Ryzen 7 1700 @3.75ghz 1.32v Cinebench R15 1627cb is that unusually good?

[deded9090]

I just overclocked my Ryzen 7 1700 and I got an extremely good score of 1627 (I think). My cpu idle with that overclock is from 37 degrees to 43 degrees but my temperature reached a max of 78 degrees. Is my voltage too high and was that dangerous for my cpu to run that hot on a single test and do so well?

 

[Hardware Brad]

That's about as high as I would go. Run something like Prime95 for about a half hour to an hour and see what the temperatures peak at. I'd try to keep it below 80c if you can.

 

[Shektron]

Looks a little above average to me. A stock clocked 1700, which has clock speed varying from 3.5-3.7 GHz, scores in the 1400 range, while one clocked at 4.0 GHz hits the 1700 range(the score is in the 1700s, that is). Yours seems closer to the 4.0 GHz range even though the clock speed is closer to the 3.5-3.7 GHz range, so yes, that is a little unusual. A maximum of 78 degrees is a little too hot, nothing dangerous of course, just not preferable.

1.32 V is fine, the safe limit for Ryzen is 1.35-1.4 V. But the temperatures are a little high. What cooler are you using?

 

[deded9090]

Just stock wrath spire cooler

 

[Shektron]

Still, that max of 78 does not look very good. As mentioned, run the Prime95 stress test and see if everything's stable. Run it for at least an hour, preferably much more than that(3-6 hours would be great). If it's stable and temps aren't over 80, I'd call it a day and use it at your current settings.

 

[deded9090]

The temp only reached 78 for three seconds and then went all the way back down to the lower 60s right after the test was finished and then returned to idle temp in the lower 40s

 

[Shektron]

Either way, run that Prime95 stress test just to be sure. You don't want your CPU running at unstable settings, and Prime95 is the best way to avoid that.

 

[deded9090]

Okay I will try Prime 95. I have oc to 3.75ghz before at a lower voltage 1.24v and was playing bf1 fine in the upper 50s to mid 60s, I was just shocked to see how well the cpu did and why the cpu was so hot as 1.24v was not enough and my computer shut down hwenI tried before upping the voltage. To be fair though all 8 cores were at 100% loads which I will never come close in any normal use and my core usages for the first four or five cores are in the 40 to 50 percentile range when playing other high tier games like bf1.

 

[andres25salan]

1700 w/ Stock Cooler @ 3.71GHz 1.2750v
1588 cb

 

[Vellinious]

That's about right where it should be.

 

[benjamin.guidez]

I think i'm lucky
I have R7 1700 too and OC'd to 3.9Ghz at 1.32v on stock cooler, i've achieved 1725cb on Cinebench R15

 

[Finstar]

Don't run prime95. It doesn't work correctly on Ryzen cpus. Use this instead: http://www.majorgeeks.com/mg/getmirror/intelburntest,1.html
Btw my 1700 does 1780cb on @4GHz

 

[Vellinious]

Don't run prime95. It doesn't work correctly on Ryzen cpus. Use this instead: http://www.majorgeeks.com/mg/getmirror/intelburntest,1.html
Btw my 1700 does 1780cb on @4GHz

IBT is a heat machine...like using that furmark crap to test GPUs. Horrible..../smh

If anything, use Realbench for light loading, and either OCCT or AIDA64 for heavy load testing.

 

[Finstar]

Don't run prime95. It doesn't work correctly on Ryzen cpus. Use this instead: http://www.majorgeeks.com/mg/getmirror/intelburntest,1.html
Btw my 1700 does 1780cb on @4GHz

IBT is a heat machine...like using that furmark crap to test GPUs. Horrible..../smh

If anything, use Realbench for light loading, and either OCCT or AIDA64 for heavy load testing.

Like it or not, IBT and FurMark are the 2 most reliable and fastest ways to determine an unstable overclock.
You don't test a winch rated for 2 tons with 1 ton load to determine if it's safe.

 

[Vellinious]

Don't run prime95. It doesn't work correctly on Ryzen cpus. Use this instead: http://www.majorgeeks.com/mg/getmirror/intelburntest,1.html
Btw my 1700 does 1780cb on @4GHz

IBT is a heat machine...like using that furmark crap to test GPUs. Horrible..../smh

If anything, use Realbench for light loading, and either OCCT or AIDA64 for heavy load testing.

Like it or not, IBT and FurMark are the 2 most reliable and fastest ways to determine an unstable overclock.
You don't test a winch rated for 2 tons with 1 ton load to determine if it's safe.

