A question for anyone with a 5600X.....

[NoddyPirate]

Just to add - any reports I’ve seen of near 5 GHz for the 5600X have been single core boost speeds with a change to the PBO in BIOS - effectively boosting just the one or two best cores rather than the whole lot together.

But that’s still an OC. I certainly haven’t read of any at 5 GHz without tweaking something...... 🤔

 

[Scott]

I’m not sure Scott if we still have the 5950 or 5900 on our minds?

AMD’s own ads say 4.6 GHz single core boost - GN saw 4.65 GHz in their review here - same as me:

5600X review

They could OC it to 4.8 GHz all core at 1.4 Vcore but that wasn’t stable for all tests so they have another entry for OC at 4.7 GHz.

And in their super tuned battle after 6 hours of tuning they again had it at 4.8 GHz all core with the occasional need to drop to 4.7 GHz for some tests.

5600X vs 10600K OC showdown

In terms of the silicon lottery however I most likely haven’t lucked out as I do have two cores with pretty low CPPC tags. But my results - stock or OC - still aren’t far off GN’s at all really.

The biggest difference is in the stock multi core frequency where they saw a higher average frequency than I do - by about 100 MHz. Given my stock multi core speeds are temperature dependent I assume their superior cooling for their tests drove the difference here? And perhaps that’s where your case lids off experiment would be an interesting one to look at!

You may well be right, it was possibly even the 5800X I had in my head. I definitely recall whimsings of single core 5ghz on the 5600X but it was perhaps with a mild OC.

Multi-core has always been limited with AMD though and there's no real change there as yet

Cooling has a huge part to play in multi-core use of the AMD chips. As said, they need to be way into optimal cooling ranges to allow for higher frequencies. You can bypass this of course though, things can get toasty pretty quick though..... hence why AMD set the bar fairly low.

 

[NoddyPirate]

But it just seems to show how far AMD have come that they’ve got their processors working so well that outside of very specific use cases, no tweaking is required.

100% true - and I am only getting the first insight into just how true this is in my thought experiement with the help of @SpyderTracks and @Scott !!

You can OC a chip like the 5600X in two ways really - you can push all cores together to reach a little past the stock single core boost frequency - but it is already so tuned out of the box that you probably won't get far past that number. You end up with a chip that does a good bit better in mutli core loads but the same or worse single core.

Or, you can OC it to max out PBO for high single core boosts - but only if your motherboard allows you exceed the AMD limits - which my ASUS board doesn't. (EDIT - turns out it can - who knew! 🙂 but only up to 200 MHz over AMD limits - some mobos allow up to 600 MHz) By pushing the best cores this way you will see single core boost a good bit higher than stock. You'll see better single core boost speeds and benchmarks results but might not really see much real life improvement it seems.

Funny how benchmarks have become such a badge of honour but those posting them then go and play their games at the exact same FPS they had beforehand!

 

[NoddyPirate]

You may well be right, it was possibly even the 5800X I had in my head. I definitely recall whimsings of single core 5ghz on the 5600X but it was perhaps with a mild OC.

Multi-core has always been limited with AMD though and there's no real change there as yet

Cooling has a huge part to play in multi-core use of the AMD chips. As said, they need to be way into optimal cooling ranges to allow for higher frequencies. You can bypass this of course though, things can get toasty pretty quick though..... hence why AMD set the bar fairly low.

Sorry @Scott - I seem to have a knack at crossing posts with others!

Yes 5 GHz seems possible for single core boosts with a BIOS tweak - which sadly my board won't allow....(EDIT - actually it allows up to a 200 MHz boost over stock but with high V’s)

I'm off to cool my board as much as possible and see if it has any effect on my stock mutli-core boosts. Damn you and your ideas!!!

 

[Scott]

You can also see the link of appropriately matched VRM capacity and how important it is for certain uses. It's not simply a matter of the maximum capacity either, it's the accuracy of the control that plays a huge part in stability with boosting.

