Question 5900x and 5950x clocks

curley81

Proper
Sep 12, 2020
100
11
115
Does anyone know what sort of all core set clock speeds we can expect from the 5900x and the 5950x when they are released in a few weeks time? or am I jumping the gun?
 
Yes you're jumping the gun, that information hasn't been reviewed or published yet. You'll have to wait till reviewers get their hands on the chips and show us what they're truly capable of.
I don't think reviewers will tell us much along those lines...it will be the overclockers who do. If it's like it has been for Zen1, 1.5 and 2 the launch reviewers, the ones who get review kits, have to follow a 'script'.

Of course we'll get crazy reports from the internet...like the youtuber who undervolted Zen2, just looked at clocks and never realized how he was killing performance. A few days or weeks later the gold will start to dribble in. After the overclockers have bought their own samples and start squeezing it under LN2 to see what it will do.

I predict it will be like Zen2...only maybe even more so. No all-core headroom to speak of and it kills lightly threaded performance so badly when you do it can also kill it for gaming.
 
Last edited:
  • Like
Reactions: curley81
I don't think reviewers will tell us much along those lines...it will be the overclockers who do. If it's like it has been for Zen1, 1.5 and 2 the launch reviewers, the ones who get review kits, have to follow a 'script'.

Of course we'll get crazy reports from the internet...like the youtuber who undervolted Zen2, just looked at clocks and never realized how he was killing performance. A few days or weeks later the gold will start to dribble in. After the overclockers have bought their own samples and start squeezing it under LN2 to see what it will do.

I predict it will be like Zen2...only maybe even more so. No all-core headroom to speak of and it kills lightly threaded performance so badly when you do it can also kill it for gaming.

We'll just have to wait and see. Though these next releases are still looking promising nonetheless, if the leaks of AMD's CPU and GPU performance are as people say they are then it'd be really impressive for AMD to take a strong foothold against two major fronts.
 
  • Like
Reactions: curley81
We'll just have to wait and see. Though these next releases are still looking promising nonetheless, if the leaks of AMD's CPU and GPU performance are as people say they are then it'd be really impressive for AMD to take a strong foothold against two major fronts.

The leaks so far paint a pretty clear picture of Zen3 taking the gaming crown, and rather convincingly. All RX6000 might do is 'be competitive' with RTX3000 (excepting the outlier 3090). But that would be enough.
 
...highly binned that not even all of the cores were capable of reaching the maximum turbo clocks. ...
That's a bit disingenuous because it only hits maximum turbo clocks on one single core at a time anway. So it doesn't really matter if it's only doing it on 6 of 8 cores (as my bronze 3700X does) as there's always one of the six available to take the boost for the next thread that needs it.

The real performance of Zen2 comes from the mid-range clocks it can hold with all cores, not the max single core turbo clocks anyway. Just tweak PBO and give it good cooling to watch it rock. That's where seeing how Zen3 works will be the most interesting. Too bad I'm not going to get one.

And I'm pretty sure the 4.9-5Ghz clocks are just that: one single core in a turbo boost. Those aren't the leaks of interest; it's the Passmark scores that bedazzle. Especially single thread results. With such a dynamic processor and boost algorithm you can't really look at clocks to gauge performance.
 
Last edited:

curley81

Proper
Sep 12, 2020
100
11
115
thanks all. just trying to make up my mind on which 5000 series to upgrade to, current is 2700x, and weather its going to be even worth setting the core clocks or as with some of the 3000 series leaving the PBO to do the work.
 
thanks all. just trying to make up my mind on which 5000 series to upgrade to, current is 2700x, and weather its going to be even worth setting the core clocks or as with some of the 3000 series leaving the PBO to do the work.
As everybody says: it's too soon to say until reviewers at least have reported on their experiences. I'd watch for GamersNexus in particular as Steve always gives a very complete analysis, especially of 'overclock-ability'. GN's also likely to buy one so they can set aside any concerns with cherry picking of reviewer samples.
 
