[SOLVED] R7 3800x vs R5 2600 temps

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
You can probably live with the 84-86 degree spikes, but if it were my CPU I wouldn't want to see anything higher than 80°C consistently. Since you probably won't see those same temps most of the time you might be ok with it, but if you run ANYTHING that makes frequent use of AVX or AVX2 instructions, and there ARE plenty of games and applications that do, then you will want to keep a close eye on temps and you may end up wanting to drop your manual OC down to 4.2Ghz if you wish to see your piece of silicon last a nice long time.

While it's true that you are unlikely to cause any immediate damage at temps between 80-90°C, consistent temperatures in that range will almost certainly contribute to premature degradation through electromigration and VT shift over the course of months or years.
While gaming the avg temp is 65c. Sometimes dropping into the higher 50s, with the all core oc.

But my whole problem is that, this is only 5c cooler than the stock cooler, with the all core OC.

I just thought it was going to cool like, much better than it is. idk. As I keep finding more info, like this video at the 3 minute mark you can see while hes using Aida64, on a 4.4ghz OC while using a custom water loop, he hits 76c, so that makes my 84c with 4.3ghz on air cooling seem really good.
View: https://www.youtube.com/watch?v=-6VlrlS0f4Q&t=26s


But idk what to believe because I keep seeing different results everywhere and my personal experience with this cooler vs the stock cooler doesnt seem to be that major.
 
Last edited:

M3rKn

Proper
Nov 13, 2019
139
42
120
4
Huh? He has a Scythe Mugen 5 rev.B cooler. Where did the comment about the Prism cooler come from?
Well he also has a Cooler Master T2 that he was using with his 2600. The Prism came with his 3800x. Its earlier in the thread. So he has 3 coolers at the moment.
 

Darkbreeze

Titan
Moderator
Yes, but he isn't trying to USE any of those other coolers either. And I'm sure he's unlikely to be sending this one back.

As far as the temperatures go, IDK what to tell you man. I know that is a good cooler BUT you are also using an 8/16 CPU with a STOCK 105w TDP that does not factor in boosting, it is a (imaginative) calculation of the highest NON-boosted P-state, per AMD internal documents, and contrary to what many (Most) here would try to tell you. It is the same (Although grossly different, considering both AMD and Intel tend to use imaginary and creative ways to determine exactly WHAT constitutes a watt for the purpose of determining TDP, and neither of them do this even loosely the same.) for Intel. The TDP you see is the highest non-boosted P-state thermal condition.

So anyhow, regardless of all that, it is impossible for me to offer any logical reason why you are not seeing much better thermals using the Macho rev.B than you did with the stock cooler. Every system I've EVER installed a good aftermarket cooler like that on has shown significant improvements in thermal compliance, and the three Ryzen systems have as well. Granted, those were not systems using the 3800x, two were the 3600x and one was a 3600 but that shouldn't matter at all. COMPARATIVELY there should have been a correlation between a 3600x using the Wraith spire and a good 120mm aftermarket cooler and a 3800x differential between the Wraith prism and a good 120mm cooler.

If however you have a fairly high ambient in the room, there may not BE much room for improvement because ALL coolers are only capable of dissipating a certain amount of rise over ambient. The higher the ambient, the less effective the cooler is going to seem to be. If for example you have a room that has a 26.6°C (80°F) ambient temperature you are probably going to see a significantly higher "temperature" reading on your full load readings than you would if you tested the same system in a room with an 18°C (65°F) ambient temperature. What somebody else gets, even if they are using the exact same motherboard, CPU and CPU cooler, and even if you eliminate the case cooling by testing in an open air bench, is not going to be the same if you have different ambient temperatures AND even further compounding the problem is the fact that every single CPU sample is different as well. Some, simply are not as good as others.

I'm assuming you have made no alterations to the heatsink in any way. You'd be surprised. I've seen people try to shave just a teeny bit off the tops of the heat pipes to make it fit in a case better, which completely ruins the cooler. I'm sure that's not something you have done, but I have to mention it because you won't be the only person that reads this thread in the entirety of the threads lifespan.

Also, you said you said that "paste is perfect amount".

Ok, so HOW exactly did you determine that? WHAT is your idea of "a perfect amount"? Pasting methods for Ryzen CPUs are somewhat different, optimally, than for some other architectures, especially on the 3000 series. Obviously, any accepted method will "work" but what "works" isn't necessarily optimal AND there is still a place in my mind that believes your mount may not have the full amount of pressure that it ought to have or that some other aspect of it could be to blame.

In some cases, especially on motherboards with less thickness, there can be issues with some coolers. It may be necessary to create additional distance from the motherboard surface to the backplate by installing fiber or hard plastic washers between the backplate and the motherboard, so that on the other side the mounting brackets that attach to hardware actually attached to the CPU cooler are a bit "lower" or closer to the surface of the motherboard so that when the fasteners are snugged down or bottomed out, there is slightly more mounting pressure. This may not be the case, in your case, but it is certainly worth investigating.

I am unconvinced that this cooler is not capable of keeping your CPU below 80 degrees in the stock configuration and I believe that if it is there is either something wrong with the cooler, something wrong with the mount or something wrong with the CPU. Either that or Scythe has seriously dropped the performance of it's parts and this will be the last time I recommend one because the VERY TINY Thermaltake True spirit direct 140 I put on those 3600x systems dropped the tested temperature by about 11 degrees, and the surface area of that Mugen 5 rev.b due to it's much thicker heatsink has to be at least somewhat comparable to that, even given the fact that your CPU has two extra cores and four extra threads by comparison. The Prism cooler you used originally also has somewhat better performance than the Spire that came with the 3600x.

