get the most out of my chip.

Dude, there's a disconnect here. You're suggesting that it was small change but in reality when you are running your chip to the point of a 7600 range score in R20, it's not a small thing. 7600 in R20 means you're clocked pretty far into the red zone so obviously giving it more voltage would have a more apparent effect. And using EDC bug is not a minor tweak lol. Also, it seems you may be missing the point of setting llc to a static range. It's not strictly to create performance but to remove variation in vdroop so that other settings work with consistency.
 
Last edited:
  • Like
Reactions: x509
like this
Joined
Feb 6, 2013
Messages
737
Overclocking dummy here. If I understood the video they were going for a 4ghz all core overclock while undervolting and memory tuning on a 3950x. I was under the impression that all core overclocking made the ryzen processors perform worse and it was best to leave all on auto. What am I not understanding?
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
Overclocking dummy here. If I understood the video they were going for a 4ghz all core overclock while undervolting and memory tuning on a 3950x. I was under the impression that all core overclocking made the ryzen processors perform worse and it was best to leave all on auto. What am I not understanding?

Worse... not exactly instead makes it not as well rounded. A locked all core overclock will give gains in a multi threaded workload while obviously giving up a large amount in single threaded workloads. That is if the overclocked all core frequency is higher than the base all core frequency. It all depends on what you do and what you expect. That said in the video a 4ghz all core overclock goal is like spare change since a stock chip will run a 3.9ghz all core frequency all on its own anyways. Thus in this case running a 4ghz all core won't be much faster but you'd lose a crap ton on low/single threaded workloads.
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,711
Worse... not exactly instead makes it not as well rounded. A locked all core overclock will give gains in a multi threaded workload while obviously giving up a large amount in single threaded workloads. That is if the overclocked all core frequency is higher than the base all core frequency. It all depends on what you do and what you expect. That said in the video a 4ghz all core overclock goal is like spare change since a stock chip will run a 3.9ghz all core frequency all on its own anyways. Thus in this case running a 4ghz all core won't be much faster but you'd lose a crap ton on low/single threaded workloads.
pretty much nailed it, but to put in perspective, if you have a 3700x which will boost to 4.4GHz single core, and you overclock it to 4GHz all core, you are essentially losing 10% performance in single core tasks (while gaining a little more than 10% in multithreaded tasks). Comes down to choosing where you want your CPU to perform the best.
 
Joined
Feb 6, 2013
Messages
737
So if I understand correctly, this application is geared towards gains in multi-threaded workloads like cinebench but will lower performance in a gaming scenario(me.) Can it just be used to undervolt while keeping clocks the same?
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,711
So if I understand correctly, this application is geared towards gains in multi-threaded workloads like cinebench but will lower performance in a gaming scenario(me.) Can it just be used to undervolt while keeping clocks the same?
The program is to improve multithreaded performance, yes, but many games are multithreaded at this point. Unless you are playing something really old that only uses one or two threads.

However, if you want to keep your cpu clocks stock but want to undervolt, thst is pretty easy to do in bios using a voltage offset.
 

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
309
Dude, there's a disconnect here. You're suggesting that it was small change but in reality when you are running your chip to the point of a 7600 range score in R20, it's not a small thing. 7600 in R20 means you're clocked pretty far into the red zone so obviously giving it more voltage would have a more apparent effect. And using EDC bug is not a minor tweak lol. Also, it seems you may be missing the point of setting llc to a static range. It's not strictly to create performance but to remove variation in vdroop so that other settings work with consistency.

In my mind it is a minor tweak. I set the EDC to a lower level, in my case 16, and the TDC and PPT to the motherboard's max. Thus allowing the motherboard to power the CPU with whatever it wants to draw within its spec. I left all voltages on Auto and really just then changed the LLC. Not like what this software seems to do or what I would need to manually do changing voltages and other settings to get the best static over clock. This software seems to basically keeping running R20 and then undervolting the CPU while chandgins the all core clock frequencies until it is stable. While I may play with it for shits and giggles, just like when I first played with an all core OC, I will probably go back to what I have since it gives the best all around MT and ST performance for my CPU. And I really only played with the LLC values to play with them and then noticed that at 8 on my MSI board it gave the best ST performance. Actually I could get it to hit 525 ST with actually changing some other stuff but it would not stay stable in P95 and would random reboot so 511-515 is what it gets now.