OCCT and AIDA64 are more than capable for heavy load testing, even more so than IBT. IBT has it's uses, for very short stability tests in order to dial in settings for an overclock, but to use it for any kind of long term testing is just catastrophic stupidity. Same with furmark.

 

[Finstar]

OCCT and AIDA64 are more than capable for heavy load testing, even more so than IBT. IBT has it's uses, for very short stability tests in order to dial in settings for an overclock, but to use it for any kind of long term testing is just catastrophic stupidity. Same with furmark.

I know many cases of where several hours of Aida64 testing has not given away an unstable oc.
Aida64 used to be my go to testing tool until i realized that it sucked when my pc kept crashing during games even though Aida64 would never crash.
OCCT seems a bit better but still far from IBT.

And please explain to me: How is using IBT or FurMark for long term testing "catastrophic stupidity"?
I've so far never seen a crash on a system that was proven to be stable by either one.

 

[Vellinious]

OCCT and AIDA64 are more than capable for heavy load testing, even more so than IBT. IBT has it's uses, for very short stability tests in order to dial in settings for an overclock, but to use it for any kind of long term testing is just catastrophic stupidity. Same with furmark.

I know many cases of where several hours of Aida64 testing has not given away an unstable oc.
Aida64 used to be my go to testing tool until i realized that it sucked when my pc kept crashing during games even though Aida64 would never crash.
OCCT seems a bit better but still far from IBT.

And please explain to me: How is using IBT or FurMark for long term testing "catastrophic stupidity"?
I've so far never seen a crash on a system that was proven to be stable by either one.

If you're implying that either of those are the end all, beat all stability test, you couldn't be more wrong. There are no stability tests that cover everything (especially IBT and Furmark), and no amount of testing will guarantee complete stability. Which is why using multiple different tests, that test on a wider variety of instruction sets, it's ALWAYS the best option. Configurable testing platforms like AIDA64 and OCCT for heavy load testing, and Realbench for lighter load testing (some will even argue that it can be used for heavy load testing, because it does include AVX instrucitons), are all superior, and won't JUST create an ungodly amount of heat for a LINPACK based test. Granted, both AIDA64 AND OCCT both use LINPACK, but also have other instruction sets, and are also kept updated with new AVX instruction sets. Something IBT hasn't done....as it hasn't been updated since like 2012.

For GPU stability, I've found the best combination to be Superposition on 1080 extreme, Firestrike Ultra (graphics test 1 and 2 on loop), and Time Spy Extreme (graphics test 1 and 2 on loop)...each for at least an hour. If there's instability in your overclock, that'll very likely find it, without submitting your GPU to such harsh environs that furmark offers. The only thing furmark is good for, is cooking eggs...and GPUs. Any additional testing should be done in the game you're playing, or the hashing program that you're using...because as I stated before, there are NO stability testing programs that will catch everything. None.

The thing is....unless you're building a rig for a very specific purpose, like data mining or video editing, you aren't going to be pushing your rig as hard as ANY of these stress tests will. Most people are just gaming, and doing light compute work, so Realbench for a few hours is MORE than enough. But if you're data mining, video editing or doing anything else that requires precision and stability, overclocking is probably the worst thing you could possibly do. I have a rig built specifically for data mining using both CPU and GPU compute, and I know better than to overclock it....because there is one simple truth to overclocking stability, and that is, that there is NO test that will guarantee it.

 

[Finstar]

If you're implying that either of those are the end all, beat all stability test, you couldn't be more wrong. There are no stability tests that cover everything (especially IBT and Furmark), and no amount of testing will guarantee complete stability. Which is why using multiple different tests, that test on a wider variety of instruction sets, it's ALWAYS the best option. Configurable testing platforms like AIDA64 and OCCT for heavy load testing, and Realbench for lighter load testing (some will even argue that it can be used for heavy load testing, because it does include AVX instrucitons), are all superior, and won't JUST create an ungodly amount of heat for a LINPACK based test. Granted, both AIDA64 AND OCCT both use LINPACK, but also have other instruction sets, and are also kept updated with new AVX instruction sets. Something IBT hasn't done....as it hasn't been updated since like 2012.

For GPU stability, I've found the best combination to be Superposition on 1080 extreme, Firestrike Ultra (graphics test 1 and 2 on loop), and Time Spy Extreme (graphics test 1 and 2 on loop)...each for at least an hour. If there's instability in your overclock, that'll very likely find it, without submitting your GPU to such harsh environs that furmark offers. The only thing furmark is good for, is cooking eggs...and GPUs. Any additional testing should be done in the game you're playing, or the hashing program that you're using...because as I stated before, there are NO stability testing programs that will catch everything. None.

