AMD asks mobo makers to remove OC options for 5800X3D

That's why I wonder why amd didn't just make a new io die with more cache or put it off to the side or something...
One reason for stacking is that it puts the cache physically closer to the cores. This is important because at the stupidly high clock speeds that modern CPUs run at, even the small distances across the die become significant; at 5GHz a signal travelling at the speed of light can only go 6cm in one clock cycle, and electrical signals travel quite a lot slower than that through silicon.
 
Looks like a major fail to me - disabling OC functionality and then refusing to explain why.
 
Looks like a major fail to me - disabling OC functionality and then refusing to explain why.
They did explain it and it’s not like stacked cache is a new technology, just new to AMD. It’s benefits and drawbacks are well known and as it currently stands it’s fragile, your voltage and frequency ranges are very narrow. And the extra layers of heat producing silicon becomes hard to manage with out increasing surface size.

 
Last edited:
This is why I just don’t see this thing being a better choice over the 5900x given the pricing/availability. That or save the money and get a normal 5800x while you can or a 5700x.
 
and for anyone that's not looking to sit through a whole episode of hot hardware the 5800X3D stuff starts at ~45:00

and one of the reasons why i can't get into that show is they go from complaining that they don't make a 5950X3D to complaining about them not making desktop versions of whatever apu's!!? c'mon ..really? even though they say that they would buy multiples of each i seriously doubt that. i'm with amd on this one. i'm sure they are to the point of seeing through bullsh.. too
 
They did explain it and it’s not like stacked cache is a new technology, just new to AMD. It’s benefits and drawbacks are well known and as it currently stands it’s fragile, your voltage and frequency ranges are very narrow. And the extra layers of heat producing silicon becomes hard to manage with out increasing surface size.


I think my issue is why brand is a 5800X3D instead of a 58003D. The X variants have always implied unlocked and fully overclockable, so this naming to me feels very disingenuous. Imagine Intel releasing a K variant you couldn't overclock; people on this board would be rioting.
 
That's why I wonder why amd didn't just make a new io die with more cache or put it off to the side or something.
Shame they killed off tr that massive heat spreader would have been a great thing for this type of stuff
Because AMD, just like Intel and Nvidia have to start working on stacked technology and eventually you have to be able to mass produce it just like AMD is doing here. Nothing allows you to learn and grow faster than an actually shipping product.
 
They did explain it and it’s not like stacked cache is a new technology, just new to AMD. It’s benefits and drawbacks are well known and as it currently stands it’s fragile, your voltage and frequency ranges are very narrow. And the extra layers of heat producing silicon becomes hard to manage with out increasing surface size.


Stacked cache / memory is NOT new to AMD. Stacking that memory with other components with different performance characteristics is a universal problem. There's a reason why Foveros is limited to laptops and there's like 1 SKU (see Lakefield).
 
I think my issue is why brand is a 5800X3D instead of a 58003D. The X variants have always implied unlocked and fully overclockable, so this naming to me feels very disingenuous. Imagine Intel releasing a K variant you couldn't overclock; people on this board would be rioting.
Well memory OC will still be available just nothing that modifies voltage or clock speeds on the chips. It’s just going to be more subtle, but yes I agree the X monicker here is confusing but it is the most expensive letter to print so it signifies quality.
 
Stacked cache / memory is NOT new to AMD. Stacking that memory with other components with different performance characteristics is a universal problem. There's a reason why Foveros is limited to laptops and there's like 1 SKU (see Lakefield).
Has AMD previously released a stacked cache product or any product using Through Silicon Vias? Because AMD’s own videos say “this is the first time we’ve tried something like this” and have said they could either “wait for the technology to mature more before releasing it or release it now to give customers the performance and improve it over time.”

IBM, Intel, Samsung, Apple, Qualcomm, they all have products using various forms of 3D stacking and lots of patents related to them but I can’t find any from AMD using this tech set in their lineups, CPU or GPU.
 
Has AMD previously released a stacked cache product or any product using Through Silicon Vias? Because AMD’s own videos say “this is the first time we’ve tried something like this” and have said they could either “wait for the technology to mature more before releasing it or release it now to give customers the performance and improve it over time.”

IBM, Intel, Samsung, Apple, Qualcomm, they all have products using various forms of 3D stacking and lots of patents related to them but I can’t find any from AMD using this tech set in their lineups, CPU or GPU.
You might want to look up HBM memory. AMD has quite a few patents around stacked technology. This is their first CPU w/ stacked memory. Not their first 3D stacked product.
 