That's a bit disingenuous because it only hits maximum turbo clocks on one single core at a time anway. So it doesn't really matter if it's only doing it on 6 of 8 cores (as my bronze 3700X does) as there's always one of the six available to take the boost for the next thread that needs it.

The real performance of Zen2 comes from the mid-range clocks it can hold with all cores, not the max single core turbo clocks anyway. Just tweak PBO and give it good cooling to watch it rock. That's where seeing how Zen3 works will be the most interesting. Too bad I'm not going to get one.

And I'm pretty sure the 4.9-5Ghz clocks are just that: one single core in a turbo boost. Those aren't the leaks of interest; it's the Passmark scores that bedazzle. Especially single thread results. With such a dynamic processor and boost algorithm you can't really look at clocks to gauge performance.

You can try manually forcing each core to see if each of your cores is stable at the highest boost clock. Maybe you lucked out and all of your cores are capable. You can also check in Ryzen Master to see which cores are the strongest. I'm lucky that I have a late batch CPU that can not only reach, but exceed boost clocks, but I also have a very low ripple PSU and high phase motherboard.
 
You can try manually forcing each core to see if each of your cores is stable at the highest boost clock. ...
Only way it will stay stable for FAH at 4.4G is to raise voltage up to the 1.375-1.4 range...that's SVI2 TFN core voltage. I'm not interested in exploring early degradation phenomenon, so no thanks to that. Especially since it makes zero difference in gaming benches and insignificant difference in FAH TPF's, with thermals through the roof even under a 240mm AIO.
 

curley81

Proper
Sep 12, 2020
100
11
115
all good info. at the moment im running a my 2700x at 4.225ghz on all cores which is only 125mhz under the boost so minimal loss too single core but obviously upgrading to r9 5000 series there may be more single core loss so and trying to find decide exactly which cpu to buy and what to do with once i do buy it. but as you say we will find out more in the coming weeks

i have MSI x570 carbon pro for reference and corsair rm1000x psu
 

curley81

Proper
Sep 12, 2020
100
11
115
Only way it will stay stable for FAH at 4.4G is to raise voltage up to the 1.375-1.4 range...that's SVI2 TFN core voltage. I'm not interested in exploring early degradation phenomenon, so no thanks to that. Especially since it makes zero difference in gaming benches and insignificant difference in FAH TPF's, with thermals through the roof even under a 240mm AIO.

thermals are not a problem to me and being a 20 year amd user i know they can be a bit over safe with their recommendations on max temp and voltage i have a duel chamber duel open loop system with 1 240rad and 1 120 rad just for the cpu so im willing to push it to the max but even then not willing to lose 300 - 400 mhz on single core performance.
 
Only way it will stay stable for FAH at 4.4G is to raise voltage up to the 1.375-1.4 range...that's SVI2 TFN core voltage. I'm not interested in exploring early degradation phenomenon, so no thanks to that. Especially since it makes zero difference in gaming benches and insignificant difference in FAH TPF's, with thermals through the roof even under a 240mm AIO.

You don't have to test all of the cores for stability at once. You can temporaily disable the ones you are not testing.
 
You don't have to test all of the cores for stability at once. You can temporaily disable the ones you are not testing.
I tried that, didn't see the point in it. Also disabling SMT can make a major difference. In the end I'm more interested in all-around performance more than just seeing big clocks. For all the problems and risk of early degrading that comes with overclocking, the payoff just wasn't enough compared to an aggressive PBO. The best performing overclock loses too much single core performance with only a minor gain in multithreaded performance in some scenarios...mainly those that really heat up the CPU like Handbrake video encodes.
 
I tried that, didn't see the point in it. Also disabling SMT can make a major difference. In the end I'm more interested in all-around performance more than just seeing big clocks. For all the problems and risk of early degrading that comes with overclocking, the payoff just wasn't enough compared to an aggressive PBO. The best performing overclock loses too much single core performance with only a minor gain in multithreaded performance in some scenarios...mainly those that really heat up the CPU like Handbrake video encodes.