The thing is....unless you're building a rig for a very specific purpose, like data mining or video editing, you aren't going to be pushing your rig as hard as ANY of these stress tests will. Most people are just gaming, and doing light compute work, so Realbench for a few hours is MORE than enough. But if you're data mining, video editing or doing anything else that requires precision and stability, overclocking is probably the worst thing you could possibly do. I have a rig built specifically for data mining using both CPU and GPU compute, and I know better than to overclock it....because there is one simple truth to overclocking stability, and that is, that there is NO test that will guarantee it.

You're putting words in my mouth here.
All i was saying is that 24h Aida64 tests are pointless when IBT does the same thing in literally 10 minutes.
Same thing for FurMark, i often have it running in the background while i overclock so that the gpu crashes instantly when i go too high.

No matter how much you want to prove me wrong, I never said that there's a single program that can determine stability. It's just that running the heaviest workload available shows a pretty well which direction to go to, especially if you're looking for a stable voltage.

 

[Vellinious]

If you're implying that either of those are the end all, beat all stability test, you couldn't be more wrong. There are no stability tests that cover everything (especially IBT and Furmark), and no amount of testing will guarantee complete stability. Which is why using multiple different tests, that test on a wider variety of instruction sets, it's ALWAYS the best option. Configurable testing platforms like AIDA64 and OCCT for heavy load testing, and Realbench for lighter load testing (some will even argue that it can be used for heavy load testing, because it does include AVX instrucitons), are all superior, and won't JUST create an ungodly amount of heat for a LINPACK based test. Granted, both AIDA64 AND OCCT both use LINPACK, but also have other instruction sets, and are also kept updated with new AVX instruction sets. Something IBT hasn't done....as it hasn't been updated since like 2012.

For GPU stability, I've found the best combination to be Superposition on 1080 extreme, Firestrike Ultra (graphics test 1 and 2 on loop), and Time Spy Extreme (graphics test 1 and 2 on loop)...each for at least an hour. If there's instability in your overclock, that'll very likely find it, without submitting your GPU to such harsh environs that furmark offers. The only thing furmark is good for, is cooking eggs...and GPUs. Any additional testing should be done in the game you're playing, or the hashing program that you're using...because as I stated before, there are NO stability testing programs that will catch everything. None.

The thing is....unless you're building a rig for a very specific purpose, like data mining or video editing, you aren't going to be pushing your rig as hard as ANY of these stress tests will. Most people are just gaming, and doing light compute work, so Realbench for a few hours is MORE than enough. But if you're data mining, video editing or doing anything else that requires precision and stability, overclocking is probably the worst thing you could possibly do. I have a rig built specifically for data mining using both CPU and GPU compute, and I know better than to overclock it....because there is one simple truth to overclocking stability, and that is, that there is NO test that will guarantee it.

You're putting words in my mouth here.
All i was saying is that 24h Aida64 tests are pointless when IBT does the same thing in literally 10 minutes.
Same thing for FurMark, i often have it running in the background while i overclock so that the gpu crashes instantly when i go too high.

No matter how much you want to prove me wrong, I never said that there's a single program that can determine stability. It's just that running the heaviest workload available shows a pretty well which direction to go to, especially if you're looking for a stable voltage.

There's no such thing as a 10 minute stability test, period...doesn't matter what program you're using. If you're stability testing for 10 minutes, and calling it good, you're doing it wrong.

 

[Finstar]

If you're implying that either of those are the end all, beat all stability test, you couldn't be more wrong. There are no stability tests that cover everything (especially IBT and Furmark), and no amount of testing will guarantee complete stability. Which is why using multiple different tests, that test on a wider variety of instruction sets, it's ALWAYS the best option. Configurable testing platforms like AIDA64 and OCCT for heavy load testing, and Realbench for lighter load testing (some will even argue that it can be used for heavy load testing, because it does include AVX instrucitons), are all superior, and won't JUST create an ungodly amount of heat for a LINPACK based test. Granted, both AIDA64 AND OCCT both use LINPACK, but also have other instruction sets, and are also kept updated with new AVX instruction sets. Something IBT hasn't done....as it hasn't been updated since like 2012.