You might want to look up HBM memory. AMD has quite a few patents around stacked technology. This is their first CPU w/ stacked memory. Not their first 3D stacked product.
AMD does not manufacture HBM nor do they own the patents on it, using somebody else’s pre manufactured parts and attaching them onto a chip is very different than designing your own. AMD’s stacked cache is their first attempt at a design in house directly on a processor. Also dealing with stacked silicon on something with a surface area like an EPYC or a GPU is very different than fitting it into a relatively small package like a consumer CPU.

AMD is not shy in saying that this is their first time using this technology on their chips and their inexperience with it is their straight up excuse for why they are removing OC from the chips. They say haven’t figured out how to manage the strict voltage and frequency requirements their implementation because they are inexperienced with it, and once they have worked out these problems they say they will be happy to bring back those OC options.

TSMC’s side of the stacked cache.

“TSMC held their Technology Symposia the same week as Computex, and one of the announcements was that their chip-on-wafer (CoW) version of SoIC™ (System on Integrated Chips) will be qualified for N7-on-N7 this year, with production slated for next year. Given that both the CCD and the V-cache are 7-nm, the new Ryzen 5000 sounds like the first product to use this technology.”

https://www.semiconductor-digest.co...firms-current-ryzen-5950x-is-tsv-capable/amp/
 
Last edited:
AMD does not manufacture HBM nor do they own the patents on it, using somebody else’s pre manufactured parts and attaching them onto a chip is very different than designing your own. AMD’s stacked cache is their first attempt at a design in house directly on a processor. Also dealing with stacked silicon on something with a surface area like an EPYC or a GPU is very different than fitting it into a relatively small package like a consumer CPU.

AMD is not shy in saying that this is their first time using this technology on their chips and their inexperience with it is their straight up excuse for why they are removing OC from the chips. They say haven’t figured out how to manage the strict voltage and frequency requirements their implementation because they are inexperienced with it, and once they have worked out these problems they say they will be happy to bring back those OC options.
Dude it was co-developed with Hynix AMD DOES have patents on HBM. It's an open standard but AMD actually has patents on it and quite few other 3D stacked technologies. Saying that they don't manufacture (fab it) doesn't prevent one from designing the technology.
You don't remember the articles saying that AMD would charge royalties?

"Advanced Micro Devices owns a number of patents covering HBM, but as that intellectual property is a part of JEDEC’s JESD235 standard, it has to be licensed to applicants desiring to implement the standard “either without compensation or under reasonable terms and conditions that are free of any unfair discrimination.” Moreover, AMD and Nvidia have a broad cross-licensing agreement, which largely prevents royalty demands."

https://wccftech.com/amd-squashes-rumors-hbm-ip-licensing-fees-memory-standard-free/.

And again first to stack memory on top of a CPU? Yes. First stacked product? No.

Here's literally an interview with AMD on how it was created.

"Making this sort of innovation happen was a broadly collaborative effort. AMD did much of the initial the heavy lifting, designing the interconnects, interposer, and the new DRAM type. Hynix partnered with AMD to produce the DRAM, and UMC manufactured the first interposers. JEDEC, the standards body charged with blessing new memory types, gave HBM the industry’s blessing, which means this memory type should be widely supported by various interested firms. HBM made its way onto Nvidia’s GPU roadmap some time ago, although it’s essentially a generation behind AMD’s first implementation."

https://techreport.com/review/28294/amds-high-bandwidth-memory-explained/
 
Last edited:
Dude it was co-developed with Hynix AMD DOES have patents on HBM. It's an open standard but AMD actually has patents on it and quite few other 3D stacked technologies. Saying that they don't manufacture (fab it) doesn't prevent one from designing the technology.
You don't remember the articles saying that AMD would charge royalties?

"Advanced Micro Devices owns a number of patents covering HBM, but as that intellectual property is a part of JEDEC’s JESD235 standard, it has to be licensed to applicants desiring to implement the standard “either without compensation or under reasonable terms and conditions that are free of any unfair discrimination.” Moreover, AMD and Nvidia have a broad cross-licensing agreement, which largely prevents royalty demands."

https://wccftech.com/amd-squashes-rumors-hbm-ip-licensing-fees-memory-standard-free/.

And again first to stack memory on top of a CPU? Yes. First stacked product? No.

Here's literally an interview with AMD on how it was created.

"Making this sort of innovation happen was a broadly collaborative effort. AMD did much of the initial the heavy lifting, designing the interconnects, interposer, and the new DRAM type. Hynix partnered with AMD to produce the DRAM, and UMC manufactured the first interposers. JEDEC, the standards body charged with blessing new memory types, gave HBM the industry’s blessing, which means this memory type should be widely supported by various interested firms. HBM made its way onto Nvidia’s GPU roadmap some time ago, although it’s essentially a generation behind AMD’s first implementation."

https://techreport.com/review/28294/amds-high-bandwidth-memory-explained/
Dude AMD and TSMC both say this is the first time they have tried using this technology on their chips. HBM uses a different stacking technology.

HBM uses microbumps, AMD is using a copper to copper bond. The copper to copper bonding allows them to improve density almost 15x over the micro bump technology used by HBM and other similarly stacked chips. They are not remotely comparable.
 
Dude AMD and TSMC both say this is the first time they have tried using this technology on their chips. HBM uses a different stacking technology.

HBM uses microbumps, AMD is using a copper to copper bond. The copper to copper bonding allows them to improve density almost 15x over the micro bump technology used by HBM and other similarly stacked chips. They are not remotely comparable.
Again CPU plus stacked cache is new for them. Don't know how many times it needs to be said but there is a big difference between stacking memory together off die and putting it onto the same package as the CPU. That's what AMD is talking about. Who said the technology was identical? Not me.

First you said they didn't have any patents on HBM which is provably wrong. Now you're saying they aren't identical. Who said HBM stacked technology and what AMD is doing here is identical? Not me. This doesn't change the fact that the memory is stacked in HBM and AMD did a lot of the work on that.

The overall fact is AMD has a number of patents covering 3D stacking. Pretending like this is the first attempt at 3D stacking is just wrong. Factually it's not historically accurate at all.

The main rub I have with this is that this technology is not easy to pull off. It's hard for everyone. Not just AMD. If it was so easy Intel would be mass producing Foveros. They are not.
 
Last edited:
Again CPU plus stacked cache is new for them. Don't know how many times it needs to be said but there is a big difference between stacking memory together off die and putting it onto the same package as the CPU. That's what AMD is talking about. Who said the technology was identical? Not me.

First you said they didn't have any patents on HBM which is provably wrong. Now you're saying they aren't identical. Who said HBM stacked technology and what AMD is doing here is identical? Not me. This doesn't change the fact that the memory is stacked in HBM and AMD did a lot of the work on that.

The overall fact is AMD has a number of patents covering 3D stacking. Pretending like this is the first attempt at 3D stacking is just wrong. Factually it's not historically accurate at all.

The main rub I have with this is that this technology is not easy to pull off. It's hard for everyone. Not just AMD. If it was so easy Intel would be mass producing Foveros. They are not.
I’m not the one saying it’s their first time doing this, AMD and TSMC are.

AMD’s patents for HBM are related to memory interfaces nothing that I can find retaining to manufacturing. Microbump tech is pretty old though so I may not have searched far enough back but I do consider patents to interfacing unrelated to manufacturing.

I’m not sure though but I think we’re arguing the same thing…
 
I think my issue is why brand is a 5800X3D instead of a 58003D. The X variants have always implied unlocked and fully overclockable, so this naming to me feels very disingenuous. Imagine Intel releasing a K variant you couldn't overclock; people on this board would be rioting.
uhh...

Non-x parts have all the oc features enabled. They are just slower

Ah, that's right, the X is for XFR (extended freq. range) -- They just boost a bit higher.

https://pcper.com/2017/03/psa-amd-xfr-enabled-on-all-ryzen-cpus-x-skus-have-wider-range/
 
I’m not the one saying it’s their first time doing this, AMD and TSMC are.

AMD’s patents for HBM are related to memory interfaces nothing that I can find retaining to manufacturing. Microbump tech is pretty old though so I may not have searched far enough back but I do consider patents to interfacing unrelated to manufacturing.

I’m not sure though but I think we’re arguing the same thing…
Did you not say this?
IBM, Intel, Samsung, Apple, Qualcomm, they all have products using various forms of 3D stacking and lots of patents related to them but I can’t find any from AMD using this tech set in their lineups, CPU or GPU.
Also I provided a link from AMD itself going into HBM. They did far more than just the memory controller. I hope you don't think that just because a company is fabless that means they don't have manufacturing patents / knowledge in manufacturing because Rambus has tons of those.
 
Did you not say this?

Also I provided a link from AMD itself going into HBM. They did far more than just the memory controller. I hope you don't think that just because a company is fabless that means they don't have fab manufacturing patents because Rambus has tons of those.
Best I can tell from your link is that AMD developed the interposer which sits as an intermediary packaging layer connecting their CPU die to the HBM logic die. So that’s part of the packaging process not the silicon manufacturing process. But if you want to include HBM as a stacked chip tech in there as previous 3D stacking tech then so be it.
 
Everyone here knows what it stands for, but the average user does not and equates it to akin of a k series intel chip that is unlocked and overclock-able. You wouldn't believe just how many simpletons out there don't even know the difference between memory speeds let alone the difference between a 5800 vs a 5800X. So again, their naming nomenclature is dubious on this chip.
 
Everyone here knows what it stands for, but the average user does not and equates it to akin of a k series intel chip that is unlocked and overclock-able. You wouldn't believe just how many simpletons out there don't even know the difference between memory speeds let alone the difference between a 5800 vs a 5800X. So again, their naming nomenclature is dubious on this chip.
I agree, but saying it implies something that it doesn't, doesn't help dispell that misconception. AMD has said what it means, it's up to us to not keep spreading misinformation.

All current AMD CPUs are overclockable and unlocked (with the exception of this one, now). The motherboard you use may not expose these features, but the CPUs support them. X or no.
 
So my question is, Why lock it?

Like, Ryzen chips keep getting less and less OC Headroom with each generation, 5000 series OC headroom is non-existent. Just enable PBO and keep it cool and it will do a better job than a human ever will. If you get greedy and fry your new chip, so what? That's Overclocking. That's what happens when you fly too close to the sun.

Ryzen processors won't do much beyond their rated boost clocks, in fact, they will likely run SLOWER with manual tuning.

Overclocking has ALWAYS carried a risk to damaging your hardware. It can and DOES happen.

So why lock this one? What was there to gain?
 
So my question is, Why lock it?

Like, Ryzen chips keep getting less and less OC Headroom with each generation, 5000 series OC headroom is non-existent. Just enable PBO and keep it cool and it will do a better job than a human ever will. If you get greedy and fry your new chip, so what? That's Overclocking. That's what happens when you fly too close to the sun.

Ryzen processors won't do much beyond their rated boost clocks, in fact, they will likely run SLOWER with manual tuning.

Overclocking has ALWAYS carried a risk to damaging your hardware. It can and DOES happen.

So why lock this one? What was there to gain?
They are leaving fabric and memory clocking enabled, they are just disabling anything that touches voltage and frequency of the CPU. But I mean you usually get better performance gains tweaking those over frequency anyways.

From what they have stated the 3D stacked chips have a voltage range of 1.30 to 1.35v and anything beyond that causes catastrophic damage and the chips already use up to that max. If that is true then there really isn’t anything left there to squeeze out of it frequency wise.

If they don’t disable the options especially the UEFI auto overclocks then that would be negligent, I can only imagine how many would do it only to burn up the chip then simply RMA them stating the CPU was DOA.
 
They are leaving fabric and memory clocking enabled, they are just disabling anything that touches voltage and frequency of the CPU. But I mean you usually get better performance gains tweaking those over frequency anyways.

From what they have stated the 3D stacked chips have a voltage range of 1.30 to 1.35v and anything beyond that causes catastrophic damage and the chips already use up to that max. If that is true then there really isn’t anything left there to squeeze out of it frequency wise.

If they don’t disable the options especially the UEFI auto overclocks then that would be negligent, I can only imagine how many would do it only to burn up the chip then simply RMA them stating the CPU was DOA.
Maybe you are right, but maybe you don't.

If we can make 5800x3D to boost to 5GHz it will be very close to the next-gen in gaming. You can be sure that AMD doesn't like that.

What happens to 5600x ? They lock the boost to up to 200MHz and no more. Why ? Is there an official statement about that (I never met)?
The only reason I can see is that 5600x became too fast and buying an expensive chip become pointless.

Do you see any similarities between both things? ;)
 
Maybe you are right, but maybe you don't.

If we can make 5800x3D to boost to 5GHz it will be very close to the next-gen in gaming. You can be sure that AMD doesn't like that.

What happens to 5600x ? They lock the boost to up to 200MHz and no more. Why ? Is there an official statement about that (I never met)?
The only reason I can see is that 5600x became too fast and buying an expensive chip become pointless.

Do you see any similarities between both things? ;)
I see what your trying to say but I think your thinking about it too much. AMD has just pushed TSMC’s 7nm process as far as it’s going to go. TSMC released a lot of info about the 3D stacking independently of AMD and their design guidelines for using the technology and it is pretty clear on the limitations and benefits.

I get people think the good old days when there was still OC room on chips was a good thing. But really that signified a lack of competition in the market, as it currently stands neither Intel nor AMD have the wiggle room to leave performance behind so they are pushing the silicon as far as it goes from the box. This is a good thing, yeah it takes away from the hobby a little bit it means that there’s an actual fight between them to make sure each part performs as well as it can right from the box.

If we get back to a stage where one of them feels they can leave some performance on the shelf on a product then that means the other doesn’t have anything they feel is pushing them and they can be lazy about their launches and I don’t think any of us want that.
 
I think my issue is why brand is a 5800X3D instead of a 58003D. The X variants have always implied unlocked and fully overclockable, so this naming to me feels very disingenuous. Imagine Intel releasing a K variant you couldn't overclock; people on this board would be rioting.
Everyone here knows what it stands for, but the average user does not and equates it to akin of a k series intel chip that is unlocked and overclock-able. You wouldn't believe just how many simpletons out there don't even know the difference between memory speeds let alone the difference between a 5800 vs a 5800X. So again, their naming nomenclature is dubious on this chip.
Which is it,
"I don't understand this naming convention" or"
"Everyone understands this naming convention"

Not dubious at all, I have no sympathy for people who can't take the time to read and understand simple things. Ask questions when unsure, etc.
 
Which is it,
"I don't understand this naming convention" or"
"Everyone understands this naming convention"

Not dubious at all, I have no sympathy for people who can't take the time to read and understand simple things. Ask questions when unsure, etc.
The problem here is this X edition part partially supports Overclocking, Fabric, and Memory are good to go it's only frequency and voltage they are locking out so it still has more OC options than a non-X part but fewer than previous X parts so do they instead give it a V moniker since half the options are still there?

5800x3d
5800v3d ...

I think I like the V3D better, they missed an opportunity for an epic pun here and I am mad at them for it.
 
The problem here is this X edition part partially supports Overclocking, Fabric, and Memory are good to go it's only frequency and voltage they are locking out so it still has more OC options than a non-X part but fewer than previous X parts so do they instead give it a V moniker since half the options are still there?

5800x3d
5800v3d ...

I think I like the V3D better, they missed an opportunity for an epic pun here and I am mad at them for it.
non-x parts are fully overclockable. I'm somewhat shocked that so many here don't know that. non-X parts are just as overclockable as X parts. This new 3d cache part is less overclockable than every other AMD cpu.

I did however miss the pun, huge miss for AMD as well.
 
Last edited:
non-x parts are fully overclockable. I'm somewhat shocked that so many here don't know that. non-X parts are just as overclockable as X parts.
Huh, I also assumed people just knew that. It's one of the big pros of buying AMD, always able to overclock. No need to pay for the extra letter just to do it. Intel marketing has been bad for the brain, apparently.
 
Huh, I also assumed people just knew that. It's one of the big pros of buying AMD, always able to overclock. No need to pay for the extra letter just to do it. Intel marketing has been bad for the brain, apparently.
Its a pro until you realize twiddling the settings makes your system slower and more unstable.... Not like the old days, toggle bios switch, gain 100% performance... (celly 300A @ 500mhz)
 
Its a pro until you realize twiddling the settings makes your system slower and more unstable.... Not like the old days, toggle bios switch, gain 100% performance... (celly 300A @ 500mhz)
I mean pbo is still a form of overclocking that isn't available to non K Intels if I remember right.
 
  • Like
Reactions: travm
like this
The truth is that we are expecting AMD to have consistent naming in their products.

I've given up on this.

AMD has come a long way, and become a real powerhouse of innovation again... But they still suck at keeping their SKU naming concise
 
The truth is that we are expecting AMD to have consistent naming in their products.

I've given up on this.

AMD has come a long way, and become a real powerhouse of innovation again... But they still suck at keeping their SKU naming concise
faster cpu = X... how is this hard to follow? their naming system hasn't had anything to do with overclocking since?? black editions?
 
Back
Top