Beyond that, it's probably worth mentioning that people spend far more than you spent on this cooler to achieve a 5°C reduction in temperatures, all the time.

And 84-86c with the 4.3ghz all core OC.

The Prism is 92-95c with the all core OC.
You do realize that a maximum temperature with the Mugen 5 of 86°C compared to a maximum temperature with the Wraith prism of 95°C is a 9°C difference, not a five degree difference, right? That is about what I told you to expect in the beginning I believe, with about an average of a ten degree difference. Somewhere in the neighbor hood of 8-12°C difference depending on ambient. I'm not sure where the problem is if that is actually what you are seeing. Rarely does a CPU cooler offer more compared to the Wraith coolers unless we're talking one of the big twin finstack 140mm coolers or liquid cooling. Especially if you have a moderately warm ambient room temperature. The temperature IN your case, is irrelevant. The ambient temp of the ROOM itself, is what matters, because that is where the air is going to be coming from to cool inside and you can not only not cool below ambient without some kind of refrigeration process but the higher the ambient is, the less effective the cooler is going to be as well.
 
Reactions: Jason H.

M3rKn

Proper
Nov 13, 2019
139
42
120
4
My windows is up to date, there is no AMD Ryzen Balanced plan. Also yes, while gaming my temps are about 55-65c at stock, PBO disabled, but with the Sycthe Mugen 5, not stock cooler. (Im downloading the chipset from AMDs website now for the Power Plan)

but also, thats not really my main concern. My main concern is if these are normal temps, with this cooler, on the 3800x. The Sycthe Mugen 5 rev.b
Yes, but he isn't trying to USE any of those other coolers either. And I'm sure he's unlikely to be sending this one back.

As far as the temperatures go, IDK what to tell you man. I know that is a good cooler BUT you are also using an 8/16 CPU with a STOCK 105w TDP that does not factor in boosting, it is a (imaginative) calculation of the highest NON-boosted P-state, per AMD internal documents, and contrary to what many (Most) here would try to tell you. It is the same (Although grossly different, considering both AMD and Intel tend to use imaginary and creative ways to determine exactly WHAT constitutes a watt for the purpose of determining TDP, and neither of them do this even loosely the same.) for Intel. The TDP you see is the highest non-boosted P-state thermal condition.

So anyhow, regardless of all that, it is impossible for me to offer any logical reason why you are not seeing much better thermals using the Macho rev.B than you did with the stock cooler. Every system I've EVER installed a good aftermarket cooler like that on has shown significant improvements in thermal compliance, and the three Ryzen systems have as well. Granted, those were not systems using the 3800x, two were the 3600x and one was a 3600 but that shouldn't matter at all. COMPARATIVELY there should have been a correlation between a 3600x using the Wraith spire and a good 120mm aftermarket cooler and a 3800x differential between the Wraith prism and a good 120mm cooler.

If however you have a fairly high ambient in the room, there may not BE much room for improvement because ALL coolers are only capable of dissipating a certain amount of rise over ambient. The higher the ambient, the less effective the cooler is going to seem to be. If for example you have a room that has a 26.6°C (80°F) ambient temperature you are probably going to see a significantly higher "temperature" reading on your full load readings than you would if you tested the same system in a room with an 18°C (65°F) ambient temperature. What somebody else gets, even if they are using the exact same motherboard, CPU and CPU cooler, and even if you eliminate the case cooling by testing in an open air bench, is not going to be the same if you have different ambient temperatures AND even further compounding the problem is the fact that every single CPU sample is different as well. Some, simply are not as good as others.

I'm assuming you have made no alterations to the heatsink in any way. You'd be surprised. I've seen people try to shave just a teeny bit off the tops of the heat pipes to make it fit in a case better, which completely ruins the cooler. I'm sure that's not something you have done, but I have to mention it because you won't be the only person that reads this thread in the entirety of the threads lifespan.

Also, you said you said that "paste is perfect amount".

Ok, so HOW exactly did you determine that? WHAT is your idea of "a perfect amount"? Pasting methods for Ryzen CPUs are somewhat different, optimally, than for some other architectures, especially on the 3000 series. Obviously, any accepted method will "work" but what "works" isn't necessarily optimal AND there is still a place in my mind that believes your mount may not have the full amount of pressure that it ought to have or that some other aspect of it could be to blame.

In some cases, especially on motherboards with less thickness, there can be issues with some coolers. It may be necessary to create additional distance from the motherboard surface to the backplate by installing fiber or hard plastic washers between the backplate and the motherboard, so that on the other side the mounting brackets that attach to hardware actually attached to the CPU cooler are a bit "lower" or closer to the surface of the motherboard so that when the fasteners are snugged down or bottomed out, there is slightly more mounting pressure. This may not be the case, in your case, but it is certainly worth investigating.

I am unconvinced that this cooler is not capable of keeping your CPU below 80 degrees in the stock configuration and I believe that if it is there is either something wrong with the cooler, something wrong with the mount or something wrong with the CPU. Either that or Scythe has seriously dropped the performance of it's parts and this will be the last time I recommend one because the VERY TINY Thermaltake True spirit direct 140 I put on those 3600x systems dropped the tested temperature by about 11 degrees, and the surface area of that Mugen 5 rev.b due to it's much thicker heatsink has to be at least somewhat comparable to that, even given the fact that your CPU has two extra cores and four extra threads by comparison. The Prism cooler you used originally also has somewhat better performance than the Spire that came with the 3600x.