By far not major tweaking like people do with Ryzen where they are clocking CCD's at different speeds to get the most out of it. Just a few settings and let the rest do it on auto pilot.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,617
In my mind it is a minor tweak. I set the EDC to a lower level, in my case 16, and the TDC and PPT to the motherboard's max. Thus allowing the motherboard to power the CPU with whatever it wants to draw within its spec. I left all voltages on Auto and really just then changed the LLC. Not like what this software seems to do or what I would need to manually do changing voltages and other settings to get the best static over clock. This software seems to basically keeping running R20 and then undervolting the CPU while chandgins the all core clock frequencies until it is stable. While I may play with it for shits and giggles, just like when I first played with an all core OC, I will probably go back to what I have since it gives the best all around MT and ST performance for my CPU. And I really only played with the LLC values to play with them and then noticed that at 8 on my MSI board it gave the best ST performance. Actually I could get it to hit 525 ST with actually changing some other stuff but it would not stay stable in P95 and would random reboot so 511-515 is what it gets now.

By far not major tweaking like people do with Ryzen where they are clocking CCD's at different speeds to get the most out of it. Just a few settings and let the rest do it on auto pilot.

You can clock your chip any which way you like, that's all good. But you are disconnected from the costs associated with how you're clocking your chip. The settings you are using are seriously hard on the chip, it's not just some minor tweak just cuz you think it is. You're forcing the chip to hold boost for much longer and harder to achieve that range of scoring. For ex. a 4.3ghz all core overclock achieves a 7600 range score. And a 4.3ghz overclock is getting rather serious, it's just a tick away from the max of very well binned chips of 4.4ghz or 4.5ghz for a golden chip. The fact that you're scoring in the same range, it is NOT from minor tweaking. You'repushing the chip into the red zone with PBO. Using a ratio overclock is not anymore serious, it's just a different way of doing it.
 

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
309
You can clock your chip any which way you like, that's all good. But you are disconnected from the costs associated with how you're clocking your chip. The settings you are using are seriously hard on the chip, it's not just some minor tweak just cuz you think it is. You're forcing the chip to hold boost for much longer and harder to achieve that range of scoring. For ex. a 4.3ghz all core overclock achieves a 7600 range score. And a 4.3ghz overclock is getting rather serious, it's just a tick away from the max of very well binned chips of 4.4ghz or 4.5ghz for a golden chip. The fact that you're scoring in the same range, it is NOT from minor tweaking. You'repushing the chip into the red zone with PBO. Using a ratio overclock is not anymore serious, it's just a different way of doing it.

But mine only sits in that supposed 'Red Zone' during a moment when tested, and that is only when R20 is set to 'highest' priority in Task Manager. Otherwise it hits in the 7400-7500 range and maxes at 4.275Ghz all core according to HWInfo64, when R20 is just run at 'below normal' priority. It is not sitting at that speed all day and all night. Sure it is doing better than stock when a test is run. Either way I do not plan to keep my chip for 'years' so I really do not care what a few spikes here and there do to it.

Also the only reason it hits those speeds is due to me cranking the pump and fan speeds on the AIO and case fans. If I leave them at my normal settings it does not boost as high, about 50-75Mhz lower to be honest and of course drops the score. Showing how much the Ryzen chips like to boost when kept cool. I should have my stuff for my loop next week to build that and see what happens. I doubt it will boost much more if any.
 
Last edited:

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
309
I have really enjoyed just using precision boost. I have found overall for me the best performance.

I wish mine would have worked better with just using PBO. My chip could not break 7000 MT in R20 or 498 ST scores. No matter what settings I used. It only hit 7100 and 501 once I changed the settings to motherboard power settings instead of the AMD ones for PPT, EDC and TDC. I saw reviewers and others posting they were getting at least 7200 and 511 in R20 and other getting upwards of 7300 with just PBO enabled.
 

bobzdar

[H]ard|Gawd
Joined
Jun 6, 2003
Messages
1,888
Fine if you don't game, but no all core o/c will match what the 3950X can do on it's own when only 4-6 cores are loaded..I guess it's worth running this and setting a profile in Ryzen master for when I do photogrammetry or video work and load all the cores, but then I'd have to reboot to go back to stock boost for VR gaming. It'd be nice if you could tune the boost curve for each core instead of just locking the clock or messing with the PBO limits - which honestly seems broken on zen 2.
 
Top