Rumor: AMD to go LGA with AM5, PCIe Gen 5 limited to only EPYC CPUs for Zen 4 Update: More leaks

[Vishera]

11 hours ago, Lurick said:

New architectures need new BIOS and other support stuff and AMD doesn't want to land themselves into the beta BIOS support crap where boards have to remove some CPU support to support new stuff and all the other crap that went on with the Ryzen 5000 CPUs on older boards.

That's not accurate,the whole thing is up to board partners to decide,not AMD.

User choice is a good thing,and so far the the beta BIOS support only applies to first gen 300 series boards.

On my B450 board i can install a 5600X without a beta BIOS,and i can install a Ryzen 7 1700 as well.

So far it wasn't bad,choice is good but AMD just wants you to buy their newest chipsets.

 

DDR5 is a valid reason for a new socket,the BIOS stuff is not a good reason - because board partners end up dealing with it and not AMD themselves.

[Brooksie359]

34 minutes ago, Vishera said:

That's not accurate,the whole thing is up to board partners to decide,not AMD.

User choice is a good thing,and so far the the beta BIOS support only applies to first gen 300 series boards.

On my B450 board i can install a 5600X without a beta BIOS,and i can install a Ryzen 7 1700 as well.

So far it wasn't bad,choice is good but AMD just wants you to buy their newest chipsets.

 

DDR5 is a valid reason for a new socket,the BIOS stuff is not a good reason - because board partners end up dealing with it and not AMD themselves.

That is the case for your motherboard and not for all motherboards. Alot of motherboards don't work like that simply because the bios memory is so small. Trying to figure out what motherboards would work like yours vs ones that wouldn't would be to hard for most consumers to the point where its not worth it. You also have to realize that it isn't a simple engineering problem to continually use the same socket while changing the architecture drastically and adding more cores. At some point they might need the extra pins and am4 couldn't provide that. Its honestly good enough that they had 4 gens of cpus on one socket. 

[Doobeedoo]

Well, not too surprised moving away from PGA with a new socket, especially since it was expected there would be more pins as well. On the other hand PCIe gen 5.0 not initially for AM5 is an interesting move, if it actually is true. Probably a move for platform launch, to keep the cost lower and manufacturing easier as well as that there are no new hardware to even take advantage of it. We'll see.

[Dabombinable]

10 minutes ago, Doobeedoo said:

Well, not too surprised moving away from PGA with a new socket, especially since it was expected there would be more pins as well. On the other hand PCIe gen 5.0 not initially for AM5 is an interesting move, if it actually is true. Probably a move for platform launch, to keep the cost lower and manufacturing easier as well as that there are no new hardware to even take advantage of it. We'll see.

When you look at how few consumer orientated devices use gen 4.0, you only end up with unecessary expenses. That being said, if the chipset link was gen 5.0, that'd be great. You could have far more m.2 SSD hanging off it.

[GreatnessRD]

13 hours ago, akio123008 said:

So putting all logical arguments aside for a second,

 

Can I just say, I really don't like LGA. AMD please don't.

I think I'm with you here. LGA is fine, but I like the PGA method.

[Chiyawa]

Huh?

 

I actually like PGA better, but with the pin count, I can see why they decided to go LGA. Still, I'm PGA all the way, I think PGA gives AMD its charm even when they are not famous.

[da na]

I mean back in the day you could swap AMD and Intel CPUs freely as AMD pretty much directly copied everything Intel did... For example, you could swap the Intel 8088 and an AMD D8088. Same socket. I don't know which socket was the last to do this, maybe socket 7? (You could put literally any company's CPU in Socket 7 and it'd still run,,,)

[da na]

Also, I think that LGA is a bit easier to screw up by an inexperienced user. I don't have experience bending pens back on a motherboard but it isn't that hard to do on a CPU, I would assume it would be a good bit harder inside the socket. But both have their advantages and disadvantages.

[WhitetailAni]

13 minutes ago, wat3rmelon_man2 said:

Also, I think that LGA is a bit easier to screw up by an inexperienced user. I don't have experience bending pens back on a motherboard but it isn't that hard to do on a CPU, I would assume it would be a good bit harder inside the socket. But both have their advantages and disadvantages.

It's MUCH harder to fix an LGA socket.

I tried to fix the socket on my Optiplex 3020 MT board (now deceased due to a failed pin bend) and I got 2 of the 6 bent ones.

On the third, the screwdriver went too far down and scraped one off.

 

On the other hand with PGA (even an AM4 CPU which has dense pins) I was able to fix all the bent ones on my Ryzen 5 1600X.

[cj09beira]

15 hours ago, 8-Bit Ninja said:

I've kept mine in anti static bags and dropped them countless times and have yet to kill one 

i kept one on my pocket for a week and it was fine . (it was a P4)

11 hours ago, Brooksie359 said:

I don't really see why pcie 5.0 is necessary at this point. Would it be good to have? Yeah so long as it doesn't cost to much but they can also wait a couple of generations of motherboard before adding if as well. 4.0 Is already pretty fast and should be more than enough for awhile to come for consumers. 

thing about pcie 5 is not pcie 5, its the possibility of CXL, gen-z etc, which run on the pcie electrical spec but with their own protocols, and will allow for much better flexibility.

 

i would have preferred to see pga with a retention mechanism, but i guess we cant have everything

[staticpage]

RIP finding good working mobos at the scrapyard. 100% of all the intel LGA mobos that would work great are destroyed because LGA is fragile and the socket cover never gets put back on. I find piles of good Z170s and have taken the time to repair a few but they are mostly all destroyed pins, so much E-waste. I find an A320 in the gaylords and works every time.

[leadeater]

58 minutes ago, cj09beira said:

thing about pcie 5 is not pcie 5, its the possibility of CXL, gen-z etc, which run on the pcie electrical spec but with their own protocols, and will allow for much better flexibility.

You won't find CXL implemented on any consumer devices though. Save maybe Apple doing something funky with memory tiering. CXL like a lot of things is still optional and you'll find it behind the likes of Quadro and Tesla paywalls for years to come.

[StDragon]

27 minutes ago, leadeater said:

You won't find CXL implemented on any consumer devices though.

Doesn't CXL have to be implemented in the CPU anyways for support?

[AndreiArgeanu]

17 hours ago, Juular said:

What's the point of changing the socket if not for PCIe 5.0 then ? I mean, 3 generations of the forward compatibility is still very good but now this just looks like an attempt to force motherboard change artificially.

Edit: ah, i forgot about DDR5, maybe that.

PCIE 5.0 is overkill for consumer hardware and will be for a while. You don't need PCIE 5.0. Also AMD can't support the same socket forever, there's already been 4 generation of cpu's on the same socket, far more than what Intel has ever done, and it's already muddy when it comes to compatibility between cpu's and motherboards, I think 4 generations is more than enough for a single socket, and it's fair for AMD to move on to a different socket so they don't have to deal with all the backwards compatbility bs.

[AndreiArgeanu]

17 hours ago, Random_Person1234 said:

PCIe Gen 5 will be limited to only EPYC Genoa CPUs on Zen 4. Intel is rumored to have PCIe Gen 5 on all Alder Lake chips.

It makes sense to be honest, I find it difficult to believe that 99.9% of consumers actually need PCIE 5.0.

 

17 hours ago, Random_Person1234 said:

AM5 will have an LGA1718 socket

this sounds like good news, the CPU is less likely to break since all the fragile pins are on the motherboard and it will be far less likely for people to just rip their cpus out of the socket when they remove their coolers, assuming they use a similar locking mechanism to Intel.

 

Though with Intel being LGA 1700 and AMD LGA1718, plus both will be on 600 series chip sets I wouldn't be surprised if a few more people than usual will end up buying the wrong motherboard. Unless the sockets will have a noticeable difference in size.

[porina]

4 hours ago, wat3rmelon_man2 said:

I mean back in the day you could swap AMD and Intel CPUs freely as AMD pretty much directly copied everything Intel did... For example, you could swap the Intel 8088 and an AMD D8088. Same socket. I don't know which socket was the last to do this, maybe socket 7? (You could put literally any company's CPU in Socket 7 and it'd still run,,,)

I know for sure it was a thing in 486 land. AMD offered the same CPUs but at higher clocks. I also know it wasn't a thing for Pentium 2, where we had Slot 1 vs. Slot A (and some slocket fun). See what they did there? The generation in between, I don't recall clearly any more. I did have a Pentium MMX 120 MHz but don't recall that AMD was ever an option for my build.

[ravenshrike]

K... TBH, if you need the storage capabilities of PCIe5.0 over 4.0 you can almost invariably afford the EPYC line of processors and EF also makes no mention of TR. If you have to step up to TR in order to get your PCIe 5.0 fix a gen early that's what, a 400-600 dollar cost difference at most core for core?

[sof006]

Interesting, i've always preferred LGA to PGA

[leadeater]

10 hours ago, StDragon said:

Doesn't CXL have to be implemented in the CPU anyways for support?