Beyond that, it's probably worth mentioning that people spend far more than you spent on this cooler to achieve a 5°C reduction in temperatures, all the time.



You do realize that a maximum temperature with the Mugen 5 of 86°C compared to a maximum temperature with the Wraith prism of 95°C is a 9°C difference, not a five degree difference, right? That is about what I told you to expect in the beginning I believe, with about an average of a ten degree difference. Somewhere in the neighbor hood of 8-12°C difference depending on ambient. I'm not sure where the problem is if that is actually what you are seeing. Rarely does a CPU cooler offer more compared to the Wraith coolers unless we're talking one of the big twin finstack 140mm coolers or liquid cooling. Especially if you have a moderately warm ambient room temperature. The temperature IN your case, is irrelevant. The ambient temp of the ROOM itself, is what matters, because that is where the air is going to be coming from to cool inside and you can not only not cool below ambient without some kind of refrigeration process but the higher the ambient is, the less effective the cooler is going to be as well.
what I was talking about applies to all coolers.
 

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
Yes, but he isn't trying to USE any of those other coolers either. And I'm sure he's unlikely to be sending this one back.

As far as the temperatures go, IDK what to tell you man. I know that is a good cooler BUT you are also using an 8/16 CPU with a STOCK 105w TDP that does not factor in boosting, it is a (imaginative) calculation of the highest NON-boosted P-state, per AMD internal documents, and contrary to what many (Most) here would try to tell you. It is the same (Although grossly different, considering both AMD and Intel tend to use imaginary and creative ways to determine exactly WHAT constitutes a watt for the purpose of determining TDP, and neither of them do this even loosely the same.) for Intel. The TDP you see is the highest non-boosted P-state thermal condition.

So anyhow, regardless of all that, it is impossible for me to offer any logical reason why you are not seeing much better thermals using the Macho rev.B than you did with the stock cooler. Every system I've EVER installed a good aftermarket cooler like that on has shown significant improvements in thermal compliance, and the three Ryzen systems have as well. Granted, those were not systems using the 3800x, two were the 3600x and one was a 3600 but that shouldn't matter at all. COMPARATIVELY there should have been a correlation between a 3600x using the Wraith spire and a good 120mm aftermarket cooler and a 3800x differential between the Wraith prism and a good 120mm cooler.

If however you have a fairly high ambient in the room, there may not BE much room for improvement because ALL coolers are only capable of dissipating a certain amount of rise over ambient. The higher the ambient, the less effective the cooler is going to seem to be. If for example you have a room that has a 26.6°C (80°F) ambient temperature you are probably going to see a significantly higher "temperature" reading on your full load readings than you would if you tested the same system in a room with an 18°C (65°F) ambient temperature. What somebody else gets, even if they are using the exact same motherboard, CPU and CPU cooler, and even if you eliminate the case cooling by testing in an open air bench, is not going to be the same if you have different ambient temperatures AND even further compounding the problem is the fact that every single CPU sample is different as well. Some, simply are not as good as others.

I'm assuming you have made no alterations to the heatsink in any way. You'd be surprised. I've seen people try to shave just a teeny bit off the tops of the heat pipes to make it fit in a case better, which completely ruins the cooler. I'm sure that's not something you have done, but I have to mention it because you won't be the only person that reads this thread in the entirety of the threads lifespan.

Also, you said you said that "paste is perfect amount".

Ok, so HOW exactly did you determine that? WHAT is your idea of "a perfect amount"? Pasting methods for Ryzen CPUs are somewhat different, optimally, than for some other architectures, especially on the 3000 series. Obviously, any accepted method will "work" but what "works" isn't necessarily optimal AND there is still a place in my mind that believes your mount may not have the full amount of pressure that it ought to have or that some other aspect of it could be to blame.

In some cases, especially on motherboards with less thickness, there can be issues with some coolers. It may be necessary to create additional distance from the motherboard surface to the backplate by installing fiber or hard plastic washers between the backplate and the motherboard, so that on the other side the mounting brackets that attach to hardware actually attached to the CPU cooler are a bit "lower" or closer to the surface of the motherboard so that when the fasteners are snugged down or bottomed out, there is slightly more mounting pressure. This may not be the case, in your case, but it is certainly worth investigating.

I am unconvinced that this cooler is not capable of keeping your CPU below 80 degrees in the stock configuration and I believe that if it is there is either something wrong with the cooler, something wrong with the mount or something wrong with the CPU. Either that or Scythe has seriously dropped the performance of it's parts and this will be the last time I recommend one because the VERY TINY Thermaltake True spirit direct 140 I put on those 3600x systems dropped the tested temperature by about 11 degrees, and the surface area of that Mugen 5 rev.b due to it's much thicker heatsink has to be at least somewhat comparable to that, even given the fact that your CPU has two extra cores and four extra threads by comparison. The Prism cooler you used originally also has somewhat better performance than the Spire that came with the 3600x.

Beyond that, it's probably worth mentioning that people spend far more than you spent on this cooler to achieve a 5°C reduction in temperatures, all the time.



