Thursday, March 17th 2022

AMD's Robert Hallock Confirms Lack of Manual CPU Overclocking for Ryzen 7 5800X3D

In a livestream talking about AMD's mobile CPUs with HotHardware, Robert Hallock shone some light on the rumours about the Ryzen 7 5800X3D lacking manual overclocking. As per earlier rumours, something TechPowerUp! confirmed with our own sources, AMD's Ryzen 7 5800X3D lacks support for manual CPU overclocking and AMD asked its motherboard partners to remove these features in the UEFI. According to the livestream, these CPUs are said to be hard locked, so there's no workaround when it comes to adjusting the CPU multiplier or Voltage, but at least AMD has a good reason for it.

It turns out that the 3D V-Cache is Voltage limited to a maximum of 1.3 to 1.35 Volts, which means that the regular boost Voltage of individual Ryzen CPU cores, which can hit 1.45 to 1.5 Volts, would be too high for the 3D V-Cache to handle. As such, AMD implemented the restrictions for this CPU. However, the Infinity Fabric and memory bus can still be manually overclocked. The lower Voltage boost also helps explain why the Ryzen 7 5800X3D has lower boost clocks, as it's possible that the higher Voltages are needed to hit the higher frequencies.
That said, Robert Hallock made a point of mentioning that overclocking is a priority for AMD and the Ryzen 7 5800X3D is a one off when it comes to these limitations. The reason behind this is that AMD is limited by the manufacturing technology available to the company today, but it wanted to release the technology to consumers now, rather than wait until the next generation of CPUs. In other words, this is not a change in AMD's business model, as future CPUs from AMD will include overclocking.

Hallock also explained why AMD didn't go with more cores for its first 3D V-Cache CPU and it has to do with the fact that most workloads outside of gaming don't reap much of a benefit. This is large due to how different applications use cache memory and when it comes to games, a lot of the data is being reused, which is a perfect scenario for a large cache, whereas something like video editing software, can't take advantage of a large cache in the same way. This means that AMD's secret to boosting the performance in games is that more game data ends up sitting closer to the CPU, which results in a 12 ns latency for the CPU to retrieve that data from the L3 cache, compared to 60-80 ns when the data has to be fetched from RAM. Add to this the higher bandwidth of the cache and it makes sense how the extra cache helps boost the performance in games.

For more details, please see video below. The interesting part starts around the 45:30 mark.

Add your own comment

222 Comments on AMD's Robert Hallock Confirms Lack of Manual CPU Overclocking for Ryzen 7 5800X3D

#26
DeathtoGnomes
Closer to the very end, last 5 minutes or so, they tried to pry info from Robert H. on the plans of AM5s longevity. Although nothing is concrete yet, my magic 8-ball™ says there might be something more to that discussion.
TomgangLet's just hope that is not the case for Zen 4 as well. Can you imagine all Zen 4 cpu being locked for OC?
if thats the case with zen4, it will be incremental updates instead of a full upgrade. But, always a but, will AMD divide into the big cache for gaming and continue normalcy, per se, for the rest of the chip line up.
Posted on Reply
#27
bug
So, this Frankenstein's creation isn't sewed together very well, is it?
Posted on Reply
#28
Ruru
S.T.A.R.S.
MelvisIts the Xtreme!!! Gaming Processor ;)
Well, for Intel the X actually standed for Extreme :D
TheLostSwede20th of April apparently.
Ended up ordering a 5800X, as they dropped to $15 more than the MSRP of the 5700X here, pretty much over night.
I guess that they were high as hell when they priced this and it's launching 20th of April..
Posted on Reply
#29
SL2
DeathtoGnomesI was kind of thinking to not wait and buy that 5800x3d, almost. I want to upgrade now, its so hard not to go nuts now, but I think the wait for next gen and am5 will be well worth that wait.
AM4 will stay for quite some time according to Su. At the same time, I don't expect AMD to launch AM5 CPU's with the kind of high prices that Vermeer had, now that the competition is real.
I do expect AM5 boards to cost more, tho.
Posted on Reply
#30
DeathtoGnomes
MaenadFINI guess that they were high as hell when they priced this and it's launching 20th of April..
better than April 1st. :D:cool:
Posted on Reply
#31
Valantar
So it's pretty safe to assume a CPU architecture more fundamentally designed with this feature in mind (rather than having connectors for it added because it was in development and needed a test vehicle) will have a separate cache voltage rail, right? This is probably non-trivial given how closely tied cache is to the cores, but that would seem like the logical way forward.


As for this news more broadly: if it only means there's no manual, multiplier-based OC, but PBO and CO are still there ... who cares? All-core fixed-multiplier OC on Ryzen is a pretty bad idea unless you're consistently running 100% load nT workloads anyhow. There's no reason why anyone should use that over PBO and CO tuning, unless what you're going for is worse performance and higher power consumption.
Posted on Reply
#32
Tomgang
DeathtoGnomesCloser to the very end, last 5 minutes or so, they tried to pry info from Robert H. on the plans of AM5s longevity. Although nothing is concrete yet, my magic 8-ball™ says there might be something more to that discussion.


if thats the case with zen4, it will be incremental updates instead of a full upgrade. But, always a but, will AMD divide into the big cache for gaming and continue normalcy, per se, for the rest of the chip line up.
It's difficult to say. Maybe 5800X3d is some sort of a test drive. How many addopt it compared to 5800X. There may be Zen 4 with and with out 3d cashe as well.

But no oc on Zen 4 will be a huge disappointment for me.
Posted on Reply
#33
SL2
MaenadFINI guess that they were high as hell when they priced this and it's launching 20th of April..
Yeah, but it is indeed a higher chip(s) inside that CPU.

"Our highest chip launches on 4/20"

Makes sense.
Posted on Reply
#34
Valantar
TomgangIt's difficult to say. Maybe 5800X3d is some sort of a test drive.
It clearly is - it's the first product to hit the market with a brand-new technology; it's relatively limited in scope (one SKU, the last product for a five-year-old platform, etc.) and they have announced no plans for further models for this platform with the feature. Definitely a test drive.
TomgangHow many addopt it compared to 5800X. There may be Zen 4 with and with out 3d cashe as well.
Almost definitely. It doesn't make sense on all SKUs, but a wider roll-out on a chip more thoroughly adapted to this (with a separate cache voltage rail, for example) would make a lot of sense. Something like every tier from x6xx or x8xx upwards having a 3D cache-enabled top-end SKU would make sense (i.e. 7600 65W, 7600X3D 105W, 7800 65W, 7800X3D 105W, etc.). This would make a lot of sense if they can make the cache die on 7nm even when the CCDs move to 5nm, as that would free up capacity to churn out more cache dice.
TomgangBut no oc on Zen 4 will be a huge disappointment for me.
Just to be clear: no multiplier-based, fixed frequency and voltage OC is not "no OC". PBO and Curve Optimizer are still ways of overclocking, and they seem to be supported here. They also deliver better results in general on Zen2 and Zen3 than old-school OC techniques.
Posted on Reply
#36
Chaitanya
Last time I overclocked a CPU was a AM2 CPU and havent touched OC either on my 4770k or 3700x.
Posted on Reply
#37
Chomiq
MatsYeah, but it is indeed a higher chip(s) inside that CPU.

"Our highest chip launches on 4/20"

Makes sense.
"It smokes... the competition".
Posted on Reply
#38
Cutechri
Who needs manual OC anymore on Zen 2 and above. Just PBO and Curve Optimizer if on Zen 3. Or do as I do and completely ignore overclocking because... why bother.
Posted on Reply
#39
Bwaze
So comparing new 3D cache Ryzen at fixed frequency (4 GHz) in AMD presentation and stating 15% uplift in gaming was quite dishonest, since new 5800X3D will not achieve 5800X frequency!

Posted on Reply
#40
TheLostSwede
News Editor
BwazeSo comparing new 3D cache Ryzen at fixed frequency (4 GHz) in AMD presentation and stating 15% uplift in gaming was quite dishonest, since new 5800X3D will not achieve 5800X frequency!

What's dishonest with that? It was a technology demo of a CPU they never launched.
Posted on Reply
#41
SL2
BwazeSo comparing new 3D cache Ryzen at fixed frequency (4 GHz) in AMD presentation and stating 15% uplift in gaming was quite dishonest, since new 5800X3D will not achieve 5800X frequency!
That's a 12 (6+6) core, which gives it more cache per core than an 8 core. Do you know what a prototype is?
Posted on Reply
#42
Bomby569
I'm missing something here, every CPU ever launched has a limit he shouldn't be pushed beyond.
If we can't even try to OC is because it's already at that limit. If 1.35 is the limit is should be at something more moderate and we could use sylicon lottery to try and push it a bit further like any CPU before it.

So i assume this is just a CPU really pushed to it's limits just to compete with Intel.
Posted on Reply
#43
mama
bugSo, this Frankenstein's creation isn't sewed together very well, is it?
We will see...
Posted on Reply
#44
Valantar
Bomby569I'm missing something here, every CPU ever launched has a limit he shouldn't be pushed beyond.
If we can't even try to OC is because it's already at that limit. If 1.35 is the limit is should be at something more moderate and we could use sylicon lottery to try and push it a bit further like any CPU before it.

