News AMD Confirms Ryzen 7 5800X3D Is Not Overclockable

saunupe1911

Distinguished
Apr 17, 2016
203
74
18,660
So would PBO automatically be disabled in the BIOS when this chip is installed?

Does this chip require a special BIOS from MB manufacturers?

If not then what's stopping someone from enabling PBO unless the chip just wouldn't take in the IO instructions.

Oh nevermind...MB manufactures are rolling out AGESA 1.2.0.6b in their BIOS to support this.
 
  • Like
Reactions: salgado18

setx

Distinguished
Dec 10, 2014
224
149
18,760
If overvoltages can easily kill new chips and they can't block only voltage control – it might be reasonable to disable overclocking.

Motherboards just love to rise random voltages when they think you are "overclocking" if you didn't set pretty much all of them to fixed values. Better be safe than gain bad publicity from fried chips.

Overall just one chip and even locked is quite sad for getting to know new technology.
 
  • Like
Reactions: Roland Of Gilead
This sounds more and more like an experimental chip. If it does work well, we may see the tech in the next generation. If not, it will remain in history (and a few people's pcs).
This makes a lot of sense to me. AMD's probably having a lull with their Zen 4 production and they really want to put this new technology in the field and seeing what works and whatnot.

Similar to when NVIDIA made the GTX 750.

No can do, the increase in rma would kill any profit from it.
Well, using PBO voids the warranty.
 
  • Like
Reactions: Roland Of Gilead
This sounds more and more like an experimental chip. If it does work well, we may see the tech in the next generation. If not, it will remain in history (and a few people's pcs).
With AM5 and Ryzen 6000(?) set to launch and of 2022, the window for changing architectural things now is little/none.

I would agree that this smells like a guinea pig from AMD. Its likely/possible that it's a limitation of strapping Vcache to an architecture that wasn't explicitly designed for it. Whereas, Ryzen 6000 could have been designed around these voltage pitfalls (time will tell). But heck, if they can slap Vcache on previous-gen architecture to bide themselves a good stopgap "flagship" against Intel for the majority of 2022...why not!? Keep in mind, the 5800X3D probably won't draw much/any more power than the ~100W that the 5800X draws. Compare that to the 150-200W of the 12700K/12900K that it's competing against. Also, if the 5800X3D can beat or match the 12900KS while consuming nearly half the power, it will be a resounding win for AMD and an utter embarrassment for Intel.

The 5800X3D still has a 4.5GHz boost frequency (5800X @ 4.7GHz). Which, based on their claim of 1.35V maximum and my own testing with my 5600G of 4.6GHz @ 1.44V single core and 4.0GHz @ 1.2V all core, seems respectable. I'd also hope/assume that AMD just caps voltage at 1.35V where someone could technically achieve a 4.5GHz all-core frequency. How much more does the average user want/need? Most reviews' overclocking results of Ryzen 5000 conclude with the recommendation that PBO and undervolting is at/near the best you can achieve without extreme cooling.
 
Last edited:
  • Like
Reactions: punkncat

hannibal

Distinguished
For real v-cache cpus we have to wait Zen5 or Zen4+ or similar. This needs separate voltage controller for the cache and that can not be done afterwards. So any zen4 with same technology would suffer from same problem. But it can be done eventually, by making different voltage regulation to this part of the chip. So Zen4+ or Zen5 what ever comes after Zen4 could use vcache just nicely if it is sensible otherwise. Maybe adding more cache to the chip itself makes vcache obsolete?
 
  • Like
Reactions: tennis2

PiranhaTech

Reputable
Mar 20, 2021
134
86
4,660
This sounds more and more like an experimental chip. If it does work well, we may see the tech in the next generation. If not, it will remain in history (and a few people's pcs).
I was thinking something along these lines. This might be a version 1 part, so they might not want to push things. It sounds like they went from doing this to the 5900X to doing it to the 5800X, and then cutting the clock. There must have been good reasons.
 

wifiburger

Distinguished
Feb 21, 2016
613
106
19,190
my problem with this is that it's not just 200mhz off the top frequency vs 5800x

5800x is cheaper & easily gets to 4.9-5.0 with PBO

so you're looking at 500mhz freq difference; too big of a gap for games that don't care about cache size

There's no way 1.3v or 1.35v being the limit. Changing LLC for example could get this thing to overshoot and kill it during AVX .

What happened here;my best guess is that they are using unused Epyc dies that are binned for 1.3v and not Ryzen dies that are binned for 1.5v
And as such they are blocking PBO, OC out of fear of RMAs and nothing to do with cache or extra filler on top of the cores.
 
Last edited:
my problem with this is that it's not just 200mhz off the top frequency vs 5800x

5800x is cheaper & easily gets to 4.9-5.0 with PBO

so you're looking at 500mhz freq difference; too big of a gap for games that don't care about cache size

There's no way 1.3v or 1.35v being the limit. Changing LLC for example could get this thing to overshoot and kill it during AVX .

What happened here;my best guess is that they are using unused Epyc dies that are binned for 1.3v and not Ryzen dies that are binned for 1.5v
And as such they are blocking PBO, OC out of fear of RMAs and nothing to do with cache or extra filler on top of the cores.
What if the v-cache makes up for the ~500mhz loss and lower voltage Wouldn't that make it a more efficient CPU?