Has to be implemented at all links in the chain. Intel's next generation Xeons are already confirmed to have CXL.

[Beerzerker]

I still prefer PGA vs LGA.
Over the years of using both I've had alot less problems from PGA than LGA. Also, unless they start to enlarge the chips and socket themselves, the more pins you've got in the same sized socket, the thinner these will get and that also means weaker, meaning more easily damaged or perhaps even burned at times.

Older LGA sockets like 775 for example were fine, they hold up rather well but the newer ones like 1151 as an example are more prone to have issues such as the usual bent/broken pins and another thing I've rarely seen mentioned, thats a pin(s) shifting to one side instead of just being bent or broken. Never saw that until I got my Z77 OCF and right out of the box a pin was shifted slightly to one side making it throw a code 55. It was only when I did some careful checking that I found it. Easily fixed and did that to get it going but the annoyance is it still does this with the same pin sometimes. It's like the pin isn't fitting tight into the socket, able to wiggle around a bit and I've got another board with the same problem (Maximus IX Apex), had to fix that the other day for the same reason and it too has the loose pin issue like the OCF.

Never had any of that with PGA and as long as you don't drop the chip it's fine, plus it's easier to fix a bent pin on the chip than it is down in a socket.

That's my whatever cents it's worth or not worth about it.
 

[cj09beira]

11 hours ago, leadeater said:

You won't find CXL implemented on any consumer devices though. Save maybe Apple doing something funky with memory tiering. CXL like a lot of things is still optional and you'll find it behind the likes of Quadro and Tesla paywalls for years to come.

dont break my heart like that 

[Chiyawa]

12 hours ago, AndreiArgeanu said:

It makes sense to be honest, I find it difficult to believe that 99.9% of consumers actually need PCIE 5.0.

Well, if it reduce the latency by a fraction, it is considered a big bonus for competitive gamers. The reason PCIe 4 became so wide spread is because of this. even nVidia who laughed at AMD for adopting PCIe 4 nonsense since GPU bandwidth only utilised half of PCIe 3.0 x16 bandwidth jumped into the band wagon.

 

Still, with PCIe 4, We have Smart Array Memory support, so it's actually a good thing to be honest. There's a lot of application can take advantage of this technology since the CPU can address every single memory cell inside the GPU directly. Maybe when PCIe 5 comes up, VRAM will be a thing of the past, as the GPU can share the super fast RAM that DDR5 offers. Of course they still need super fast cache.

[AndreiArgeanu]

34 minutes ago, Chiyawa said:

VRAM will be a thing of the past, as the GPU can share the super fast RAM that DDR5 offers.

Current GDDR6X memory found in an RTX 3090 can reach a theoritcal maximum speed of 936GB/s, while with DDR5 the information I could says that it can achieve a peak 51.2GB/s per module which is double DDR4-3200mhz. To add more context GDDR5X memory could reach peak speeds of about 325GB/s. I genuinely don't expect for DDR5 to replace any kind of memory on mid to high end gpu's. The most I expect is for GPU's to maybe use DDR5 as additional, slower memory when they run, kind of in a way that the page file in Windows works, or kind of in the way that the GTX 970 '4'gb worked. 

 

It might be good enough for lower end cards, and especially good for APU's. If AMD and Intel would work to make higher end integrated graphics for their APU's, with DDR5 I could definitely see them taking on the lower end GPU market.

 

 

35 minutes ago, Chiyawa said:

Well, if it reduce the latency by a fraction, it is considered a big bonus for competitive gamers.

I doubt that PCIE 4.0 causes a measurable difference in latency over PCIE 5.0. Even if PCIE 5.0 would be released I highly doubt NVIDIA or AMD would be quick to switch to it anyway, so all that latency improvement would go to waste.

[Chiyawa]

15 minutes ago, AndreiArgeanu said:

I genuinely don't expect for DDR5 to replace any kind of memory on mid to high end gpu's.

Indeed. High end GPU really need that speedy GDDR RAM. Low end GPU, such as GT1030, can just share the main system RAM. That would significantly reduce the cost of GPU card.

[Dabombinable]

34 minutes ago, Chiyawa said:

Indeed. High end GPU really need that speedy GDDR RAM. Low end GPU, such as GT1030, can just share the main system RAM. That would significantly reduce the cost of GPU card.

Look at how poorly the GT 1030 performed with dedicated DDR4 on its 64bit bus. Even low end cards such as the GT 1030 should not be forced to share system RAM. Otherwise we'll end up with the Geforce 6200 Turbocache all over again - a POS that is barely good as a display adapter, and with far worse performance than the standard version..