You do realize that a maximum temperature with the Mugen 5 of 86°C compared to a maximum temperature with the Wraith prism of 95°C is a 9°C difference, not a five degree difference, right? That is about what I told you to expect in the beginning I believe, with about an average of a ten degree difference. Somewhere in the neighbor hood of 8-12°C difference depending on ambient. I'm not sure where the problem is if that is actually what you are seeing. Rarely does a CPU cooler offer more compared to the Wraith coolers unless we're talking one of the big twin finstack 140mm coolers or liquid cooling. Especially if you have a moderately warm ambient room temperature. The temperature IN your case, is irrelevant. The ambient temp of the ROOM itself, is what matters, because that is where the air is going to be coming from to cool inside and you can not only not cool below ambient without some kind of refrigeration process but the higher the ambient is, the less effective the cooler is going to be as well.
Yea with all the previous info I seemed to have gotten some numbers mixed up. Im sorry about that.

For the last test, I recorded a video with the 4.3ghz 1.325v OC doing 6 passes (5 minutes) of Cinebench. Here are the results. This is on par for the cooler? Im not sure why its only 79c now and not 82c... Maybe because I was testing so much before?

Im using Cinebench because someone told me Prime95 or Aida takes months, if not years off lifespan everytime its used... They said to do Cinebench infinite loop, but for immediate results just set it for 5 minutes.

FULL SCREEN RECOMMENDED FOR TEXT CLARITY

View: https://www.youtube.com/watch?v=5YVqvtWhQO4
 

Darkbreeze

Titan
Moderator
Im using Cinebench because someone told me Prime95 or Aida takes months, if not years off lifespan everytime its used...
Completely untrue AND inaccurate. Complete BS, and yes, I do wish I was allowed to spell that out fully and capitalize it, and I'd say the same thing directly to the face of ANYBODY who says otherwise. Furthermore, I would invite them, wholeheartedly, to show the PROOF of that, which of course they and others who've made similar claims in the past, certainly cannot do. There are applications using AVX and AVX2, not to mention AVX-512, that potentially create greater than 100% TDP loads while Prime95 Small FFT will ONLY create 100% TDP loads, steady state, on all cores, so long as the AVX instructions are disabled.

Even with AVX enabled, and regardless of any other considerations, if Prime95 were capable of creating thermal conditions (And it might be, depending on the level of your overclock and what kind of cooling you have) that could potentially be thermally damaging, it would require having a failure of the thermal protections that are a part of the architecture in order for permanent damage to occur, in which case you would have bigger problems than what thermal testing utility you were running anyhow.

Let me explain something, in exactly the way it was explained to me, many years ago. And this comes from somebody who has hundreds and hundreds, if not THOUSANDS, of man hours invested in testing CPU architectures and has in fact LITERALLY written the book on it, as seen here:



This was the clarification he gave me, those many years ago, and it is important that you actually read through ALL of it, and absorb what is said, so that you will understand that Prime95 Small FFT with no AVX instructions enabled is specifically compliant with the Intel datasheets regarding thermal compliance testing. If there was ANY possibility that running it was detrimental to the CPU, you can be 3000% sure that would NOT be the case. As well, the evidence is entirely clear that Cinebench is NOT a steady state testing utility, and is NOT a suitable metric for testing thermal compliance. It is a benchmark. Nothing more, nothing less. It is not suitable for thermal testing and it is not suitable for stress testing. It is suitable for benchmarking performance. Period.

As quoted from Computronix:


I can think of several reasons why x264 encoding or AVX / AVX2 / FMA3 apps won't work as a unilateral metric for thermal testing.

(1) A steady-state workload gives steady-state temperatures; encoding does not.

(2) Simplicity in methodology; most users would find encoding apps unfamiliar and cumbersome to accomplish a simple task.

(3) Most users such as gamers never run any apps which use AVX / FMA, so adaptive or manual voltage aside, it makes no sense to downgrade your overclock to accommodate those loads and temps.

(4) Standardization; Prime95 has been around since 1996; many users are familiar with it.

For the minority of users who routinely run AVX / FMA apps, then P95 with AVX/AVX2 can be useful for tweaking BIOS for thermal and stability testing.
Regardless of architecture, Prime95 Small FFT (With AVX/AVX2 disabled) works equally well across all platforms. Steady-state is the key. How can anyone extrapolate accurate Core temperatures from workloads that fluctuate like a bad day on the Stock Market?

I'm aware of 5 utilities with steady-state workloads. In order of load level they are:

(1) P95 (No AVX/AVX2) - Small FFT's
(2) HeavyLoad - Stress CPU
(3) FurMark - CPU Burner
(4) Intel Processor Diagnostic Tool - CPU Load
(5) AIDA64 - Tools - System Stability Test - Stress CPU

AIDA64's Stress CPU fails to load any overclocked / overvolted CPU to get anywhere TDP, and is therefore useless, except for giving naive users a sense of false security because their temps are so low.

HeavyLoad is the closest alternative. Temps and watts are within 3% of Small FFT's.
Prior to the initial stability testing for an OC, I run a full system backup. I then use an SSD dedicated for testing (also backed up), because crashing Windows and BSOD's are an integral part of the OC process. This is especially critical when OC'ing memory, because when crashed, it's the absolute fastest way to corrupt Windows and / or any software.

Like thermal testing, CPU stability testing is an extremely controversial topic. However, unlike thermal testing, for which I've established a methodology that "complies" with the datasheets, I think the initial objective for stability testing is to approach the process by adhering to tests that run workloads around 100% TDP, +/- about 10%. I generally follow the %TDP scale in my guide in section 11. I'm sure you've previously done so, but you might want to take another look at it just to refresh your memory.

