- Jan 29, 2005
- 5,196
- 197
- 106
This... could be a long post so be prepared (just get two cups of coffee ready, should do the trick).
Lately I've been over-clocking my CPU, it's very humble stuff. It's also the first time I OC in many years.
So I got a Ryzen 5 1600. Default is 3.2Ghz. I still have the default AMD cooler (it's decent though, I do like it). With that in mind, I decided to "just" go with a +400 Mhz boost, so went with a 3.6Ghz OC. I read about it OC'ing this particular chip of course, and figured out I'd start with only bumping the vCore by two 'notches' in the BIOS (goes up by pressing directional arrow keys, I don't decide the actual voltage; it just goes up by very small increments with each key press).
According to HMonitor, it looks like this right now for 3.6Ghz:
vCore Min 0.480v (varies a bit, stays under 1.0v), and vCore Max 1.286v (does not vary, stays there)
The only thing I did with the vCore (in the BIOS) was to increase it by two ticks as I said, that's it. The other thing I did (outside of vCore) was to change the Loadline Calibration - I think it's called - from Normal to 'Medium', and that's about it (not sure if it helps, but some OC'ing guides I've seen for this chip recommended it, some didn't even mention it; it seems to just be a question of testing on a per-chip basis). After that, nothing else.
Then of course I made tests. The first one was Prime95, Small FFTs, test lasted for about 13 hours, no errors and HMonitor and MSI Afterburner both reported a max CPU Temp of 78ºC (never went above). As of now it idles at 31ºC (of course, it varies based on room temp a bit; right now here it's Autumn, temps outside are cool, so indoor temps are sitting around the usual 25ºC to 26ºC). So, my other "tests" were games and some Benchmark softwares (usual 3DMark, Heaven, etc). My games run fine, no CTDs, no blue screens, no freezes (tested a bunch of them, modern ones, older ones, DX9, 10, 11, etc).
Good so... I "got my over-clock" then, great; it's gonna stay there (I might go higher but not unless I get a better cooler one day, if I ever bother with that).
So, you might be asking yourself then: "Wtf then is the point of this if everything works?"
Well, what an astute question there, my friend.
The thing is all the CPU Benchmark scores (specifically the CPU scores, not including GPU scores) I've had so far are pretty much all relatively lower than exact same systems I compared them to. Just taking 3DMark CPU score results, for example. In all of the ones I compared my scores with, the other Ryzen 5 1600s I've seen had a higher score (as a final score, and on-paper in terms of actual performance during the Benchmark with frames-per-second) by a couple of hundred points. In fact, most of the scores I compared 'same systems' with, those systems had better CPU scores with a Stock Ryzen 5 still sitting at 3.2Ghz, while mine trailed behind at 3.6Ghz. For example, my latest total score was something like 7150 (give or take a few points), while other identical systems' total score would be around 7800+.
Now, that's the total score (it does include GPUs), and of course some people OC their GPU as well and it does help bring the numbers up of course. However, as I specified earlier I then only pay attention to the CPU results (score and FPS). In about 95% of cases my Ryzen 5 definitely under performs, sometimes even performs less than Stock speed Ryzen 5s (to repeat myself).
Alright, so 'normally' I don't quite care that much about Benchmarking scores. However, I took the time to OC a bit, so... might as well check to see if it was a successful OC, right? Now I'm just scratching my head since I'm not sure what variables here need to be taken in consideration and that inevitably brings dozens of questions in my cranium, and being the person I am I must comply with my curiosity and ask away.
1) First of all, not sure if that's even a 'thing' in OC in general. I'm not qualified to know if it is, but I still wonder about the following: Is there such a thing as a Stable-But-Underperforming OC? In other words, and for example, someone would OC their CPU with just the right amount of power for it to be Stable, BUT... somehow not enough juice (vCore / power) for said OC value to actually perform 'as it should at that speed' so to speak? Does that happen at all? Or is it really a case of "If you don't have enough power going for any OC, it's just going to crash, period" and there's never going to be lower performance involved?
I'm asking this because maybe only increasing my vCore by 'two ticks' was good enough to get it Stable, but not "high enough" to actually get my 'real' 3.6Ghz performance (I.E. I'm "choking" my CPU? damn, poor thing).