Only time will tell
 
Then they should have called it the 58003D, NOT X.
The X has nothing to do with overclocking for AMD's processors, and even the non-X Ryzen CPUs have had overclocking unlocked. It's largely just there on some models to indicate a higher-end version of a processor at a given core count. And in the case of X3D, that stands for "Hybrid 3D" at least according to AMD when they announced it over two years ago.
 
500 MHz less when compared with 5800X top clock speed with PBO, that's 10%. Humongous cache should improve IPC by 5 to 20 %, and that increases with how much more of a contention RAM throughput can be (hint : the higher the CPU clock, the higher the contention). Meaning that the 5800X3D is likely at worst as powerful as the 5800X, and at best 10% faster.
 

rluker5

Distinguished
Jun 23, 2014
605
367
19,260
With AM5 and Ryzen 6000(?) set to launch and of 2022, the window for changing architectural things now is little/none.

I would agree that this smells like a guinea pig from AMD. Its likely/possible that it's a limitation of strapping Vcache to an architecture that wasn't explicitly designed for it. Whereas, Ryzen 6000 could have been designed around these voltage pitfalls (time will tell). But heck, if they can slap Vcache on previous-gen architecture to bide themselves a good stopgap "flagship" against Intel for the majority of 2022...why not!? Keep in mind, the 5800X3D probably won't draw much/any more power than the ~100W that the 5800X draws. Compare that to the 150-200W of the 12700K/12900K that it's competing against. Also, if the 5800X3D can beat or match the 12900KS while consuming nearly half the power, it will be a resounding win for AMD and an utter embarrassment for Intel.
In gaming the 12700k/12900k don't use the same watts as when running p95 or cinebench. They are on par with AMD's efficiency when gaming and the 12700k exceeds it in watts per frame per Igor's testing.

The 5800X3D is advertised as having the best gains in gaming so comparisons in power draw should be with that. It should also have less cache misses and may be able to move a couple more things out of memory into cache. This will reduce cpu cycles wasted by waiting for ram and reduce power wasted by unneeded speculative prediction operations. It doesn't increase ipc at all, but it can look like it. And it will likely be more efficient than Alder in games but not faster when Alder is overclocked.
 
As long as the performance is right why does overclocking matter apart from the ‘fun’ aspect?
Talking from personal experience going from a 2700X to a 3800XT and having a 5600X as an HTPC system, I can tell you why: when you're running a Ryzen CPU that is not thermally constrained PBO increases performance by a lot, plus whatever the motherboard considers "safe" in therms of OC headroom. Again, this is when keeping temperatures in check.

What AMD is saying here is simple to me: the cache will put a constraint in power delivery (locked/capped SoC-CPU voltage), no matter the thermal headroom of the CPU. So if you've already invested in a good cooling solution, be it AIO, a beefy HSF or whatever, it'll mean the CPU will run very cool, but won't give you that "step up" in performance based on the money you have already invested/spent on the cooling solution. Sure, it may be really efficient and run cool, but what good is that to me when I can't translate it into more performance?

As people has already mentioned, ALL Ryzen CPUs have a lot of performance headroom you don't even need to fiddle much with to extract, so the already small extra gap this CPU brings will be even smaller using a properly cooled (and in turn, self-OC'ed) 5800X or better.

The only saving grace of this CPU will be HTPC? Something where thermal constraints are more important than whatever extra headroom you can get with a beefy cooling solution? That's hardly an enticing proposition for me at least. And I sincerely wonder* how many here can honestly say, with a straight face, this CPU is what they were really expecting. I can certainly say it is not what I expected.

Regards.
 
Talking from personal experience going from a 2700X to a 3800XT and having a 5600X as an HTPC system, I can tell you why: when you're running a Ryzen CPU that is not thermally constrained PBO increases performance by a lot, plus whatever the motherboard considers "safe" in therms of OC headroom. Again, this is when keeping temperatures in check.

What AMD is saying here is simple to me: the cache will put a constraint in power delivery (locked/capped SoC-CPU voltage), no matter the thermal headroom of the CPU. So if you've already invested in a good cooling solution, be it AIO, a beefy HSF or whatever, it'll mean the CPU will run very cool, but won't give you that "step up" in performance based on the money you have already invested/spent on the cooling solution. Sure, it may be really efficient and run cool, but what good is that to me when I can't translate it into more performance?

As people has already mentioned, ALL Ryzen CPUs have a lot of performance headroom you don't even need to fiddle much with to extract, so the already small extra gap this CPU brings will be even smaller using a properly cooled (and in turn, self-OC'ed) 5800X or better.

The only saving grace of this CPU will be HTPC? Something where thermal constraints are more important than whatever extra headroom you can get with a beefy cooling solution? That's hardly an enticing proposition for me at least. And I sincerely wonder* how many here can honestly say, with a straight face, this CPU is what they were really expecting. I can certainly say it is not what I expected.

Regards.
Sorry but that doesn’t answer it. It doesn’t matter what any other cpu does. If I am paying X and I know I will get Y performance I can make a decision based on that, overclocking does not change this.

I have a 3700X and I use PBO. However if I can get a 5800X3D and benchmarks show it has an average 30% performance improvement in my use case then that’s all I need to know. As long as I know what I am getting beforehand and can compare to what I have now I really don’t see the problem.