I created the %TDP scale to give users a better perspective of test utility workloads, because most don't realize what a HUGE difference there is between %TDP "Workload", and %CPU "Utilization", which is merely an indication of processor resource activity. It would be a big improvement if Microsoft would upgrade Windows Task Manager to show CPU Power Consumption (Watts) as well as %TDP Workload, because %CPU Utilization is very misleading since it tops out at 100%, and can't read excessively high workloads such as P95 with AVX at nearly 130%.

One of the first utilities I run is Intel Processor Diagnostic Tool, so I know that I'm working with a good CPU. Then for the initial OC quick stability test I use P95 v26.6 Small FFT's. After BIOS tweaking gets it past 20 minutes, I run 25 passes of IBT on "Standard" and "High", then RealBench for 15 minutes and AIDA64 for 20 minutes with CPU + FPU selected. (AIDA64 can still be used when expired by disabling your internet connection, then changing the computer's date to within 30 days of when AIDA64 was installed).

I'm not particularly fond of AIDA64 because there are 4 possible test selections (CPU, FPU, Cache, RAM) involving the CPU which can potentially yield15 different test results. That's a lot of variables ... and almost no one bothers to mention which test(s) they ran (like not mentioning which P95 version and torture test, or ambient temperature). Note that the individual CPU test is less than 70% TDP while the individual FPU test is about 115% TDP workload, but only the CPU + FPU test combination is about 100% TDP workload.

I typically don't run some of the more popular utilities until later in the process, such as IETU, CPU-Z, AIDA64 test CPU, and Cinebench R15 (auto-looped for 20 runs using a batch file) because I know their workloads are too light. However, the new Cinebench R20 is interesting because they increased the workload closer to 100% TDP, and included a new feature for setting a time interval for looping, such as 1200 seconds (20 minutes) or whatever you like.

I also don't test a gaming OC with any P95 AVX versions because I know that without an AVX offset, it'll crash the rig. That's just always a given.

After I've fine-tweaked BIOS and I'm satisfied that my initial OC is stable, I go back to P95 for a minimum of 2 hours for Small FFT's, and 2 hours for Blend, then later for 8 to 12 hours. I pretty much work my way up and down the utilities shown on the %TDP scale in my guide, eventually finishing up with everything I can throw at it from my test utility inventory, including all the memory (P95 non-AVX Blend, Memtest overnight) and graphics (3DMark and Unigine) tests, as well as running games and apps.

It takes me several days and nights to a week or so of testing to validate any rig's stability and thermal performance to my complete satisfaction. However, in a practical sense, there's seldom the luxury of having enough time to assemble, test and deliver an OC'd rig with a comfortably high level of confidence ... so only real-world usage time will tell. I actually prefer working with known-good used parts rather than unknown new parts, because as we all know, in the world of computer electronics, there's about a 10% failure rate within the first few months, not to mention DOA's.
It appears that mersenne.org's server no longer hosts P95 v26.6. The link works, but when you click on 64 bit to download the .zip file, it displays "This site can’t be reached" followed by "The webpage at ftp://mersenne.org/gimps/p64v266.zip might be temporarily down or it may have moved permanently to a new web address. ERR_FTP_FAILED".

It's been like that for several days now:

• Prime95 v26.6 - http://www.mersenneforum.org/showthread.php?t=15504

Since P95's newest version, 29.8, was recently released, it's likely that the absence of 26.6 was deliberate. 29.8's new features include "click" options to disable AVX, AVX2 & AVX-512, and Small FFT's applies the same workload in terms of power and heat as 26.6, I see these new features as long-needed improvements to P95's user interface. They've also changed their test options:

*Smallest FFT's
Small FFT's
*Medium FFT's
Large FFT's
Blend
Custom

Additionally, the default FFT lengths and values are slightly different, which has no appreciable thermal effects.

So with these updates, there's no further need to differentiate between versions; just run Small FFT's on the latest version with all AVX options disabled. Simple.

Accordingly, I've updated the text and illustrations in sections 10, 11 & 12 of the guide to reflect 29.8. Most of the changes are in Section 11, so take a look at it. If you'd like, I can share my updated Prime95 29.8.jpg file with you for your Overclocking Guide.

I may also post a new thread as a "P95 Notice" in the CPUs and Mods forums.
I assure you, if there was ANY chance of causing damage by running Prime95 for the purpose of thermal testing, or in the even you happen to be testing memory stability for which the custom Blend mode test is a staple in the memory overclocking and tweaking community, that Computronix and myself, not to mention thousands of other technicians, engineers, hardware reviewers and general enthusiasts, would definitely not be using it. I have periodically run Prime95 on all my machines, to check for continued thermal compliance, at six month intervals, AFTER establishing initial thermal compliance after a build, for all my builds, for YEARS, and have never had any indication of any kind of problem, nor will I ever.

Even Puget Systems, arguably one of the most respected systems builders and PC think tanks that exist, uses Prime95 in it's testing arsenal.


As do too many others to list. There are some detractors out there but they largely offer snake eats tail, politician speak, unproven theoretical reasoning to suggest against the use of Prime95 with either no real understanding of how the utility can and should be used, or a misunderstanding of it's behaviors and properties, and with no "proof" of it causing any of the supposed damage in any case.