2) Now, yes it's a Ryzen (and first Gen at that). I am fully aware that Ryzen CPUs absolutely adore fast RAM. The problem with Benchmarks, however, is that there's never a completely separate system RAM score on its own. So, even though I DO know that my own system RAM's speed COULD negatively affect my CPU, I don't actually know how much that might be the case here. So here my main questions are: How do I know how much my current RAM might be having a negative impact on my CPU performance? and What do I do if that ends up being the case, between simply trying to increase the RAM speed, or changing the RAM timings?
There's a potential secondary 'set' of issues with this, however. My own RAM is from Hynix. It is rated to run at 'up to' 3000Mhz. From the amount of OC I did try to do with them so far I simply cannot get them any higher than 2666Mhz even at 1.4v regardless of what else I try (and believe me, I tried). If THIS is the number one and perhaps ONLY reason as to why my CPU overall 'under performs' (because of my RAM and its maximum speed I managed to get it to) then my main question about RAM still imposes itself: HOW do I make sure that it IS the case?
It's easy to say (such as it is stated in reviews often) something like "We recommend getting your RAM going at least at 2666Mhz for Ryzen..." and so on, but the problem with that is that we should technically have an impossibly-long sheet of ALL compatible RAM sticks from specific vendors at specific Rated / Base speeds combined with specific Mobos with actual Benchmark results to REALLY put such statements in perspective for a gamer's point of view. I can't just have a 'feeling' for it, that my RAM probably does "slow down" my system. Well... oh, ok. By how much then? How "better" then would it perform if THIS particular RAM would be running at 2999Mhz instead?
To be honest CPU-to-RAM voltage / frequency performance ratios (or vice versa) always confused me (then again yeah, I'm a beginning with OC essentially; but I have questions nonetheless).
3 (Conclusion):
The gist is... all across the Interweb's scores I compared mine to (with identical systems), my CPU-specific scores are always lower (even against those running it at Stock, which is probably THE one part that confuses me). And I don't know why. If the RAM is causing it, I'm not sure by how much it impacts the 'real' performance I would get otherwise. If it's not the RAM, then I don't know what to do other than to just blindly increase the vCore on my CPU OC "just because", even though it is apparently very stable. If it's not due to the vCore, or even due to my RAM, then at that point I really don't know what to think about.
Is there perhaps anything else outside of OC per se I might have to look for? Maybe a specific setting or two in the BIOS I should check for? Is there something in Windows 10 itself I should activate or disable? Maybe the Power options? (side note: it's currently set to AMD's own Power options, as it was pretty much recommended to do everywhere I looked for those running AMD Ryzen CPUs)
Basically I just want to make sure that there's either 'nothing else' I can do at this point and just be happy with the scores I have. Or there's in fact a few things I could look into and would just appreciate some guidance from you OC'ing magicians around.
Thanks.
EDIT: Alright so I decided to 'fight' it and got a Memory OC to 2933 instead (at specified Corsair specs). I thought "Ok, so if my Memory is the culprit, let's just OC it to at least what Corsair says it can run to", and see where the score goes from there and if it does help the CPU perform 'normally' since Memory would run faster (which is what Ryzen CPUs crave for; fast Memory).
So, generally-speaking, in Desktop for most normal activities it seems to be just a tad bit 'snappier', maybe (could be placebo, even though technically-speaking the speed is indeed higher). However, in my scores coming from three Benchmarks (with comparable results online) still have a noticeably lower CPU-related score than most (although not all) other identical systems (only taking into account the ones that do not have a higher CPU overclock to start with).
The total scores I got were a little bit higher, to negligible extent in all cases. But, yeah, still lower than most others. So at this point I'll stop worrying about all this. It only reminds me why I "quit" over-clocking years ago, too stressful, too many variables to look for and worry about; too much time to dedicate for investigating issues. I did end up getting stable OCs in all this story anyway so... yay happy panda time.
The gist is now please do ignore all this (I suppose at this point that Wall Of Text x2 did the job anyway) since I'll just proceed to delete all my Benchmark software and such from this point and just keep my current specs untouched from now on.