 

[NoddyPirate]

You can also see the link of appropriately matched VRM capacity and how important it is for certain uses. It's not simply a matter of the maximum capacity either, it's the accuracy of the control that plays a huge part in stability with boosting.

Yes 100% @Scott - my motherboard is a perfect example of one incapable of proper OC - for stability as you say no doubt - but even just the lack of monitoring sensors and so on - means it’s shot in the dark stuff with it when it comes to power supplied.

The 5600X is a low power chip - but the thoughts of trying what I’m doing on my mobo with a big power draw CPU gives me shivers.....

 

[NoddyPirate]

Right so - my PC room was only at 17 degrees this morning - so I could get a little better temps with full fans than previously....I didn't remove any case panels because I couldn't be bothered!

For a stock all core boost, it seems at lower temps it can push a higher voltage through the CPU giving higher clock speeds for the same power supplied. It makes sense I guess from an electronics standpoint? Resistance goes up with increasing temperature - so with the same limit on power sent from the VRM, the CPU actually gets less juice at the business end? Perhaps?

I got the following with fans at min, low and max speeds - all showing the exact same stock 76 W supplied.

76 degrees - Core Voltage 1.169 V - Sustained Freq 4.100 GHz - Max Freq 4.250 GHz (was closer to 4.000 GHz in the past nearer 80 degrees)
64 degrees - Core Voltage 1.175 V - Sustained Freq 4.200 GHz - Max Freq 4.325 GHz
55 degrees - Core Voltage 1.194 V - Sustained Freq 4.300 GHz - Max Freq 4.350 GHz

The single core boosts were all identical at 4.650 GHz no matter what - but I imagine the time they would be held would change accordingly.

So I think it all shows that cooling affects Ryzen stock behaviour too - even when you are miles away from the CPU temperature limit. Who knew! I always assumed that below some fairly high upper threshold, stock performance would always be the same. (GN were seeing 4.4 GHz which makes sense with their fancy pants AIO they were using I suppose.)

So cooling does appear to impact even stock performance when all cores are involved and even at relatively low temperatures. Interesting! (...to me anyway)

So - that actually answers my original question at the start of this thread. Thanks so much @Scott !

 

[NoddyPirate]

Just for fun - Manual PBO to increase the PPT to 86 W in BIOS - limiting the current also just to be safe. Fans still at max:

61 degrees - 1.250 V - Sustained all core at 4.425 GHz.

 

[Scott]

Just for fun - Manual PBO to increase the PPT to 86 W in BIOS - limiting the current also just to be safe. Fans still at max:

61 degrees - 1.250 V - Sustained all core at 4.425 GHz.

That's a good return, with all things considered

 

[NoddyPirate]

That's a good return, with all things considered

Well apart from my cooling being in Airplane Mode - as in, it sounds like an airplane!

With my normal fan curves - it's all quiet again at 64 degrees and 4.375 GHz, which isn't too bad but not much to write home about at the same time....maybe 125 MHz better than my personal stock.

I'll play with the PPT a little more perhaps and see how it all behaves. I've deleted Clocktuner for now as I couldn't get it to behave itself and have uninstalled it as it was throwing errors at startup too.

Learned loads here - so thanks again to both of you!!!

 

[Scott]

Just imagine you had got the 240 lite, you would be pushing 6Ghz

 

[NoddyPirate]

Just imagine you had got the 240 lite, you would be pushing 6Ghz

You may be right!

On the cooling (I'm reluctant to mention the war here) as far as I can tell my yoke is supposed to rated for 180 TDP. I know TDP and Watts are not the same thing, particularly with Ryzen, but extrapolating out my results in terms of deg/W, I would expect my cooler to explode at a TDP equaivalent of about 125 or so. At 110 W drawn from the VRM I head into the mid to high 70's. And if you want my system to quiet while it's at 110W you could try having in a different room to your mouse, keyboard and monitor!