All of that being said, yes, if you run Prime95 Small FFT with AVX and AVX2 ENABLED, and no offset in the BIOS, you are very likely to exceed safe thermal parameters and create potential for both crashing or forced system shut downs due to exceeding thermal spec. So, don't do that. If you KNOW you are going to be running applications that make heavy use of AVX instructions, of any kind, then you will want to specifically tune the AVX offset in the BIOS along with the use of AVX instructions in Prime95, so that you are able to safely use AVX with no problems to your hardware.

For stock configurations this should actually not be a problem to start with, because there should already be enough headroom for some amount of AVX usage, but even so, if you know there will be heavy usage, tuning for it is advisable.
 
Reactions: Jason H.

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
Completely untrue AND inaccurate. Complete BS, and yes, I do wish I was allowed to spell that out fully and capitalize it, and I'd say the same thing directly to the face of ANYBODY who says otherwise. Furthermore, I would invite them, wholeheartedly, to show the PROOF of that, which of course they and others who've made similar claims in the past, certainly cannot do. There are applications using AVX and AVX2, not to mention AVX-512, that potentially create greater than 100% TDP loads while Prime95 Small FFT will ONLY create 100% TDP loads, steady state, on all cores, so long as the AVX instructions are disabled.

Even with AVX enabled, and regardless of any other considerations, if Prime95 were capable of creating thermal conditions (And it might be, depending on the level of your overclock and what kind of cooling you have) that could potentially be thermally damaging, it would require having a failure of the thermal protections that are a part of the architecture in order for permanent damage to occur, in which case you would have bigger problems than what thermal testing utility you were running anyhow.

Let me explain something, in exactly the way it was explained to me, many years ago. And this comes from somebody who has hundreds and hundreds, if not THOUSANDS, of man hours invested in testing CPU architectures and has in fact LITERALLY written the book on it, as seen here:



This was the clarification he gave me, those many years ago, and it is important that you actually read through ALL of it, and absorb what is said, so that you will understand that Prime95 Small FFT with no AVX instructions enabled is specifically compliant with the Intel datasheets regarding thermal compliance testing. If there was ANY possibility that running it was detrimental to the CPU, you can be 3000% sure that would NOT be the case. As well, the evidence is entirely clear that Cinebench is NOT a steady state testing utility, and is NOT a suitable metric for testing thermal compliance. It is a benchmark. Nothing more, nothing less. It is not suitable for thermal testing and it is not suitable for stress testing. It is suitable for benchmarking performance. Period.

As quoted from Computronix:










I assure you, if there was ANY chance of causing damage by running Prime95 for the purpose of thermal testing, or in the even you happen to be testing memory stability for which the custom Blend mode test is a staple in the memory overclocking and tweaking community, that Computronix and myself, not to mention thousands of other technicians, engineers, hardware reviewers and general enthusiasts, would definitely not be using it. I have periodically run Prime95 on all my machines, to check for continued thermal compliance, at six month intervals, AFTER establishing initial thermal compliance after a build, for all my builds, for YEARS, and have never had any indication of any kind of problem, nor will I ever.

Even Puget Systems, arguably one of the most respected systems builders and PC think tanks that exist, uses Prime95 in it's testing arsenal.


As do too many others to list. There are some detractors out there but they largely offer snake eats tail, politician speak, unproven theoretical reasoning to suggest against the use of Prime95 with either no real understanding of how the utility can and should be used, or a misunderstanding of it's behaviors and properties, and with no "proof" of it causing any of the supposed damage in any case.

All of that being said, yes, if you run Prime95 Small FFT with AVX and AVX2 ENABLED, and no offset in the BIOS, you are very likely to exceed safe thermal parameters and create potential for both crashing or forced system shut downs due to exceeding thermal spec. So, don't do that. If you KNOW you are going to be running applications that make heavy use of AVX instructions, of any kind, then you will want to specifically tune the AVX offset in the BIOS along with the use of AVX instructions in Prime95, so that you are able to safely use AVX with no problems to your hardware.

For stock configurations this should actually not be a problem to start with, because there should already be enough headroom for some amount of AVX usage, but even so, if you know there will be heavy usage, tuning for it is advisable.
Ty for all that info! I will be sure to study it!

Also, something very weird happened today. Someone suggested on my video to turn on PBO as their Cinebench score on a 3700x with no OC was higher than my 3800x with the manual oc. So I turned on PBO with the manual OC and Im seeing a max temp of 75c with a score about 200cb higher than the one in my video....

This is very strange to me.
 

Darkbreeze

Titan
Moderator
Not strange at all. A lot of Cinebench benchmarks are likely to rely on single core performance, and with PBO enabled it's likely that for short periods of time the system might allow a higher single core clock speed than what you have with a manual overclock.
 

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
Not strange at all. A lot of Cinebench benchmarks are likely to rely on single core performance, and with PBO enabled it's likely that for short periods of time the system might allow a higher single core clock speed than what you have with a manual overclock.
PBO is enabled WHILE the manual OC is enabled. meaning my clock is set to 4.3 and it never goes above or below that under load.

I was under the assumption PBO only works without a manual OC, and if thats the case, idk how my temp is 75c today, and a 200 higher cb score.
 

Darkbreeze

Titan
Moderator
IDK man. You got me there. I'll look into it but that seems as though it flies in the face of known behavior, not that I doubt you, I don't, but I'm not sure how performance could increase and thermals could decrease, by enabling PBO on top of an already manual overclock.
 
Reactions: Jason H.