For GPU stability, I've found the best combination to be Superposition on 1080 extreme, Firestrike Ultra (graphics test 1 and 2 on loop), and Time Spy Extreme (graphics test 1 and 2 on loop)...each for at least an hour. If there's instability in your overclock, that'll very likely find it, without submitting your GPU to such harsh environs that furmark offers. The only thing furmark is good for, is cooking eggs...and GPUs. Any additional testing should be done in the game you're playing, or the hashing program that you're using...because as I stated before, there are NO stability testing programs that will catch everything. None.

The thing is....unless you're building a rig for a very specific purpose, like data mining or video editing, you aren't going to be pushing your rig as hard as ANY of these stress tests will. Most people are just gaming, and doing light compute work, so Realbench for a few hours is MORE than enough. But if you're data mining, video editing or doing anything else that requires precision and stability, overclocking is probably the worst thing you could possibly do. I have a rig built specifically for data mining using both CPU and GPU compute, and I know better than to overclock it....because there is one simple truth to overclocking stability, and that is, that there is NO test that will guarantee it.

You're putting words in my mouth here.
All i was saying is that 24h Aida64 tests are pointless when IBT does the same thing in literally 10 minutes.
Same thing for FurMark, i often have it running in the background while i overclock so that the gpu crashes instantly when i go too high.

No matter how much you want to prove me wrong, I never said that there's a single program that can determine stability. It's just that running the heaviest workload available shows a pretty well which direction to go to, especially if you're looking for a stable voltage.

There's no such thing as a 10 minute stability test, period...doesn't matter what program you're using. If you're stability testing for 10 minutes, and calling it good, you're doing it wrong.

Did i say that 10 mins will tell if your pc is stable? OR did i say that IBT does in 10 minutes what Aida64 does in 24 hours?
Maybe you should learn to read before trying to argue with someone.

 

[javierpizarro.r.jp]

RYZEN 1700 3.7ghz 1.2875v cinernbench r15 score 1941 , single core 177 , stock cooler

 

[Vellinious]

If you're implying that either of those are the end all, beat all stability test, you couldn't be more wrong. There are no stability tests that cover everything (especially IBT and Furmark), and no amount of testing will guarantee complete stability. Which is why using multiple different tests, that test on a wider variety of instruction sets, it's ALWAYS the best option. Configurable testing platforms like AIDA64 and OCCT for heavy load testing, and Realbench for lighter load testing (some will even argue that it can be used for heavy load testing, because it does include AVX instrucitons), are all superior, and won't JUST create an ungodly amount of heat for a LINPACK based test. Granted, both AIDA64 AND OCCT both use LINPACK, but also have other instruction sets, and are also kept updated with new AVX instruction sets. Something IBT hasn't done....as it hasn't been updated since like 2012.

For GPU stability, I've found the best combination to be Superposition on 1080 extreme, Firestrike Ultra (graphics test 1 and 2 on loop), and Time Spy Extreme (graphics test 1 and 2 on loop)...each for at least an hour. If there's instability in your overclock, that'll very likely find it, without submitting your GPU to such harsh environs that furmark offers. The only thing furmark is good for, is cooking eggs...and GPUs. Any additional testing should be done in the game you're playing, or the hashing program that you're using...because as I stated before, there are NO stability testing programs that will catch everything. None.

The thing is....unless you're building a rig for a very specific purpose, like data mining or video editing, you aren't going to be pushing your rig as hard as ANY of these stress tests will. Most people are just gaming, and doing light compute work, so Realbench for a few hours is MORE than enough. But if you're data mining, video editing or doing anything else that requires precision and stability, overclocking is probably the worst thing you could possibly do. I have a rig built specifically for data mining using both CPU and GPU compute, and I know better than to overclock it....because there is one simple truth to overclocking stability, and that is, that there is NO test that will guarantee it.

You're putting words in my mouth here.
All i was saying is that 24h Aida64 tests are pointless when IBT does the same thing in literally 10 minutes.
Same thing for FurMark, i often have it running in the background while i overclock so that the gpu crashes instantly when i go too high.

No matter how much you want to prove me wrong, I never said that there's a single program that can determine stability. It's just that running the heaviest workload available shows a pretty well which direction to go to, especially if you're looking for a stable voltage.

There's no such thing as a 10 minute stability test, period...doesn't matter what program you're using. If you're stability testing for 10 minutes, and calling it good, you're doing it wrong.

Did i say that 10 mins will tell if your pc is stable? OR did i say that IBT does in 10 minutes what Aida64 does in 24 hours?
Maybe you should learn to read before trying to argue with someone.

Man...you LITERALLY said exactly that.

"All i was saying is that 24h Aida64 tests are pointless when IBT does the same thing in literally 10 minutes."