For a 5600X it's fine for sure, but for anything other than a fairly low power CPU, say 115W max supplied, including peaks - which therefore excludes even just the 5800X - I think we all agree that these yokes are not up to the task.

(Forgot to add also that with simply changing the BIOS PBO clock override to 200 MHz I can get single core boosts to 4.825 GHz which sounds great - but with ugly voltages over 1.42 V sustained - so it's not for me. 😱)

 

[Scott]

TDP and Watts should be close, but there are losses to consider and, as you can imagine, tonnes of assumptions made.

I take things like this, always, with a pinch of salt. Like the 0-62mph time listed for cars.

I would be curious if the TDP is max or sustained also. The great advantage with air coolers is the dissipation before being heat soaked. They do amazingly well for short periods of time.

 

[NoddyPirate]

TDP and Watts should be close, but there are losses to consider and, as you can imagine, tonnes of assumptions made.

Yep - as an example the 5600X TDP is 65W buts is set to 76W PPT stock. So about a 17% difference. I'm finding it hard to find the correct numbers for the 5800X - someone who has one would be best placed - but it seems a 105W rating alongside a 142W PPT stock setting which is kind of mad - I assume that's not right....

(EDIT - Deleted the rest of this post as it was just waffle! 🙂)

 

[NoddyPirate]

Edited some posts of mine above to keep things consistent - and to reduce some of the waffle!! 🤦‍♂️😀

 

[NoddyPirate]

Just some interesting additional info from the one and only GN..... A very ineteresting read and explains why quoted figures can be so misleading and confusing when compared to actual results.....

What does TDP mean?

This is one of their few articles where they don't have or even recommend a TLDR - but a key takeaway nonetheless from the conlcusion:

"TDP is an overly reductive number, and it’s only becoming more difficult to define with features like AMD’s Precision Boost 2 that make every individual processor perform differently in different situations. The safe assumption is that any TDP rating, whether it’s from AMD or Intel, is just an easy-to-swallow marketing device that looks good on a box."

 

[NoddyPirate]

Right so - my last epic I promise. (This one really is huge though!) I found it really tough to get information on overclocking from the web or from the forum, so I felt it would be useful to lay out everything I've learned and played with in case anyone else wants the info.

I was going to put this in a new thread rather than have it buried in Page 2/3 of a waffly thread - Mods you can decide what you want to do with it.........

 

[NoddyPirate]

OVERCLOCKING RYZEN 5000

This was an idiots voyage given my newbie status and the potential to harm my system – but my lust for understanding won out in the end. Zen 3 Ryzen is a clever piece of kit I must say.

My 5600X has an advertised base clock of 3.7 GHz with max single core boosts to 4.6 GHz – although it seems common to be getting boosts to 4.65 GHz which is what I see.

Multi-Core boosts are dependent on various elements, not the least of which is temperature. With my PCS Tower Cooler holding the CPU at 62 degrees under 100% load, I see all core boosts typically around 4.25 GHz. From what I can tell that’s about 150 MHz worse than you might see with a decent AIO strapped on instead keeping it in the mid 50’s or better.

Voltage wise, my cores were supplied with a max of around 1.180 V under sustained multi-core loads and 1.356 V for single core loads. But in general, voltages could vary between 0.96 and 1.44 V under normal use. AMD seem to say up to 1.5 V is normal but I imagine only as transients.

When it comes to overclocking then, there are a few options. But I’ll just explain the stock settings hiding in the background….

The 5600X has TDP rating of 65W. On my motherboard that equated to a max power draw of 76W, a max Peak Current of 90A and a max sustained current of 60A. These numbers are referred to as PPT, EDC and TDC respectively – and they are standard Ryzen numbers for a 65W CPU.

(Previously the standard PPT for a 65W TDP Chip was apparently 88W it seems – but this was a bit pointless as the CPU would hit the EDC limit before it got to 88W anyway, so 76W makes much more sense – maxing out both PPT and EDC at about the same time.)


