News AMD: Ryzen 3000-Series CPUs Lack Manual Overclocking Headroom

jimmysmitty

Polypheme
Moderator
Thats all fine and dandy but they really take the fun out of being a PC enthusiast. Overclocking is one of the best ways to get a bit of extra performance and its fun to learn how to do it.

When I got my Q6600 and tweaked it to 3GHz using less than the stock voltage it was great. Ran cool and performed well for years.

I guess this means no reason to buy any of the really high end AMD boards.
 
Reactions: DMAN999 and splave

ingtar33

Illustrious
this is a bit disappointing, but we knew that Ryzen's base design was never a high GHZ design. debauer proved back when Ryzen1 came out that the design is HIGHLY efficient at low clock speeds (per him, the sweet spot was right around 2.4-3.3ghz), this was how AMD could release an 8 core 16 thread cpu overclocked to 4ghz and have it use about the same power at the socket as an intel i7 with 4 cores 8 threads overclocked to 5ghz. very efficient cores that just don't like overclocking.

It reminds me a bit of their PhenomII cpus, those disliked overclocking immensely. Well, when the node matures I'm sure the clock speeds will go up a bit, that's what usually happens. furthermore it is looking like AMD's auto clocking tools are getting more and more effective.
 
Maybe it's just me, but I don't think it's all that big of a deal.

Overclocking has generally been hit or miss. Now, the CPU overclocks itself, but is paying attention to what it can and can't do, and, no warranty gets voided. No trial and error needed. No wondering if you've won or gone bust on the silicon lottery.

I always found it baffling when people were looking to overclock a chip the moment they got it. At least, in recent years. Back in the day, yeah, I could see the desire for it. When you overclocked a Pentium 133 successfully to 166, that was a 25% jump in CPU speed. And likely, your CPU was officially a 133 because Intel wanted specific numbers of chips at each tier.

These days it's what? Maybe a 5% increase without exotic measures for cooling? And these days, the GPU is more likely to be holding you back, rather than the CPU, in a lot of cases where gaming is involved.
 

TJ Hooker

Illustrious
Herald
Thats all fine and dandy but they really take the fun out of being a PC enthusiast. Overclocking is one of the best ways to get a bit of extra performance and its fun to learn how to do it.

When I got my Q6600 and tweaked it to 3GHz using less than the stock voltage it was great. Ran cool and performed well for years.
To be honest, I think this is ultimately a move in the right direction for the consumer. Rather than playing the silicon lottery and potentially spending extra for OC-enabled components so that you can get some unknown level of performance increase, you can instead buy a CPU that provides its best right out of the box with little variability between one chip to the next. It's been years since you could really get major performance gains out of CPU overclocking anyway.

I agree that it's not great if you find the acting of overclocking enjoyable in and of itself though.
 

jimmysmitty

Polypheme
Moderator
To be honest, I think this is ultimately a move in the right direction for the consumer. Rather than playing the silicon lottery and potentially spending extra for OC-enabled components so that you can get some unknown level of performance increase, you can instead buy a CPU that provides its best right out of the box with little variability between one chip to the next. It's been years since you could really get major performance gains out of CPU overclocking anyway.

I agree that it's not great if you find the acting of overclocking enjoyable in and of itself though.
Yes and no. For the majority of consumers Intels idea is perfect. Regular CPUs with no OCing. But for enthusiasts, whcih AMD is marketing with a lot of this, they like having those features.

It also takes away from their competitive aspect if performance is similar to the competition but the competition can just overclock and perform better.

You're ragging on TTH for a statement that AMD made?

Typical redgarl...
He is never a fan of anything somewhat negative about AMD even if its true or and official statement from AMD.
 

TJ Hooker

Illustrious
Herald
It also takes away from their competitive aspect if performance is similar to the competition but the competition can just overclock and perform better.
Oh, I agree that if two CPUs have the same stock performance but one can overclock and the other can't then the OC-able one is better. But let's say you have a CPU that a performance level X, and can typically be OC'd 5-10%. Now let's say we have a different CPU that has X+7.5% performance at stock, but virtually no OC-ability. What I was trying to say is that I think the latter would be a superior product, even for enthusiasts, as the average performance is the same but without the variability (and added cost) of overclocking. Again, just from a practical perspective, putting aside any enjoyment you get out of the tweaking process itself.

So I don't think AMD's lack of OC-ability is a problem in itself. It's only a drawback because Intel's otherwise similarly performing CPUs are able to eke into the lead because they can be OC'd.