Lately I've been over-clocking my CPU, it's very humble stuff. It's also the first time I OC in many years.
So I got a Ryzen 5 1600. Default is 3.2Ghz. I still have the default AMD cooler (it's decent though, I do like it). With that in mind, I decided to "just" go with a +400 Mhz boost, so went with a 3.6Ghz OC. I read about it OC'ing this particular chip of course, and figured out I'd start with only bumping the vCore by two 'notches' in the BIOS (goes up by pressing directional arrow keys, I don't decide the actual voltage; it just goes up by very small increments with each key press).
According to HMonitor, it looks like this right now for 3.6Ghz:
vCore Min 0.480v (varies a bit, stays under 1.0v), and vCore Max 1.286v (does not vary, stays there)
The only thing I did with the vCore (in the BIOS) was to increase it by two ticks as I said, that's it. The other thing I did (outside of vCore) was to change the Loadline Calibration - I think it's called - from Normal to 'Medium', and that's about it (not sure if it helps, but some OC'ing guides I've seen for this chip recommended it, some didn't even mention it; it seems to just be a question of testing on a per-chip basis). After that, nothing else.
Then of course I made tests. The first one was Prime95, Small FFTs, test lasted for about 13 hours, no errors and HMonitor and MSI Afterburner both reported a max CPU Temp of 78ºC (never went above). As of now it idles at 31ºC (of course, it varies based on room temp a bit; right now here it's Autumn, temps outside are cool, so indoor temps are sitting around the usual 25ºC to 26ºC). So, my other "tests" were games and some Benchmark softwares (usual 3DMark, Heaven, etc). My games run fine, no CTDs, no blue screens, no freezes (tested a bunch of them, modern ones, older ones, DX9, 10, 11, etc).
Good so... I "got my over-clock" then, great; it's gonna stay there (I might go higher but not unless I get a better cooler one day, if I ever bother with that).
So, you might be asking yourself then: "Wtf then is the point of this if everything works?"
Well, what an astute question there, my friend.
The thing is all the CPU Benchmark scores (specifically the CPU scores, not including GPU scores) I've had so far are pretty much all relatively lower than exact same systems I compared them to. Just taking 3DMark CPU score results, for example. In all of the ones I compared my scores with, the other Ryzen 5 1600s I've seen had a higher score (as a final score, and on-paper in terms of actual performance during the Benchmark with frames-per-second) by a couple of hundred points. In fact, most of the scores I compared 'same systems' with, those systems had better CPU scores with a Stock Ryzen 5 still sitting at 3.2Ghz, while mine trailed behind at 3.6Ghz. For example, my latest total score was something like 7150 (give or take a few points), while other identical systems' total score would be around 7800+.
Now, that's the total score (it does include GPUs), and of course some people OC their GPU as well and it does help bring the numbers up of course. However, as I specified earlier I then only pay attention to the CPU results (score and FPS). In about 95% of cases my Ryzen 5 definitely under performs, sometimes even performs less than Stock speed Ryzen 5s (to repeat myself).
Alright, so 'normally' I don't quite care that much about Benchmarking scores. However, I took the time to OC a bit, so... might as well check to see if it was a successful OC, right? Now I'm just scratching my head since I'm not sure what variables here need to be taken in consideration and that inevitably brings dozens of questions in my cranium, and being the person I am I must comply with my curiosity and ask away.
1) First of all, not sure if that's even a 'thing' in OC in general. I'm not qualified to know if it is, but I still wonder about the following: Is there such a thing as a Stable-But-Underperforming OC? In other words, and for example, someone would OC their CPU with just the right amount of power for it to be Stable, BUT... somehow not enough juice (vCore / power) for said OC value to actually perform 'as it should at that speed' so to speak? Does that happen at all? Or is it really a case of "If you don't have enough power going for any OC, it's just going to crash, period" and there's never going to be lower performance involved?
I'm asking this because maybe only increasing my vCore by 'two ticks' was good enough to get it Stable, but not "high enough" to actually get my 'real' 3.6Ghz performance (I.E. I'm "choking" my CPU? damn, poor thing).