OVERCLOCK OPTIONS

AUTO-OC

The easiest appears to be simply to click on “Auto OC” on Ryzen Master, or enable it in BIOS and sit back and watch it. This is a crude method really - It sets the PPT, EDC and TDC limits to the maximum allowed by the motherboard – for my PRIME B550 PLUS that means 395 W, 190A and 160A. Other motherboards have MUCH higher limits – more like perhaps 1000W for the TUF X570 for example. This is all a bit silly – as the CPU will never ask for that much in any category anyway. Really, you are just unleashing the CPU to do whatever it wants and to draw as much power as it can handle. I don’t think that’s the best for longevity personally of course no matter how cool you can keep it.

For my 5600X with “Auto OC” the power (PPT) increased by 60% to 105W. EDC and TDC saw lesser gains but were still high – up to 120A for EDC. My cooler running at 75% held the temperature at 72 degrees – and for all that energy and heat? Multi-Core boosts were now at 4.525 GHz – still well below the stock single core boost number. So what about single core boosts then?

In Ryzen Master you have to set a “Precision Boost Override” (not Overdrive) value when you enable “Auto OC”. This is basically the additional boost for Single Core performance only (and Base Clocks actually) – but it does not impact all core performance as that is ultimately power limited. Since we’re looking at the extremes, I set it to the max my motherboard will allow – which is 200 MHz. Other motherboards can allow higher settings here. This gave me 4.825 GHz single core speeds – but with peak core voltages approaching 1.45 V – now I don’t fully understand the limits of the latest Ryzen chips but 1.45 is far too high for my comfort! So, on to another method…..


MANUAL OC

Manual Overclocking with Ryzen Master is simple – you set a voltage and a frequency and hope for the best. Mobo power limits apply again. I was careful here – and conservative - testing each new setting fully with benchmarks and stress tests to guarantee stability before moving on to the next one.

The problem though is that you are pushing all cores together. Single Core performance is basically irrelevant as it will also top out at whatever frequency you have set. I managed to get to 4.7 GHz all core (and single core obviously) at 1.325 V. Anything more than that was unstable. Seeing that even Gamers Nexus were also unable to get their 5600X copy beyond 4.7 GHz consistently, made me feel better though! Ultimately, the limit of the weakest core had been reached.

For the 5600X this meant I had all cores at a slightly higher frequency than even the stock single core boost speed. A great result in some ways. But the power draw was high – around 112W peak – higher than Auto-OC which I didn't expect - with my cooler holding things around 76 degrees. So, my multi-core performance saw a huge increase, but single core was basically unchanged. You would want to be doing nothing other than full time heavy multi-core work for this to be a good idea!

But for CPU’s like the 5900X and 5950X the stock single core boosts are so high to start with that you are almost certainly going to fail to get all cores close to those frequencies pushing them all at the same time. Manual Overclocking for those CPU’s will simply ensure that single core performance is significantly worse than with stock settings.

This is where Ryzen is so clever and finely tuned. Only the best few cores in your CPU do the single core boosting. The less capable ones sit back and watch. This means you won’t be able to keep them all in line for long when you push them manually.


PRECISION BOOST OVERDRIVE (PBO)

This can be enabled in Ryzen Master or BIOS. With this you can actually set the PPT, EDC and TDC limits yourself. The chip will then do whatever it can up to whichever of your limits it hits first. This is getting closer to a useful setup and my results were very interesting. First I added just 10W to my PPT limit for 86W (and 8A to EDC as my maths said that would be just enough – and the default TDC would be irrelevant). Now my all core performance was at 4.450 GHz. That’s 200 MHz better than stock and only 75 MHz less than “Auto-OC” managed with 105W! Another 10W – 96W PPT - got me only 50 MHz more for 4.500 GHz. The returns were diminishing rapidly! It really shows how inefficient “Auto-OC” is on it’s own. And that pushing any more was a bit pointless.


PRECISION BOOST OVERRIDE

This is as previously discussed under “Auto-OC” earlier – adding increased single core boosts but with high voltages. You can add this to the PBO setting at your leisure with similar results.