He is never a fan of anything somewhat negative about AMD even if its true or and official statement from AMD.
Lol, yeah I've noticed.
 
Reactions: bit_user

TCA_ChinChin

Reputable
Feb 15, 2015
279
58
4,940
21
Thats all fine and dandy but they really take the fun out of being a PC enthusiast. Overclocking is one of the best ways to get a bit of extra performance and its fun to learn how to do it.

When I got my Q6600 and tweaked it to 3GHz using less than the stock voltage it was great. Ran cool and performed well for years.

I guess this means no reason to buy any of the really high end AMD boards.
I was about to start roasting you until I actually read your comment and noticed you said AMD boards instead of processors lol. I disagree with your later comment on how Intel's idea is better for non-enthusiasts though. AMD's solution provides the best balance between consumers and a large portion of enthusiasts. Their method of PBO + auto-overclock will allow for maximum performance for everyone except overclockers (and other enthusiasts who overclock for various reasons). Everyone else won't be able to get crazy 20% clockspeed improvements from overclocking but its still there to tinker with. Not as exciting, but there. I would disagree about Intel having a competitive advantage once overclocked, Ryzen 3000 still takes the cake in most applications.
 

jimmysmitty

Polypheme
Moderator
I was about to start roasting you until I actually read your comment and noticed you said AMD boards instead of processors lol. I disagree with your later comment on how Intel's idea is better for non-enthusiasts though. AMD's solution provides the best balance between consumers and a large portion of enthusiasts. Their method of PBO + auto-overclock will allow for maximum performance for everyone except overclockers (and other enthusiasts who overclock for various reasons). Everyone else won't be able to get crazy 20% clockspeed improvements from overclocking but its still there to tinker with. Not as exciting, but there. I would disagree about Intel having a competitive advantage once overclocked, Ryzen 3000 still takes the cake in most applications.
For the majority of consumers overclocking never happens. Boost clocks are all they need. The smaller market of enthusiasts it is fine and the even smaller market of more hardcore enthusiasts it is not a good thing.

As for performance, so far on apples to apples (not counting price) comparison (8 core to 8 core) Intel has a slight, very, advantage and the overclocking gives them a much larger advantage. If you go price to price then yes in more heavily multi threaded applications AMD has the advantage.
 
It also takes away from their competitive aspect if performance is similar to the competition but the competition can just overclock and perform better.
Being able to get marginally better performance for double the price does not seem particularly competitive to me. You're not getting anything for free either, since Intel's charging an extra $50+ for the ability to overclock a processor, plus extra for an overclocking-capable chipset, in addition to them not even supplying a cooler for those parts. AMD offering i7-like performance for $200, and i9-like performance for $330, along with reasonably capable coolers for all of their processors, definitely seems far more competitive no matter how one looks at it.

Perhaps when Intel slashes prices once again with their 10-series processors, offering similar thread counts as AMD at comparable price points, they may once again be competitive, so long as one is willing to put up with higher power draw and heat output for that bit of extra performance, along with what will probably be inadequate stock coolers on the locked parts, and no coolers on the unlocked parts.

As for performance, so far on apples to apples (not counting price) comparison (8 core to 8 core) Intel has a slight, very, advantage and the overclocking gives them a much larger advantage. If you go price to price then yes in more heavily multi threaded applications AMD has the advantage.
The performance differences from overclocking these processors are not all that large. A stock 9900K can boost up to 5.0GHz on a single core, down to 4.7GHz with all cores loaded. According to SiliconLottery, only 35% of 9900Ks were able to push a 5.0GHz or higher all-core overclock under their test criteria and 13% were not even able to hit 4.9GHz, but assuming one is willing to use higher voltages and expensive cooling, we'll use 5.0GHz for our comparison. A 5.0GHz all-core overclock should result in roughly the same performance as stock for lightly-threaded workloads, up to around 6% higher performance at most in heavily threaded workloads utilizing all cores. And that's assuming perfect performance scaling, which usually isn't the case, especially in things like games, where the processor will be spending much of its time waiting for the graphics card either way.

And as far as heavily-multithreaded workloads go, you don't even have to go price to price. In many applications, a 3700X will outperform a 9900K. The two processors trade blows at application performance in general, performing similar overall. And this is for a new processor architecture that hasn't yet been optimized for. Unless the primary purpose of a system is to run one or more of those applications where the 9900K happens to pull slightly ahead, there is little reason to pay more for that processor.