2) Now, yes it's a Ryzen (and first Gen at that). I am fully aware that Ryzen CPUs absolutely adore fast RAM. The problem with Benchmarks, however, is that there's never a completely separate system RAM score on its own. So, even though I DO know that my own system RAM's speed COULD negatively affect my CPU, I don't actually know how much that might be the case here. So here my main questions are: How do I know how much my current RAM might be having a negative impact on my CPU performance? and What do I do if that ends up being the case, between simply trying to increase the RAM speed, or changing the RAM timings?
There's a potential secondary 'set' of issues with this, however. My own RAM is from Hynix. It is rated to run at 'up to' 3000Mhz. From the amount of OC I did try to do with them so far I simply cannot get them any higher than 2666Mhz even at 1.4v regardless of what else I try (and believe me, I tried). If THIS is the number one and perhaps ONLY reason as to why my CPU overall 'under performs' (because of my RAM and its maximum speed I managed to get it to) then my main question about RAM still imposes itself: HOW do I make sure that it IS the case?
It's easy to say (such as it is stated in reviews often) something like "We recommend getting your RAM going at least at 2666Mhz for Ryzen..." and so on, but the problem with that is that we should technically have an impossibly-long sheet of ALL compatible RAM sticks from specific vendors at specific Rated / Base speeds combined with specific Mobos with actual Benchmark results to REALLY put such statements in perspective for a gamer's point of view. I can't just have a 'feeling' for it, that my RAM probably does "slow down" my system. Well... oh, ok. By how much then? How "better" then would it perform if THIS particular RAM would be running at 2999Mhz instead?
To be honest CPU-to-RAM voltage / frequency performance ratios (or vice versa) always confused me (then again yeah, I'm a beginning with OC essentially; but I have questions nonetheless).
3 (Conclusion):
The gist is... all across the Interweb's scores I compared mine to (with identical systems), my CPU-specific scores are always lower (even against those running it at Stock, which is probably THE one part that confuses me). And I don't know why. If the RAM is causing it, I'm not sure by how much it impacts the 'real' performance I would get otherwise. If it's not the RAM, then I don't know what to do other than to just blindly increase the vCore on my CPU OC "just because", even though it is apparently very stable. If it's not due to the vCore, or even due to my RAM, then at that point I really don't know what to think about.
Is there perhaps anything else outside of OC per se I might have to look for? Maybe a specific setting or two in the BIOS I should check for? Is there something in Windows 10 itself I should activate or disable? Maybe the Power options? (side note: it's currently set to AMD's own Power options, as it was pretty much recommended to do everywhere I looked for those running AMD Ryzen CPUs)
Basically I just want to make sure that there's either 'nothing else' I can do at this point and just be happy with the scores I have. Or there's in fact a few things I could look into and would just appreciate some guidance from you OC'ing magicians around.
Thanks.
EDIT: Alright so I decided to 'fight' it and got a Memory OC to 2933 instead (at specified Corsair specs). I thought "Ok, so if my Memory is the culprit, let's just OC it to at least what Corsair says it can run to", and see where the score goes from there and if it does help the CPU perform 'normally' since Memory would run faster (which is what Ryzen CPUs crave for; fast Memory).
So, generally-speaking, in Desktop for most normal activities it seems to be just a tad bit 'snappier', maybe (could be placebo, even though technically-speaking the speed is indeed higher). However, in my scores coming from three Benchmarks (with comparable results online) still have a noticeably lower CPU-related score than most (although not all) other identical systems (only taking into account the ones that do not have a higher CPU overclock to start with).
The total scores I got were a little bit higher, to negligible extent in all cases. But, yeah, still lower than most others. So at this point I'll stop worrying about all this. It only reminds me why I "quit" over-clocking years ago, too stressful, too many variables to look for and worry about; too much time to dedicate for investigating issues. I did end up getting stable OCs in all this story anyway so... yay happy panda time.
The gist is now please do ignore all this (I suppose at this point that Wall Of Text x2 did the job anyway) since I'll just proceed to delete all my Benchmark software and such from this point and just keep my current specs untouched from now on.
Last edited:
Facebook
Twitter