CURVE OPTIMIZER

For me, this setting is the key to all of this process. On my setup, it is only available within BIOS. With Curve Optimizer you can set an offset to the default CPU voltage curves. In effect, you are controlling frequency indirectly, because if you reduce the voltage a little, you create more thermal and power headroom for the cores to boost within - and vice versa. If you insert a negative offset then you are under-volting – a positive one is over-volting.

In practical terms this means a negative offset will allow a higher frequency for about the same voltage and power used (and temperature). Or, for the same frequency (such as a frequency limited single core boost) it will give less voltage and power (and temperature). It’s really quite clever.

You can apply the offset across all cores at the same time, or do it core by core if you know which cores are most capable of handling high frequency with lower voltages. Ryzen Master will show you which cores are your best ones for OC. I didn't bother with that though and just applied an all core offset.

Each unit of offset equates to about 5 mV on my motherboard. So I inserted 10 negative units for around a 50mV reduction. You have to be a bit careful that you don’t insert too much as it affects all phases – and if it gets too low at idle you can suffer crashes. My system behaved perfectly with idle voltages at 0.918 V.

Adding this to the 86 W Precision Boost Settings above gave me an additional 100 MHz – with about the same voltage, power and temperature as previously. More speed for basically no additional thermal or voltage related stress. Hooray!

With the Precision Boost Override – I didn't want those high voltages I saw earlier. So, I elected to find setting that would give me more single core boost, but without increasing the boost voltage beyond the stock scenario. A 75 MHz override, with the curve optimiser keeping the voltage down achieved this. Not much gain, but again no increase in voltage suffered.

FINALLY!

So, after more tweaking, I have settled on PBO with PPT - 88W, EDC - 101A and TDC - 65A, a Curve Optimiser of Negative 15, and a Override of 125 MHz. Pushing harder got too little return to be worth it. It got me from:

Stock – 4.250 GHz all core, 4.650 GHz single core – at 62 degrees, to:

Now – 4.550 GHz all core, 4.775 GHz single core – at 67 degrees.

A 7% gain in multi-core - and a lesser 2.6% gain in single core (which was sort of free!) - all within what I consider to be a perfectly decent voltage range - with just 12 more Watts over stock and very little extra thermal load.

I see the difference in benchmarks (yawn, who cares!) and a noticeable 6% or so improvement in heavy tasks in Lightroom which is actually good for me. No real visible change in day to day use, but that’s kind of the point with PBO and Curves. I only tap into the increased power limits when I need them. Under normal conditions I now sit at 30 degrees at idle, and use the normal 35-40W under single core workloads, so my tweaks don’t really come into play – other than the lower power used from the undervolt. But the extra umph is hiding there for whenever it’s asked for.

Will I keep it? Probably not, but it was fun doing it. Besides, the RAM overclock I did yesterday got me an additional 7% performance anyway, so I don’t need all this stuff anymore!

(EDIT - final tweaks updated)

 

[NoddyPirate]

Mods - I realise now that my post above would have been better placed in the OverClocking forum really. May I have permission to delete the above and move it to there under a new thread? 🤔

 

[NoddyPirate]

Thanks Mod! Whoever you were! 🙏

Final addition to my saga above:

There are a decent number of reports I can find of even 5900X or 5950X users getting stable OC's with PBO Overdrive, Override and Curve Optimiser used altogether - giving increases in both multi and single core boosts (up to 5.1 GHz single core that I've seen - such as Here) - where manual OC couldn't get past 4.7 GHz. If these are actually more than just lucky silicon lottery winners it would be an interesting result!

The biggest issue that arises is heat naturally. A fully OC'd 5950X draws over 200W vs it's default 142W PPT, and those with 280mm AIO's or less have struggled to keep it under 95 degrees.

Would be a fun afternoon tinkering with a 360mm cooler and with someone elses motherboard and CPU though!