So i assume this is just a CPU really pushed to it's limits just to compete with Intel.
It seems to be a bit more complex than that thanks to the specifics of the cache die. The CPU cores themselves likely handle higher voltages and frequencies just fine, but if the cache die is on the same voltage rail as the cores and risks damage above 1.35V, then obviously the cores can't be allowed to go that high either. And seeing how cache and logic have quite different characteristics in silicon, especially when using different libraries (the cache die is ~2x the density of the cache on the CCD, after all), so it likely just isn't designed to scale in the same way.
BwazeSo comparing new 3D cache Ryzen at fixed frequency (4 GHz) in AMD presentation and stating 15% uplift in gaming was quite dishonest, since new 5800X3D will not achieve 5800X frequency!

That's a prototype of an unreleased product. I doubt AMD would launch the 5800X3D unless it was faster than the 5800X - if it didn't make sense, they would make a lot more money selling those stacked dice in Epyc chips where HPC/server vendors would gobble them up at much higher prices. But ultimately, we'll have to see how this pans out in real-world reviews.
Posted on Reply
#45
mb194dc
Just lock the voltage but allow the multiplier to be changed? Can't be that complicated...
Posted on Reply
#46
SL2
Bomby569If 1.35 is the limit is should be at something more moderate and we could use sylicon lottery to try and push it a bit further like any CPU before it.
How are you going to do that when manual OC is disabled?
Bomby569So i assume this is just a CPU really pushed to it's limits just to compete with Intel.
No.
Even if Alder Lake wasn't released yet, AMD would want the X3D to be as close to the 5800X clock frequencies as possible, otherwise it wouldn't make much sense in launching it.
AFAIK, the 3D cache is primarily made for Milan-X.
Posted on Reply
#47
qubit
Overclocked quantum bit
The technical reason for this limitation is interesting, but as a customer, I just it to overclock, so I'd pass on this model.
Posted on Reply
#48
illusion archives
DemonicRyzen666So why does it have an X in its name then?
it's not Ryzen 5800X 3D
it's a ryzen 5800 3D
Maybe it's Ryzen 5800 X3D?
Posted on Reply
#49
SL2
qubitThe technical reason for this limitation is interesting, but as a customer, I just it to overclock, so I'd pass on this model.
The limitations may just as well be the very reason why we don't see more X3D models. Well that and the fact that most X3D's goes to Milan-X.
Posted on Reply
#50
tabascosauz
ValantarIt clearly is - it's the first product to hit the market with a brand-new technology; it's relatively limited in scope (one SKU, the last product for a five-year-old platform, etc.) and they have announced no plans for further models for this platform with the feature. Definitely a test drive.

Almost definitely. It doesn't make sense on all SKUs, but a wider roll-out on a chip more thoroughly adapted to this (with a separate cache voltage rail, for example) would make a lot of sense. Something like every tier from x6xx or x8xx upwards having a 3D cache-enabled top-end SKU would make sense (i.e. 7600 65W, 7600X3D 105W, 7800 65W, 7800X3D 105W, etc.). This would make a lot of sense if they can make the cache die on 7nm even when the CCDs move to 5nm, as that would free up capacity to churn out more cache dice.

Just to be clear: no multiplier-based, fixed frequency and voltage OC is not "no OC". PBO and Curve Optimizer are still ways of overclocking, and they seem to be supported here. They also deliver better results in general on Zen2 and Zen3 than old-school OC techniques.
Haven't finished watching the video yet, but did he confirm PBO is actually around? Generally even Auto PBO quickly enables higher Vcore peaks (1.5V+) not possible at stock. Yes CO is fine, but seeing as the current AGESA completely removed Boost Override (only to be later reintroduced), sounds like they were prepping their firmware for the 5800X3D by removing the whole shebang. Without Override there's not much point to CO since all Vermeer chips hit their stock global limits easily

But knowing AGESA they have an abysmal record of keeping features consistent (especially vendors like GB that release and pull a billion beta BIOSes weekly). So if the V-cache really is that sensitive to voltage, then that sounds like a recipe for disaster..........but I wonder if it's that the cache really can't handle the voltage, or that heat density is too much under certain heavy loads?

Think Intel have had separate cache clock and voltage for like a decade now - obviously a bit more involved with V-cache but I'd fully expect this whole deal to be resolved on AM5. 5800X3D is just an experiment searching for guinea pigs

Currently on normal L3, cache freq is very closely correlated to core effective clock at all times, so curious to see if they have or will eventually decouple cache clock
Posted on Reply
Add your own comment
Nov 21st, 2024 12:29 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts