Wednesday, February 7th 2024

AMD Ryzen 8000G Desktop APUs Don't Support ECC Memory

AMD's newly announced Ryzen 8000G "Hawk Point" desktop APUs do not support ECC memory, contrary to what the specifications on the AMD website had initially shown, Reddit users found out. The company has since quietly edited its product pages to remove the bit about ECC support. For the overwhelming majority of desktop client use cases, including enthusiast PCs, ECC memory support is irrelevant. That said, the memory controllers of "Phoenix" in Ryzen PRO 7000 mobile processors for commercial notebooks support ECC memory, and so it stands to reason that upcoming Ryzen PRO models for both commercial desktops and notebooks might feature it.

The AMD Ryzen 7 8700G and Ryzen 5 8600G are based on the 4 nm "Hawk Point" monolithic silicon, with a more overclocker-friendly set of DDR5 memory controllers than the ones found in the Ryzen 7000 "Raphael" processors. Besides support for several high-frequency DDR5 modes, the memory controller technically supports ECC (at least "Phoenix" does, on the Ryzen PRO 7000 mobile processors). The memory controller also supports a maximum of 256 GB of memory, or 64 GB dual-rank memory modules per slot. It also supports 24 GB and 48 GB DIMM densities.
Source: Tom's Hardware
Add your own comment

17 Comments on AMD Ryzen 8000G Desktop APUs Don't Support ECC Memory

#1
Wirko
Five or eight years from now, many people would start giving old and frugal 8000G PCs a second life as NAS boxes, and the absence of ECC will be a big downside at that time.
Posted on Reply
#3
azrael
Very disappointed, if I'm honest. Especially since the PRO version, if it really would support ECC, most likely is created from the same silicon. I'm currently speccing a PC with ECC memory for a colleague and explicitly asked him to wait for the 8000G series. Seems he waited unnecessarily. Will probably advise him to go with a Ryzen 7 7700 instead.
Posted on Reply
#4
jesdals
considering that a 7600 has igpu I dont see why one would want a 8000 series cpu for a NAS unless you already have one
Posted on Reply
#5
MaMoo
jesdalsconsidering that a 7600 has igpu I dont see why one would want a 8000 series cpu for a NAS unless you already have one
I think it may have to do with people using APUs for ML development. It cuts AMD out of a GPU sale, because for many tasks, brute power is not a problem but memory capacity. High memory capacity GPUs are expensive today.

That would be a guess.
Posted on Reply
#6
bonehead123
azraelI'm currently speccing a PC with ECC memory for a colleague and explicitly asked him to wait for the 8000G series. Seems he waited unnecessarily. Will probably advise him to go with a Ryzen 7 7700 instead
Yea, I was in the same position a few months ago, but my client said "screw it man, I needz my machine NOW" so I built it with the 7700, and he has been extremely happy with it !
Posted on Reply
#7
KarymidoN
I don't think its a big deal, DDR5 already has on-die ECC, wich is not a replacement for ECC, but makes the RAM more reliable. For home aplications that should be more than enough.
Posted on Reply
#8
Wirko
jesdalsconsidering that a 7600 has igpu I dont see why one would want a 8000 series cpu for a NAS unless you already have one
Certainly not now. But once the APU is old and ripe for recycling, it could become a great processor for a DIY NAS.
Posted on Reply
#9
MaMoo
KarymidoNI don't think its a big deal, DDR5 already has on-die ECC, wich is not a replacement for ECC, but makes the RAM more reliable. For home aplications that should be more than enough.
I am not certain on this - but I thought the on-die ECC was required because DDR5 runs closer to the error margins at its lofty speeds and ECC was a forethought to ensure that DDR5 would scale further not to reduce overall errors?
Posted on Reply
#10
Wirko
MaMooI am not certain on this - but I thought the on-die ECC was required because DDR5 runs closer to the error margins
Yes but...
MaMooat its lofty speeds
... it has nothing to do with speed. It protects data at rest (sitting in memory cells), not data in movement (on the way between IMC and memory cells, in either direction).
MaMooand ECC was a forethought to ensure that DDR5 would scale further not to reduce overall errors?
Yes when it scales towards smaller, less reliable cells, which it does. No when it scales towards higher speeds, which it also does.
Posted on Reply
#11
Random_User
These aren't news. Pretty much all previous retail APUs never had ECC support either. The last one 5700G didn't had, but it was PRO 5750G that did. It won't be a surprise, if the same case repeats again with 8000G and 8050G-ish. I'm pretty sure AMD will release PRO series as well. So, this's no biggie. Considering it was possible to buy legit tray PRO variant from many stores stores, for the same price as boxed 5700G

The real problem is ECC support on the motherboards. Considering how easily MB makers ditched seven segment display, CMOS reset, even service leds, and so many other features form even X670 class mobos (that were available even on B550 ones), where's the guarantee, that these vendors would want to invest in even more niche and even less requested feature like "ECC for desktop".

AMD clearly stated, that ECC is for workstations/HEDT, and for that purpose there's Threadripper for several kilobucks. AMD is new intel.
They dethronned intel from it's own invention (HEDT), that the last did for artifical segmentations and gouging purposes.
Posted on Reply
#12
Wirko
Random_UserConsidering it was possible to buy legit tray PRO variant from many stores stores, for the same price as boxed 5700G
But without the same retail warranty.
Random_UserThe real problem is ECC support on the motherboards.
It's reasonable to expect that AMD will also release PRO versions of the processors, and also that most mobos will support them. So hardware ECC support will be there, and firmware support for non-PRO chips probably too, if unofficial and unreliable.
Posted on Reply
#13
stimpy88
Amateur AMD copying Intel's lead. Pathetic.
Posted on Reply
#14
unwind-protect
KarymidoNI don't think its a big deal, DDR5 already has on-die ECC, wich is not a replacement for ECC, but makes the RAM more reliable. For home aplications that should be more than enough.
Nah. It doesn't report errors - corrected or not - to the OS. That makes it nearly useless as you can't recognize a DIMM going bad.
Posted on Reply
#15
KarymidoN
MaMooI am not certain on this - but I thought the on-die ECC was required because DDR5 runs closer to the error margins at its lofty speeds and ECC was a forethought to ensure that DDR5 would scale further not to reduce overall errors?
unwind-protectNah. It doesn't report errors - corrected or not - to the OS. That makes it nearly useless as you can't recognize a DIMM going bad.
Traditional ECC = Ensures data integrity by handling memory errors while data is being moved.
On-die ECC = Ensures higher reliability of higher-density memory and protects the data that is in the memory chip.

Yes, different things. But, for home aplications its waaay overkill to use ECC RAM. APU's never had ECC and it never seen to stop people from using them before for home aplications.
Posted on Reply
#16
kapone32
Could it be that this allows for faster Memory support?
Posted on Reply
#17
KarymidoN
kapone32Could it be that this allows for faster Memory support?
Derbauer delided the 8700g, it's obvious they had to cut corners to fit everything in the monolithic die.
Posted on Reply
Add your own comment
May 21st, 2024 16:52 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts