News AMD Nixes Support for PCIe 4.0 on Older Socket AM4 Motherboards, Here's Why

[Ncogneto]

The issue is not that isnt does or does not support it. The issue is that AMD stated they would not lock the feature out and leave it up to motherboard manufactures,

They didn't


This is what AMD Said specifically


"Users today may find a PCIe 4.0 option available in their pre-X570 motherboards. However, users should expect this option to be disabled when final retail BIOSes are released to implement full performance and stability for new 3rd Gen Ryzen processors.

As pre-X570 motherboards were not designed with PCIe 4.0 in mind, their designs may be incapable of running PCIe 4.0 signaling with the requisite stability and performance. To ensure a reliable and consistent experience in the field, PCIe 4.0 will not be an option ultimately available to pre-X570 motherboards. Users may continue with a beta BIOS if they desire, but performance and stability cannot be guaranteed."

Reading comprehension, it's your friend. AMD is not locking the CPU's from any older motherboards, if this were the case, a beta bios would not work either.

which I guarantee some of the better X470s at least could support it, yet now they are locking it out completely. Thats the issue.

I agree with their decision but AMD has some culpability to the claims. It is not just the manufactures who made the claims, TH even got a statement from AMD on this 5 months ago.

You can personally guarantee it? Seriously?

 

[jimmysmitty]

They didn't


This is what AMD Said specifically


"Users today may find a PCIe 4.0 option available in their pre-X570 motherboards. However, users should expect this option to be disabled when final retail BIOSes are released to implement full performance and stability for new 3rd Gen Ryzen processors.

As pre-X570 motherboards were not designed with PCIe 4.0 in mind, their designs may be incapable of running PCIe 4.0 signaling with the requisite stability and performance. To ensure a reliable and consistent experience in the field, PCIe 4.0 will not be an option ultimately available to pre-X570 motherboards. Users may continue with a beta BIOS if they desire, but performance and stability cannot be guaranteed."

Reading comprehension, it's your friend. AMD is not locking the CPU's from any older motherboards, if this were the case, a beta bios would not work either.



You can personally guarantee it? Seriously?

Look at the link I posted earlier with the quote from AMD. They stated they would not lock out the feature and leave it up to board manufactures. It was a statement from AMD that TH got directly, not a third party motherboard manufacture.

Again I have not said anything about CPUs or backwards compatibility, just stated the facts which were that AMD stated they would not lock the feature out originally and have now done a 180 and will be locking the feature out. I don't know anyone who would remain on a beta BIOS.

And yes some are built with higher quality materials which would probably handle the increased bandwidth. Some, not all. Power draw is the same as PCIe 3.0 from what I can find.

 

[TJ Hooker]

@jimmysmitty I think they're doing some mental acrobatics and saying that because the beta BIOSs enabling 4.0 are still available that means that AMD hasn't actually disabled the feature (in spite of the fact that it will be disabled in future BIOSs).

 

[cryoburner]

And technically AM4 has supported 2 gens, Zen and Zen 2. I wouldn't consider Zen+ a new gen as it was just a refinement on Zen

By this logic, Intel's recent motherboards haven't even been able to make it through a single generation then, since they've been on 14nm Skylake architecture for the last four years, yet required new motherboards for their Coffee Lake processors.

In addition to moving to an improved fabrication process, Zen+ offered some notable improvements that allow chips to maintain better multi-core boost clocks, as well as improvements to how memory and cache are handled. Those could arguably be considered relatively minor improvements compared to the much larger changes we see with Zen 2, but are probably just as significant if not more so than all the minor revisions Intel has been doing to their processors over the last several years. Intel considers each year's chips to be a new generation, so by their definition, Ryzen 1000, 2000 and 3000-series processors are all distinct generations.

Pentium 4, Pentium D, First generation Core 2 (65nm) and second generation Core 2 (45nm).

What happened to not counting refinements of architectures as a new generation? Pentium D used the same Prescott architecture as the Pentium 4 processors, just with two dies glued together. It was just a dual-core Pentium IV. I'm pretty sure the 45nm Core 2's were just a die-shrink of the existing architecture as well. And we're talking about processors that all launched well over a decade ago, which isn't all that relevant to what Intel has been doing more recently.

Look at the link I posted earlier with the quote from AMD. They stated they would not lock out the feature and leave it up to board manufactures. It was a statement from AMD that TH got directly, not a third party motherboard manufacture.

The thing is, the article makes it sound like AMD is somehow backtracking on some official promise that PCIe 4.0 would be supported on existing motherboards. But it's not like this is something they announced at some official press event, and prior to the option appearing in some BIOS updates for a few motherboards, no one was even expecting the feature to make its way to existing boards.

An AMD representative was simply replying to a request for comment by the writer of the article about how some motherboards where adding a BIOS option to enable the feature. That representative's response makes it sound like the feature would not be officially supported by them, but if the motherboard manufacturers could make it work on certain boards, then they were free to do so. From the sound of it though, the results may not have been consistent enough to make the feature worth enabling on that hardware.

Now, perhaps it could make for an interesting article testing PCIe 4.0 hardware with some of these boards once the new hardware is available to see whether there might have been any notable issues with it. PCIe 4.0 SSDs are probably the only consumer hardware that could show any performance difference, but graphics cards might be worth testing as well.

 

[Ncogneto]

@jimmysmitty I think they're doing some mental acrobatics and saying that because the beta BIOSs enabling 4.0 are still available that means that AMD hasn't actually disabled the feature (in spite of the fact that it will be disabled in future BIOSs).

LOL, AMD doesn't write the Bios's, but keep trying. It is obvious that AMD hasn't "locked out" the feature.

 

[jimmysmitty]

By this logic, Intel's recent motherboards haven't even been able to make it through a single generation then, since they've been on 14nm Skylake architecture for the last four years, yet required new motherboards for their Coffee Lake processors.

In addition to moving to an improved fabrication process, Zen+ offered some notable improvements that allow chips to maintain better multi-core boost clocks, as well as improvements to how memory and cache are handled. Those could arguably be considered relatively minor improvements compared to the much larger changes we see with Zen 2, but are probably just as significant if not more so than all the minor revisions Intel has been doing to their processors over the last several years. Intel considers each year's chips to be a new generation, so by their definition, Ryzen 1000, 2000 and 3000-series processors are all distinct generations.


What happened to not counting refinements of architectures as a new generation? Pentium D used the same Prescott architecture as the Pentium 4 processors, just with two dies glued together. It was just a dual-core Pentium IV. I'm pretty sure the 45nm Core 2's were just a die-shrink of the existing architecture as well. And we're talking about processors that all launched well over a decade ago, which isn't all that relevant to what Intel has been doing more recently.


The thing is, the article makes it sound like AMD is somehow backtracking on some official promise that PCIe 4.0 would be supported on existing motherboards. But it's not like this is something they announced at some official press event, and prior to the option appearing in some BIOS updates for a few motherboards, no one was even expecting the feature to make its way to existing boards.

An AMD representative was simply replying to a request for comment by the writer of the article about how some motherboards where adding a BIOS option to enable the feature. That representative's response makes it sound like the feature would not be officially supported by them, but if the motherboard manufacturers could make it work on certain boards, then they were free to do so. From the sound of it though, the results may not have been consistent enough to make the feature worth enabling on that hardware.

Now, perhaps it could make for an interesting article testing PCIe 4.0 hardware with some of these boards once the new hardware is available to see whether there might have been any notable issues with it. PCIe 4.0 SSDs are probably the only consumer hardware that could show any performance difference, but graphics cards might be worth testing as well.

I never stated their current has. I was happier with their original setup, a new uArch and a die shrink per socket. And you can be on the same process tech forever but have new uArchs. AT some point we may actually be in that situation due to limits of the materials we use for process tech.

Technically there were two P4 and two PDs. Smithfield (Prescott based) and Presler (Cedar Mill based). Prescott to Cedar Mill was like Conroe to Penryn. Smaller process, vastly improved power and temps with performance improvements. Although not as good as Cedar Mill to Conroe, that was more like Bulldozer to Zen.

I would consider Zen+ more like Conroe -> Penryn. Die shrink with major improvements to the entire CPU, including major IPC improvements over the previous generation. Thats a lot more than a refinement. Yes pretty much every CPU since Skylake has been a refinement on Skylake but to claim Intel has never supported more than two generations of CPUs is false.

Either way it was an official statement that AMD was not going to lock out the feature and now they did a 180 and will have the feature locked out in official BIOS. They made a statement and have retracted it in essence. I have NO issue with it as my opinion is that older hardware holds a system back which is why I only buy one CPU per board. I build a new system with parts designed together.

However the fact is that they made the statement and are now reversing the statement. I don't understand why people are arguing it. AMD is a company. Yes they make mistakes and will make claims and 180 on them. They have so before. They probably will again.

 

[Redneck5439]

I think that a lot of people are getting all fired up here about nothing really. People who favor Intel are saying that AMD broke a promise and if it was poor Intel who would have done it people would be jumping all over them, and people who favor AMD are saying there was no broken promise. I think the truth is somewhere in the middle. I think that an AMD engineer gave his opinion but didn't have the authority to speak for the company's official stance and the company was forced to make a final decision on the matter. Such things happen every day, which is why fact checkers will always follow up with someone with the actual authority to speak for the company before running with the story.

Lets get one thing totally straight though, this is not on the level of Intel's recent deceptions. This is not like showcasing a 28 core 5Ghz beast and letting everyone think it was in stock configuration while the actual real world performance is much less. This is not like promising 10nm for years while keeping everyone on 14nm and watching your competition reach 14nm, refine their arch to 12nm, then to 7nm all while promising the same old "10nm is coming and will be dominant". Not to even mention the illegal "cripple AMD" complier they utilized to obtain their lopsided marketshare in the first place. Intel will say and do anything to make their products look better, even if its outright lying to the very fans who fill their pockets with money for "high end next gen processors" that are only >5% IPC improvement on the same process generation after generation. AMD's statement of not locking out PCIe 4.0 support (which can't even be used by any GPUs on the market today) is nothing compared to what Intel has done and stated.

 

[jimmysmitty]

I think that a lot of people are getting all fired up here about nothing really. People who favor Intel are saying that AMD broke a promise and if it was poor Intel who would have done it people would be jumping all over them, and people who favor AMD are saying there was no broken promise. I think the truth is somewhere in the middle. I think that an AMD engineer gave his opinion but didn't have the authority to speak for the company's official stance and the company was forced to make a final decision on the matter. Such things happen every day, which is why fact checkers will always follow up with someone with the actual authority to speak for the company before running with the story.

Lets get one thing totally straight though, this is not on the level of Intel's recent deceptions. This is not like showcasing a 28 core 5Ghz beast and letting everyone think it was in stock configuration while the actual real world performance is much less. This is not like promising 10nm for years while keeping everyone on 14nm and watching your competition reach 14nm, refine their arch to 12nm, then to 7nm all while promising the same old "10nm is coming and will be dominant". Not to even mention the illegal "cripple AMD" complier they utilized to obtain their lopsided marketshare in the first place. Intel will say and do anything to make their products look better, even if its outright lying to the very fans who fill their pockets with money for "high end next gen processors" that are only >5% IPC improvement on the same process generation after generation. AMD's statement of not locking out PCIe 4.0 support (which can't even be used by any GPUs on the market today) is nothing compared to what Intel has done and stated.

There has not been any real comparison to Intel. Each "transgression" is individual per company. I am not saying anything other than the facts and you are assuming it was an AMD engineer speaking out of place. I think it was AMD as a whole as that would have been a nice feature but they jumped the gun and found it has too many possible issues. If it was not an official company statement then they should have made it clear back in January when the idea was floating around instead of waiting.

BTW giving AMD credit for process tech feels out of place sine they haven't developed a process tech since their 45nm. After that it was all spun off in a new company, Global Foundries, for 32nm and others for everything else. ll they do now is design the CPU uArch and use the best available foundry based process tech out there.

 

[Redneck5439]

There has not been any real comparison to Intel. Each "transgression" is individual per company. I am not saying anything other than the facts and you are assuming it was an AMD engineer speaking out of place. I think it was AMD as a whole as that would have been a nice feature but they jumped the gun and found it has too many possible issues. If it was not an official company statement then they should have made it clear back in January when the idea was floating around instead of waiting.

BTW giving AMD credit for process tech feels out of place sine they haven't developed a process tech since their 45nm. After that it was all spun off in a new company, Global Foundries, for 32nm and others for everything else. ll they do now is design the CPU uArch and use the best available foundry based process tech out there.

"There has not been any real comparison to Intel" I believe it was you who said "You, I and everyone here knows if Intel did the same people would rage". You opened the door of comparison to Intel.

Aside from my assumption that it was an AMD engineer, everything else I said was facts. While some may dismiss and blow off Intel's many "transgressions" they do lay down a very good model of just how Intel does business. Basically if you can't beat them cheat them is their business model. In the past it has been their own customers who have had to suffer the consequences of their outright lies (great case in point is the before mentioned "cripple AMD" compiler that would make Intel's processors look better in benchmarks even though AMD's processors of the time were superior).

Now here we are in 2019 with yet another AMD processor generation that is set to dethrone Intel and some are dwelling on a nothing story about not bringing PCIe 4.0 to the previous generations of motherboards. PCIe 4.0 support on the older motherboards wouldn't have included M.2 support so the only thing that 4.0 would have supported would be top slot, which is the graphics card slot. There are 0 graphics cards on the market today with PCIe 4.0 controllers in them, therefore they can't support PCIe 4.0. PCIe 4.0 won't be relevant for GPUs till 2020 or 2021 so its pointless to risk issues to enable it on older motherboards.

As far as AMD "haven't developed a process tech since their 45nm. After that it was all spun off in a new company, Global Foundries, for 32nm and others for everything else. ll they do now is design the CPU uArch and use the best available foundry based process tech out there"- Faulting them for something that is working is just nuts. Their business arrangement with GF has given them an advantage over a much larger company that easily has 10X more R&D money, so it was a very smart thing for AMD to do. Perhaps Intel should give up on developing its own process tech as they clearly are falling behind and can't get 10nm right and switch to a company that can actually get the process tech right. Perhaps if they turned to Global Foundries or another company years ago they wouldn't still be stuck on 14nm. AMD in the same time frame has jumped from 32nm Piledriver to 14nm Zen (in one generation) then to 12nm Zen+ and now to 7nm Zen 2. I'd say whatever AMD is doing is clearly working, clearly winning, and to fault them for that because "they didn't develope the process tech" and just "design the CPU uArch" is just crazy.

"However the fact is that they made the statement and are now reversing the statement. I don't understand why people are arguing it. AMD is a company. Yes they make mistakes and will make claims and 180 on them. They have so before. They probably will again. " This I totally agree with. There is however a difference between making a simple mistake in a statement and doing things like showcasing a processor in a massively overclocked state needing extreme exotic cooling and passing it off as a new processor at stock in an official presentation. I believe it was also Intel who hired a company to benchmark their upcomming i9 9900K agaist the R7 2700X and even though it legitimately outperformed the 2700X they had to take it a leap further and neuter the 2700X with horrible RAM timings and speed and disabling half its threads. You can't compare a simple mistake like AMD just rectified to deliberately misleading/ lying to people like Intel does.

 

[jimmysmitty]

"There has not been any real comparison to Intel" I believe it was you who said "You, I and everyone here knows if Intel did the same people would rage". You opened the door of comparison to Intel.

Aside from my assumption that it was an AMD engineer, everything else I said was facts. While some may dismiss and blow off Intel's many "transgressions" they do lay down a very good model of just how Intel does business. Basically if you can't beat them cheat them is their business model. In the past it has been their own customers who have had to suffer the consequences of their outright lies (great case in point is the before mentioned "cripple AMD" compiler that would make Intel's processors look better in benchmarks even though AMD's processors of the time were superior).

Now here we are in 2019 with yet another AMD processor generation that is set to dethrone Intel and some are dwelling on a nothing story about not bringing PCIe 4.0 to the previous generations of motherboards. PCIe 4.0 support on the older motherboards wouldn't have included M.2 support so the only thing that 4.0 would have supported would be top slot, which is the graphics card slot. There are 0 graphics cards on the market today with PCIe 4.0 controllers in them, therefore they can't support PCIe 4.0. PCIe 4.0 won't be relevant for GPUs till 2020 or 2021 so its pointless to risk issues to enable it on older motherboards.

As far as AMD "haven't developed a process tech since their 45nm. After that it was all spun off in a new company, Global Foundries, for 32nm and others for everything else. ll they do now is design the CPU uArch and use the best available foundry based process tech out there"- Faulting them for something that is working is just nuts. Their business arrangement with GF has given them an advantage over a much larger company that easily has 10X more R&D money, so it was a very smart thing for AMD to do. Perhaps Intel should give up on developing its own process tech as they clearly are falling behind and can't get 10nm right and switch to a company that can actually get the process tech right. Perhaps if they turned to Global Foundries or another company years ago they wouldn't still be stuck on 14nm. AMD in the same time frame has jumped from 32nm Piledriver to 14nm Zen (in one generation) then to 12nm Zen+ and now to 7nm Zen 2. I'd say whatever AMD is doing is clearly working, clearly winning, and to fault them for that because "they didn't develope the process tech" and just "design the CPU uArch" is just crazy.

"However the fact is that they made the statement and are now reversing the statement. I don't understand why people are arguing it. AMD is a company. Yes they make mistakes and will make claims and 180 on them. They have so before. They probably will again. " This I totally agree with. There is however a difference between making a simple mistake in a statement and doing things like showcasing a processor in a massively overclocked state needing extreme exotic cooling and passing it off as a new processor at stock in an official presentation. I believe it was also Intel who hired a company to benchmark their upcomming i9 9900K agaist the R7 2700X and even though it legitimately outperformed the 2700X they had to take it a leap further and neuter the 2700X with horrible RAM timings and speed and disabling half its threads. You can't compare a simple mistake like AMD just rectified to deliberately misleading/ lying to people like Intel does.

What I stated was in no way a comparison. It was an observation after many years watching the industry. I believe that because AMD is a smaller company people do let AMD slide a bit more. I have no need to go into the many "transgressions" AMD has made in their time as they are not relevant nor are any of Intels which is why I wont debate them with you.

The facts are that AMD stated one thing and are doing a 180. Its not a big issue, again as I have said many times I think its better to move people to a newer platform anyways, but it is still a bad move on their part as they should have never made the original statement.

PS. the benchmarking issue? That was actually the companies fault for using AMDs optimizing software without realizing in game mode it would default turn off the SMT. The company also looks to have been amateurs. I wouldn't trust benchmarks from Intel, AMD or nVidia. I rely on TH, Anand etc for that as should any true enthusiast.

 

[Redneck5439]

What I stated was in no way a comparison. It was an observation after many years watching the industry. I believe that because AMD is a smaller company people do let AMD slide a bit more. I have no need to go into the many "transgressions" AMD has made in their time as they are not relevant nor are any of Intels which is why I wont debate them with you.

The facts are that AMD stated one thing and are doing a 180. Its not a big issue, again as I have said many times I think its better to move people to a newer platform anyways, but it is still a bad move on their part as they should have never made the original statement.

PS. the benchmarking issue? That was actually the companies fault for using AMDs optimizing software without realizing in game mode it would default turn off the SMT. The company also looks to have been amateurs. I wouldn't trust benchmarks from Intel, AMD or nVidia. I rely on TH, Anand etc for that as should any true enthusiast.

While I agree that people should be moved to newer platforms I don't believe it should be every generation/ every year forcing people to move to a new motherboard. I like what AMD has done with AM4. You can take a first gen Ryzen motherboard and put a new Zen 2 processor in it and it will work and work well. I like how AMD has managed to improve their motherboards with every generation but also keep socket support. I also like how they will be ending support for AM4 at the end of 2020, presumably with the release Zen 3 and 5nm going to AM5 (or possibly AM4+), ensuring that AM4 isn't going to hold them back.

As far as "the benchmarking issue? That was actually the companies fault for using AMDs optimizing software without realizing in game mode it would default turn off the SMT. The company also looks to have been amateurs" I don't believe for a minute that Intel didn't have their "thumb on the scale". Intel would have known very quickly that the 2700X was gimped because they test their products against AMDs flagship all the time. Intel would have known that the 2700X's scores were way off. Furthermore a professional company that does benchmarking not knowing that the 2700X was gimped is pure crap as my 12 year old nephew knows to set the RAM to its specified speed and timings (which was never done for the 2700X for those benchmarks) in bios - NOT USING SOFTWARE BUT SETTING IT IN BIOS. He also knows that gaming mode was made for the Threadripper line to lessen the impact of that generation's latency issues, it was never intended for use with the 2700X. If he knows that at 12 how can a company with professional college educated employees not know? They were told to make the i9 9900K look as good as humanly possible, and they did it by gimping the 2700X. Intel had to know by looking at the numbers just like Gamer's Nexus called foul within hours of the benchmarks being released. But Intel would never gimp a competitor's processor on purpose, would they? cough "cripple AMD" cough...

"I wouldn't trust benchmarks from Intel, AMD or nVidia. I rely on TH, Anand etc for that as should any true enthusiast"- truer words have never been spoken!! I totally 100% agree. I never trust benchmarks from the manufactures as they always cherry pick what they want released. I personally like Linus, Gamers Nexus and Hardware Unboxed the best, but also rely on other trusted reviewers. There is of course a difference between cherry picking results and paying a company to gimp the competition and release it as official benchmarks. Intel took it a step further and over the top with that one.

 

[cat1092]

This is nothing like when beginning with Ivy Bridge, Intel began using low grade thermal paste, rather than solder, of which AMD continues to use with mainstream CPU's to this date. Why fix what wasn't broken?

Intel left their $10 job per CPU on our backs, and initially never mentioned making the change. Rather, one of these tech sites revealed the issue to a lot of pissed customers. Nor have they made progress with die shrinks since just after the post Haswell era, although they've 'been working' on it since 2015. It's good to see Intel has competition these days, hopefully AMD's pricing will force Intel down further. Which is good for everyone.

As for PCIe 4.0 upgraded by UEFI firmware upgrades, there's lots of boards that would be a fire hazard if implemented. Anyone going to Newegg to look at these specs of these will see there's lots more metal in the PCIe slots, as well as improved heatsinks & cooling of the VRM area. Lastly, some of these will require the traditional 8 pin, plus an extra 4 pin ATX for power delivery. Prior MB's were never marketed as being upgradable to a newer standard when released, so I don't see the big deal. If one wants a 570 MB, these are available in limited quantities for the time being. Even the mATX models are in the $200 range, yet pricing will level off once inventory builds. For that matter, the better Ryzen CPU's are in very short supply, yet even among those in stock (the Ryzen 5 3000 series) near doubles the performance of the i7-4790K & 6700K (Passmark scores). Not bad for a 'budget' 3000 series CPU.

Hopefully in a couple of years, when PCIe 5.0 becomes released (roadmap already in place), those who purchases these PCIe 4.0 MB's won't be looking for a free upgrade via firmware. That's when I'll build new, should include an upgrade to DDR5 (or DDR6) RAM. This is why I'm not jumping on the PCIe 4.0 wagon, will be a short lived release. Just as the (too early released) NVIDIA GTX 2000 series GPU's will be, won't have half of the success the GTX 1000 line enjoyed, now they'll have to scramble to get current.

As for speed, remember, every PCIe release in essence doubled the speed of (most) everything connected. Unfortunately, SATA has hit it's wall, however there's already PCIe 4.0 NVMe SSD's in the works & for those of us still on PCIe 3.0, will mean big discounts across the board.

It's really a lot to expect to make PCIe 3.0 MB's upgradable, on the other hand I get the point. Whomever at AMD stated the possibility should had kept this to themselves & there would be no controversy.

Cat