Installing 24GB DDR5 Modules on AMD Ryzen 7000 Platform Springs Mixed Results—POSTs but Doesn't Boot

[Dr. Dro]

I haven't asked Corsair yet for the new 24GB DIMMs, I should though.

Would be looking forward to it and a review of the 48 GB modules too, your memory reviews are a delight to read. Their Vengeance RGB 2 x 48 GB kit (CMH96GX5M2B5600C40) is already in stock in some stores here in Brazil, although, the prices are insanely high for such a low-performance module. At $385/440€ (and BRL 3.499/$665 here) for a 5600/C40 kit, one must really need so much memory quite badly for it to be even remotely worth it.

 

[Redwoodz]

Think BIOS level, not OS...

Except the user reported it boots into Windows... think less.

Everytime there is a problem with new AMD hardware, it inevitably ends up being poor Windows employement, which then later gets fixed after AMD gets blasted in the media. Every time. This is no coincidence. This is corruption.

 

[Dr. Dro]

Except the user reported it boots into Windows... think less.

Everytime there is a problem with new AMD hardware, it inevitably ends up being poor Windows employement, which then later gets fixed after AMD gets blasted in the media. Every time. This is no coincidence. This is corruption.

I'm in partial agreement. AMD's platforms are quite unorthodox for x86 system architecture - just look at Ryzen 9 X3D series. To compound the problem, AGESA, which is the lowest level firmware code in an AMD system and responsible for resource management, memory training, emulation of certain instructions etc. tends to be exceptionally buggy, often taking years for a platform to reach maturity. AM4 just recently got here. Then you have the need for certain custom drivers for chipset and scheduling, both of which also contain bugs. So yes, AMD has some blame in this situation.

On the other side, Windows is at a crossroads. Its commitment to backwards compatibility causes it to be bogged down by a LOT of legacy code making it exceptionally unwieldy, and to make things even worse, contains a very significant amount of third party licensed code that Microsoft can't touch.

Does Microsoft face the wrath of losing a lot of compatibility with older Windows software and clean up their act, or are we going to receive Windows 8 reskins forever from now on? Time will tell.

 

[ir_cow]

Ram incoming @Dr. Dro / @Mussels !

 

[Mussels]

Except the user reported it boots into Windows... think less.

Everytime there is a problem with new AMD hardware, it inevitably ends up being poor Windows employement, which then later gets fixed after AMD gets blasted in the media. Every time. This is no coincidence. This is corruption.

It's certainly something, when Intel have support for their E-cores at launch yet AMD require their own chipset drivers with custom power plans and prefer-X core features in their software, yet intels stuff is seamless added in earlier


It could be as simple as intel send hardware samples to intel along with lots of money to get shit working earlier, or it could be something more nefarious - but without any evidence, conspiracies are just imagination

 

[Sora]

I'm in partial agreement. AMD's platforms are quite unorthodox for x86 system architecture - just look at Ryzen 9 X3D series. To compound the problem, AGESA, which is the lowest level firmware code in an AMD system and responsible for resource management, memory training, emulation of certain instructions etc. tends to be exceptionally buggy, often taking years for a platform to reach maturity. AM4 just recently got here. Then you have the need for certain custom drivers for chipset and scheduling, both of which also contain bugs. So yes, AMD has some blame in this situation.

On the other side, Windows is at a crossroads. Its commitment to backwards compatibility causes it to be bogged down by a LOT of legacy code making it exceptionally unwieldy, and to make things even worse, contains a very significant amount of third party licensed code that Microsoft can't touch.

Does Microsoft face the wrath of losing a lot of compatibility with older Windows software and clean up their act, or are we going to receive Windows 8 reskins forever from now on? Time will tell.

It doesn't boot into windows, it hits critical process died on the the second phase of boot, which is usually poor timings corrupting the security descriptor.

Honestly, we're at a point where there is no guarantee of stability because the RAM is always pushed too hard and the mainboard bios isn't tuned enough to mitigate it.

 

[Dr. Dro]

It doesn't boot into windows, it hits critical process died on the the second phase of boot, which is usually poor timings corrupting the security descriptor.

Honestly, we're at a point where there is no guarantee of stability because the RAM is always pushed too hard and the mainboard bios isn't tuned enough to mitigate it.

The only way I can think of to mitigate this problem is by using relaxed timings and low frequencies... but it might not be enough.

 

[Sora]

Raja has 192GB running on a ASUS ROG STRIX X670-E

 

[Dr. Dro]

Raja has 192GB running on a ASUS ROG STRIX X670-E

View attachment 288176

Wew, that is some slow DDR5, though! I wonder how long until 96 GB kits make it to current high-performance segment (6600+) tier of performance. I don't think 192 GB kits will ever quite catch up, and in the event they do it definitely won't be on current generation platforms.

 

[Mussels]

Wew, that is some slow DDR5, though! I wonder how long until 96 GB kits make it to current high-performance segment (6600+) tier of performance. I don't think 192 GB kits will ever quite catch up, and in the event they do it definitely won't be on current generation platforms.

More ram = lower speed
It's a law of the universe these days

Although i question why you think it's "slow" when the comments on the thread are excited about how fast it is given the RAM amount

 

[Dr. Dro]

More ram = lower speed
It's a law of the universe these days

Although i question why you think it's "slow" when the comments on the thread are excited about how fast it is given the RAM amount
View attachment 288289

Yeah, but I mean, in the frequency ladders DDR5-5200 C40 is equivalent to DDR4-2400 or -2666, it's only a hair above JEDEC base and that already took a binned kit to achieve at this capacity.

 

[unwind-protect]

All these hacks, and just because unbuffered RAM has been kept in a crippled state. Per module capacity should have been doubled with the switch to DDR5, in the original spec. Instead we get 1.5x which might work somewhere sometime.

 

[Sora]

More ram = lower speed

higher timings, not necessarily lower speeds.

 

[Dr. Dro]

higher timings, not necessarily lower speeds.

Well, both are true, it all depends on the capacity. When filling all ranks and banks, maximum attainable frequency will be reduced due to the strain on the CPU's memory controller. Lately with DDR5 this is of extreme importance, mind you, with differences between fully populated vs. single-rank per channel in the 2000 to 3000 MT/s range.

 

[Mussels]

higher timings, not necessarily lower speeds.

No, IMC's force you to use lower MHz/MTs as you add more ranks.