And in terms of gaming performance, the Intel processors might still manage to edge out the competition in benchmarks, but only when pairing them with a very high-end graphics card at low-resolutions that few people would use them with. At practical resolutions for a given card, there will be little to no difference, and significantly more performance could be had by putting that money toward faster graphics hardware instead.
 

bit_user

Splendid
Herald
I say: kudos to AMD for automating overclocking! I don't see how this can possibly be viewed as a negative.

The real culprit for the lack of variability is that PBO is combined with binning. To that end, I think we will see some variability in the boost speeds of the highest-binned dies, but remember that AMD is holding those for the 3950X, and possibly the next ThreadRipper and maybe some Epyc SKUs.

Anyway, the real question for builders is how much cooling you need, before PBO plateaus. It would be nice to know, so you don't waste money or create a machine that's louder or more of a dust trap than it needs to be.
 

mitch074

Distinguished
Mar 17, 2006
2,099
55
19,890
23
I say: kudos to AMD for automating overclocking! I don't see how this can possibly be viewed as a negative.

The real culprit for the lack of variability is that PBO is combined with binning. To that end, I think we will see some variability in the boost speeds of the highest-binned dies, but remember that AMD is holding those for the 3950X, and possibly the next ThreadRipper and maybe some Epyc SKUs.

Anyway, the real question for builders is how much cooling you need, before PBO plateaus. It would be nice to know, so you don't waste money or create a machine that's louder or more of a dust trap than it needs to be.
It's a bit of a change in paradigm : up until now, overclocking was a closed garden for DIYers that wanted as much bang for buck as they could get (I sure was one of them); chip makers would try to discourage overclocking through multiplier or frequency locking (Intel first among them), but since it was never really practical, Intel decided to make you pay more for overclocking - meaning that those that got a 'K' edition CPU with a 'Z' chipset were the privileged fews.

AMD never really tried hard to prevent overclocking - just enough to prevent sleazy deals with low-range chips being sold as high range ones. Now, with auto overclock, you really get what you pay for - but even back in the K8/K10 days, squeezing an extra 800 MHz out of an entry level CPU on a cheap motherboard was merely a matter of pushing the reference clock as high as the system could handle - often as much as 25%, if not 50% - those Sempron64 1600 MHz pushed as far up as 2400 MHz were sweet deals. The $100 Athlon II X4 620 was quite nice too : +800 MHz on (big) air with only a little extra voltage, heck I'm still running it @3200 MHz, no overvolt, on a puny $20 cooler...

But then, of course, the 5 years of Bulldozer killed it : you were either a classy overclocker looking for maximum performance on a high price Intel platform, or you were hitting 5 GHz on your cubicle heater trying to reach the same performance as the competition's stock... Tuned supercar VS. ricer.

But now, the "ricer" car swapped the old carburetor-fed V8 for a fusion plant-powered electric engine : max torque and max power right out of the box - while the tuned supercar is having overheating problems at high speeds.
 
Jul 12, 2019
1
0
10
0
I remember the time when Overclocking was not an "out of the box" feature. Great times.
Nowdays, overclocking is something that it is taken for granted and they simply expect it to work. Back to work guys, OC should not be just a slider ^-
 

jimmysmitty

Polypheme
Moderator
Um... redgarl -> red garl -> red girl?

Then, there's the avatar pic that's clearly the back of the red-clad redhead, Asuka Langly.

You could be right although the majority of users who use Anime avatars of females typically are males.

I say: kudos to AMD for automating overclocking! I don't see how this can possibly be viewed as a negative.

The real culprit for the lack of variability is that PBO is combined with binning. To that end, I think we will see some variability in the boost speeds of the highest-binned dies, but remember that AMD is holding those for the 3950X, and possibly the next ThreadRipper and maybe some Epyc SKUs.

Anyway, the real question for builders is how much cooling you need, before PBO plateaus. It would be nice to know, so you don't waste money or create a machine that's louder or more of a dust trap than it needs to be.
Automated OCing has never been as efficient as manual. I doubt it would be now either. One of the biggest issues I always saw was the voltage itself. Automated OCing always pushes the voltage higher than I was able to manually set it for and be stable.

Plus it takes away the feeling of accomplishment when you got a stable overclock.

And I wouldn't say AMD was the first at it. Asus and other motherboard OEMs have had those features in BIOS and software for quite a while. AMD is just the first CPU OEM to push it.
 