Rogue Leader

It's a trap!
Moderator
IDK man. You got me there. I'll look into it but that seems as though it flies in the face of known behavior, not that I doubt you, I don't, but I'm not sure how performance could increase and thermals could decrease, by enabling PBO on top of an already manual overclock.
PBO allows for higher VRM current and SoC Power so in theory even though he manually overclocked (which is locking his clock speed) anything he didn't manually set is being allowed to run higher, while the clock rate is limited to the 4.3 he manually set it at. A good well cooled 3800X under PBO with no other settings should be able to hit 4.4ghz.

PBO probably shouldn't work with a manual OC, in fact you should shut it off, in the same way when you manually OC an Intel processor or an older AMD processor you shut off Turbo. But I bet there is nothing to prevent you from using it, just like there is nothing preventing you from leaving Turbo on. The difference is PBO is way more sophisticated so it doesn't just crank the clock from wherever its at automatically. Most likely if PBO is causing any sort of improvement, there are probably changes you can make to your OC that would be better, OR some external factor like your ambient temperature went down significantly.

Your Cinebench multi score is a little better than most magazines got. However 200 points is kinda splitting hairs.
 
Reactions: Jason H.

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
man, who knew just changing my cooler would be such a learning experience lol. Seriously thanks for all the info.

I mean I see a lot of people on stock settings hitting 5000cb but mine would NOT go past 4850. IDK I mean Ill probably turn PBO off but, would you suggest maybe trying 4.4ghz :p
 

Rogue Leader

It's a trap!
Moderator
man, who knew just changing my cooler would be such a learning experience lol. Seriously thanks for all the info.

I mean I see a lot of people on stock settings hitting 5000cb but mine would NOT go past 4850. IDK I mean Ill probably turn PBO off but, would you suggest maybe trying 4.4ghz :p
I'd suggest saving a backup of your overclock settings if they are stable and then turning it all back to Auto and leaving PBO on and see if it performs better. The reality is PBO does better than manual overclocking most of the time. And its dynamic so it will scale with your cooling and temps.

I bet under PBO you'll hit 4.4ghz stable, and get the same kind of scores or better.
 
Reactions: Jason H.

Darkbreeze

Titan
Moderator
Right. Which is why for the most part I've recommended, depending on the type of setup and how much a person wants to push it, that users either simply use PBO or disable it and use the stock precision boost with no overdrive. With good cooling, PBO should be fine. With stock cooling, in my experience with the Wraith coolers, just using precision boost, no overdrive, is enough to push the stock cooler to it's limits pretty much. Manual overclocking has repeatedly been shown to really not offer that much in most cases on Ryzen unless you are either a very experienced overclocker, or willing to do a LOT of extensive research and testing to become one, or both, if you're an experienced overclocker with no previous Ryzen overclocking experience.

Considering most of the reviews show little to be gained from doing so, I'm pretty much happy to simply go with whatever is the simpler process for any client builds. If and when I get my own Ryzen platform I'll probably play around with it but I don't think that's going to happen until the next Gen is released. Seems like there is less and less room for clocks the more the process is shrunk and that is to be expected I guess.
 

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
I'd suggest saving a backup of your overclock settings if they are stable and then turning it all back to Auto and leaving PBO on and see if it performs better. The reality is PBO does better than manual overclocking most of the time. And its dynamic so it will scale with your cooling and temps.

I bet under PBO you'll hit 4.4ghz stable, and get the same kind of scores or better.
I'd suggest saving a backup of your overclock settings if they are stable and then turning it all back to Auto and leaving PBO on and see if it performs better. The reality is PBO does better than manual overclocking most of the time. And its dynamic so it will scale with your cooling and temps.

I bet under PBO you'll hit 4.4ghz stable, and get the same kind of scores or better.
With PBO ENABLED, no OC (settings on Auto) it only boosts to 4.15 in cinebench, and my score is almost 300 points lower. I hit 5111 yesterday with my OC and today, with how you told me to set it up, is 4815 and is hotter by 2c (77c) because it uses a higher voltage to achieve the 4.15 than I use to achieve 4.3 steadily accross the board.

Just tested it just now.

SO how is it not boosting higher than 4.15? Whats wrong...

Also I dont see 2 different options for just Precision Boost, Or Precision Boost Overdrive.

In my bios there is an AMD Overclocking Option, I hit that and there it says Precision Boost Overdrive, Enabled.

Then In my normal BIOS Overclocking section, it says Advanced CPU Configuration, there is says Precision Boost Overdrive, but it has different options to choose from. I believe some are higher states of PBO? I just chose, Enabled.

Then there is an option, in the same section called Core Boost Performance. And all that has is Auto or Disabled. Its set to Auto.
 
Last edited:

Darkbreeze

Titan
Moderator

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
Precision boost overdrive disabled means no PBO and it would ONLY have Precision boost enabled then. PBO enabled, should be, enabled. Shouldn't be any more to it than that.

Core boost is likely the way you can disable the Precision boost, if you wanted to for some reason. No reason really to do that though.

Did you happen to read this, I can't remember.

https://www.gamersnexus.net/guides/3491-explaining-precision-boost-overdrive-benchmarks-auto-oc
Well, PBO enabled is only hitting 4.15, with a higher voltage, than my manual OC. Why is that.

Everyone keeps telling me just leave PBO enabled but its legitly worse than my manual OC, unless something is wrong. Which I havent changed any other settings besides my fan curves, and set xmp on memory. Theres more than 1 setting under my PBO options though. I think its like "Enhanced 1" "Enhanced 2" "Enhanced 3" then it has Eco modes to run the cpu at a lower tdp.

So unless I need to do something else, it is seeming whatever PBO is doing, its not doing it properly, and my manual OC is better. At least during cinebench.
 

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
PBO is likely assuming you don't have the thermal overhead or VRM power to do it. I would try the higher enhanced modes and see if that makes the difference, probably loosens things up.
Turned on the highest enhanced pbo (enhanced 4).. No differance than just enabled. Still 4.15ghz at a higher voltage than my manual OC.

Tbh Im about to just give up because none of this is making any sense at all whatsoever.

Lower temps with a HIGHER manual overclock, but PBO cant even go past 4.2ghz in cinebench at 77c?

Im honestly getting annoyed with it. People tell me PBO, the internet shows people at stock settings hitting 5000cb scores in cinebench with the STOCK COOLER.

And I cant go over 4.15ghz with a 50$ cooler that is MUCH better than the stock one? Either someone isnt giving me enough information to solve this issue, or something is very wrong.
 

Rogue Leader

It's a trap!
Moderator
Turned on the highest enhanced pbo (enhanced 4).. No differance than just enabled. Still 4.15ghz at a higher voltage than my manual OC.

Tbh Im about to just give up because none of this is making any sense at all whatsoever.

Lower temps with a HIGHER manual overclock, but PBO cant even go past 4.2ghz in cinebench at 77c?

Im honestly getting annoyed with it. People tell me PBO, the internet shows people at stock settings hitting 5000cb scores in cinebench with the STOCK COOLER.

And I cant go over 4.15ghz with a 50$ cooler that is MUCH better than the stock one? Either someone isnt giving me enough information to solve this issue, or something is very wrong.
The problem with trusting other peoples results on the internet is you have to assume that

a) they are truthful
b) their conditions are exactly the same

I can get some amazing results with a stock cooler, in my basement, with intake air coming from the cold tile floor and the A/C blasting so the ambient room temp is around 66 deg F. Heck lets take the side panel off while we are at it.

Aside from that no two CPUs are exactly the same, some will clock better or worse, some of them their boards will clock better or worse, etc.

Like I said when you get within 200 points you're splitting hairs anyway. It sounds like you're doing better on a manual overclock, maybe your chips sensors aren't reading perfectly so PBO is less aggressive, maybe your ambient room temp is a touch higher, maybe the VRMs on your board are a little weak. The long and short of it is, you're never going to match someone elses results exactly, and you may not beat them.
 
Reactions: Jason H.

Darkbreeze

Titan
Moderator
Which I did mention previously. A poor CPU sample might result in less than favorable automatic behavior. Why doesn't it show up as a problem with your manual overclock? It does, as a higher than desirable temperature given the quality of the CPU cooler.

All it takes is that you get a CPU with a poor quality internal paste job or just plain bad sample overall and you're going to see poorer results. In this case, I guess I'd have to say that you are probably one of the few that is better off just sticking with your manual overclock so long as you can keep the temps down. The problem I see is, you can't.

As with ANY Ryzen or Intel CPU, if you can't keep the temperature below 80°C while running Prime95 Small FFT, then there IS a problem. What is the problem? Too high of voltage. Poor CPU sample. Cheap motherboard. Less than adequate cooling. Bad paste job on the cooler OR the internal TIM. Clock speed set to high for the sample and hardware combination. Take your pick.

There were a lot of people that could easily hit 5Ghz on air cooling with Haswell refresh. Then there were just as many that had samples that couldn't reliably remain stable or thermally compliant even at 4.7Ghz. This isn't a new problem. It's as old as overclocking and boost protocols.
 
Reactions: Jason H.

Jason H.

Honorable
Oct 20, 2013
1,469
75
11,590
61
I managed to get 4.4ghz at 1.28v to run stable with a max temp of 81c after like 10 runs of cinebench with a score of 5207.

That is just so nuts to me compared to with only PBO, only boosting to 4.15ghz at 1.31v, getting a max temp of 81c and only a 4807cb score.

I am so beyond confused with these new results lol.

Maybe its my motherboard. I dont really have cash for a new one and lots of people say the Tomahawk b450 is good and has good VRMs. Although I did have to update the BIOS for the new R7 chip. But even that doesnt explain the boosting issue unless the BIOS is the problem... So weird lol.
 
Reactions: Rogue Leader

Rogue Leader

It's a trap!
Moderator
Thats a good board, should be fine with overclocking and so on.

I'd try some long term testing on that OC see if it stays stable, that may be the reason. or PBO just isn't working for you which happpens. Works great on my 3700X, but as I said every chip is different
 
Reactions: Jason H.

Darkbreeze

Titan
Moderator
I have not seen ANY review site manage to get a stable overclock at that level, and they are usually using best case hand picked samples sent directly to them by the manufacturers. Cinebench is NOT an acceptable metric for "stability" testing. Which I said before I'm pretty sure. I don't much care who says it is, it's not. It's a benchmark, not a stability test.

Realbench, using the stress test option, IS a useful stability test. Run Realbench for 8 hours, with half of your memory selected. If it passes that, then I'll believe that your 4.4Ghz OC is stable. Until then, I can only believe that it is not stable, regardless of the fact that you are not getting any clearly evident bluescreens or errors. There is much more to stability than that though. Many unstable systems will never show any symptoms like that until a while down the road when microerrors accumulate and begin causing major errors.
 
Reactions: Jason H.

ASK THE COMMUNITY

TRENDING THREADS