So you do or you don't know if all of your cores can actually hit the highest boost clock? I'm not surprised you can't keep it cool. Those fisher price AIOs really need a huge reservoir to be effective.
 
So you do or you don't know if all of your cores can actually hit the highest boost clock? I'm not surprised you can't keep it cool. Those fisher price AIOs really need a huge reservoir to be effective.
oh yes i do...i know 3 of my cores are boosting 25mhz over max, 3 more are hitting right on boost and the other two rarely if ever hit 25 under. It falls in line with the core ranking displayed in HWInfo. HWinfo, btw, not only shows the silicon quality ranking AMD gives the cores during binning but also the core preference order the scheduler uses when prioritizing work load to threads. The core preference order is determined not only on ranking but shared resources, e.g., caches.

i can't expect much...my processor is overall graded 'bronze' by CTR. I got it in August of last year so it definitely not one of the gold and platinum grade samples so common now coming off a mature process.

And it's not the reservoir or radiator that's limiting it as the liquid never thermally saturates; I can run FAH for days on end with processor temp never going over 75C. Prime95, small FFT's never gets above 82C range for over 4 hours before I shut it down. The limitation is the water block, just not enough fin surface area to transfer the heat to the liquid.

I'm quite aware of liquid cooling's limitations...both AIO and CCL. AIO's can saturate the liquid (as Linus points out on 120mm rad's) and degrade as the microfins in the water block clog (I expect mine will, but it's at 3 1/2 years & holding well for now). But CCL is expensive and has to be serviced (or the same problem as the WB fins also clog).

So for the trouble and expense of a CCL I MAY get few dozen more points in multi-thread performance and the pleasure of watching ST performance decline and micro-stuttering return, or worsen, when gaming. That's simply not a bargain I can rationalize.
 
Last edited:
  • Like
Reactions: damric
oh yes i do...i know 3 of my cores are boosting 25mhz over max, 3 more are hitting right on boost and the other two rarely if ever hit 25 under. It falls in line with the core ranking displayed in HWInfo. HWinfo, btw, not only shows the silicon quality ranking AMD gives the cores during binning but also the core preference order the scheduler uses when prioritizing work load to threads. The core preference order is determined not only on ranking but shared resources, e.g., caches.

i can't expect much...my processor is overall graded 'bronze' by CTR. I got it in August of last year so it definitely not one of the gold and platinum processors that are so common now coming off a mature process.

Yeah I guess I lucked out with a silver for my Ryzen 2. I really lucked out on one of the 1600 AFs I bought early this year. Most of them were good overlockers, but one could do 4350MHz with good voltage and had a strong IMC too.
 
...
So for the trouble and expense of a CCL I MAY get few dozen more points in multi-thread performance and the pleasure of watching ST performance decline and micro-stuttering return, or worsen, when gaming. That's simply not a bargain I can rationalize.

Oh...and to add to that I'm sure I'd need a new motherboard too. I have a B450 Mortar, with a 4 phase VRM. It's a BIG 4 phases so it runs cool even with an 8 core in the midst of Prime95, but 4 is not 10 or 12. At the the edge of voltage margin more phases will keep power delivery smooth and clean with, in particular, little to none of the undershoot that makes the processor lock up. I can imagine that's also happening to me.

But again, put out the expense for one of these new 12 phase motherboard (the B550 ROG Strix-A looks nice) and case (they're all ATX) all I get for that is a few dozen more points in multithread BM performance and at BEST similar ST. It's still not a bargain I want to make.
 

curley81

Proper
Sep 12, 2020
100
11
115
thanks for the learning guys appreciate it the b450 mortar is what im putting my current 2700x into once i get the r9 5000 cpu. when i do i will have a little tweek session and see what i will do thank you again great discussion
 

curley81

Proper
Sep 12, 2020
100
11
115
I opted for a 5800x and from what ive read the boost clock 4.7ghz is acheveable on all clocks is this accurate? seen bench marks at 4.725 and 4.75 all clocks with liquid cooling on various MOBOs i would be happy with 4.6 all clocks is there anyone here with more knowledge on this?