TJ Hooker

Illustrious
Herald
Automated OCing has never been as efficient as manual. I doubt it would be now either. One of the biggest issues I always saw was the voltage itself. Automated OCing always pushes the voltage higher than I was able to manually set it for and be stable.

And I wouldn't say AMD was the first at it. Asus and other motherboard OEMs have had those features in BIOS and software for quite a while. AMD is just the first CPU OEM to push it.
AFAIK the automatic OC tools implemented my mobo OEMs were pre-determined profiles programmed into the FW. So it didn't take your individual CPUs capabilities into account, and therefore needed to err on the side of over volting so that they'd work with the majority of CPUs.

AMD's auto-OC/boost features take into account real-time data that's specific to your CPU when determining behavior. So it'd likely be more efficient (which I think is generally borne out in the power consumption numbers for Ryzen CPUs), and fundamentally different than mobo OEM auto-OCing.
 

splave

Contributing Writer
Editor
Jan 4, 2019
8
4
15
0
Thats troublesome if they must squeeze this hard on their brand new line of processors without leaving any headroom frequency wise. Hopefully they can quickly refine the process for some more clocks.
 

jimmysmitty

Polypheme
Moderator
AFAIK the automatic OC tools implemented my mobo OEMs were pre-determined profiles programmed into the FW. So it didn't take your individual CPUs capabilities into account, and therefore needed to err on the side of over volting so that they'd work with the majority of CPUs.

AMD's auto-OC/boost features take into account real-time data that's specific to your CPU when determining behavior. So it'd likely be more efficient (which I think is generally borne out in the power consumption numbers for Ryzen CPUs), and fundamentally different than mobo OEM auto-OCing.
Not quite. Asus has a built in one that sets clocks and voltages and then tests and will continue to do so until it hits what it feels is stable. At least the newer ones do.

I am not knocking the software. I just think the most efficient way to overclock is with manual. What would be nice is if they would give us the ability to alter the boost settings so we could overclock the all core boost but keep the single core boost or even alter it to higher speeds if it can handle it. That would be ideal.

Thats troublesome if they must squeeze this hard on their brand new line of processors without leaving any headroom frequency wise. Hopefully they can quickly refine the process for some more clocks.
AMD has no input into the process. It is all up to TSMC and since TSMC has many other customers to work on they wont focus only on AMDs design. They do have a 7nm+ coming (they call it 6nm) but thats due out for AMD by next year and that wont promise any more speed or headroom. It could give better efficiency but we have a year or so before we even get to see that.
 

TCA_ChinChin

Reputable
Feb 15, 2015
279
58
4,940
21
@jimmysmitty . AMD still have manual overclocking. They just also happen to include auto-overclocking as an extra. Its now a bad thing, and Intel has started to play catch up with their Performance Maximizer. Tom's themselves have reviewed the software. Intel's solution isn't even that viable as even using their own software for auto-overclocking will void warranty. PBO + auto-overclock in AMD's case is within processor spec, and thus is covered by warranty.
 

bit_user

Splendid
Herald
You could be right although the majority of users who use Anime avatars of females typically are males.
Whatever someone's avatar, if they call themselves a girl er... "garl", I'm not going to argue. In this case, the avatar only serves to help explain the name.

However, if a poster with the same avatar were named Ji... er... let's see... RickyBobby, then clearly they want to be regarded as a "he".
 

bit_user

Splendid
Herald
AMD has no input into the process. It is all up to TSMC and since TSMC has many other customers to work on they wont focus only on AMDs design.
@splave was talking about "process" improvements - not specific to AMD's designs - which are quite likely to continue.

Fabs are always trying to improve yields, for one thing, which can have some tangential benefits on efficiency. A process node has a long life, beyond the point where it's surpassed by something better. There are still chips getting fabbed at 40 nm and bigger!
 

jimmysmitty

Polypheme
Moderator
@jimmysmitty . AMD still have manual overclocking. They just also happen to include auto-overclocking as an extra. Its now a bad thing, and Intel has started to play catch up with their Performance Maximizer. Tom's themselves have reviewed the software. Intel's solution isn't even that viable as even using their own software for auto-overclocking will void warranty. PBO + auto-overclock in AMD's case is within processor spec, and thus is covered by warranty.
I never said they did not have manual overclocking. However when you manually overclock you typically kill the boost levels, check THs review where they got it to X clock speed but single threaded suffered since it was lower than the default boost rate yet higher than PBO+ would go for all core.
 

ASK THE COMMUNITY

TRENDING THREADS