Hynix A or M die for "futureproofed" AM5 build?

[Braegnok]

I have a feeling the 6000CL26 and 6400CL28 are the same quality bin. If your into memory overclocking, get really cheap 5600 A-Die KD5AGUA80-56G460D and drop it to CL24. It will need a fan and 1.5v, but it will be half the price.

AMD rules is highest frequency possible and lowest latency in Gear 1 mode. That means for 99% of users the limit is 6400 for 7000/9000 series.

Will you see the difference from 6000CL26 to CL32? Probably not. Mostly depends on the game and what is acceptable to you in 1% lows.


I just got the RGB NEO version Friday. On my test bench now. Reminds me of DDR4-4000 CL15. High bin, but pricey.

Yes, the RGB Neo are very good kit's,.. I tested the 6400CL30 kit last month, was good. The 6000CL26 was better ICs,..

 

[ir_cow]

I'm not really into memory overclocking (or overclocking in general). That doesn't mean I don't want as much speed as possible, even if it's ECC memory. And since I plan to upgrade to Zen 6 eventually (unless that turns out to be impossible) I would like for the memory not to be outdated specwise. Assuming the Zen 6 IMC is at least as capable as the current ones.

If your buying ECC memory, the last thing you want to do is overclock. That defeats the purpose of a workstation.

 

[A Computer Guy]

The full specs are these for the interesting two 2x 48GB kits. Price difference is $100, as stated above. I'm still baffled that they've managed to eke out 6400 MT/s at 1.1v.

5600MHz	PC5-44800	CL-46-45-45-90	1.1V
6400MHz	PC5-51200	CL-52-52-52-103	1.1V

Really not much of a difference in latency. The 6400MT/s is probably just a better bin of the same chips.

If your unlucky with your CPU you might need to downclock to 6200MT/s or 6000MT/s with a slight penalty if you don't tweak your timings.

 

[azrael]

If your buying ECC memory, the last thing you want to do is overclock. That defeats the purpose of a workstation.

Again, I do not want to overclock, but speed and ECC are not mutually exclusive. FWIW I'm pretty certain that at one point all memory will have ECC functionality. On-die ECC is already mandatory for DDR5, and speeds and densities will only increase.

Really not much of a difference. The 6400MT/s is probably just a better bin of the same chips.

View attachment 385505
If your unlucky with your CPU you might need to downclock to 6200MT/s or 6000MT/s with a slight penalty if you don't tweak your timings.

View attachment 385508

Pretty much my thoughts. So are the slightly better binned ICs worth $100 more. The kits are already very expensive as it is.

 

[A Computer Guy]

Again, I do not want to overclock, but speed and ECC are not mutually exclusive. FWIW I'm pretty certain that at one point all memory will have ECC functionality. On-die ECC is already mandatory for DDR5, and speeds and densities will only increase.


Pretty much my thoughts. So are the slightly better binned ICs worth $100 more. The kits are already very expensive as it is.

$100 for 400MHz (800MT/s) faster RAM. Will it matter I suppose it depends if you run RAM speed sensitive workloads. I wouldn't overthink it. If you have the cash - get the best performance you can use for your platform.

 

[ir_cow]

Again, I do not want to overclock, but speed and ECC are not mutually exclusive.

It is to a point. You will only find JEDEC memory with true ECC support.

FWIW I'm pretty certain that at one point all memory will have ECC functionality. On-die ECC is already mandatory for DDR5, and speeds and densities will only increase.

On-site ECC for DDR5 is only in the resting state, not transferring.

 

[azrael]

It is to a point. You will only find JEDEC memory with true ECC support.

Tell that to v-color...

On-site ECC for DDR5 is only in the resting state, not transferring.

I'm fully aware of that. Actually, if there's something I'm familiar with it's ECC. My proposition stands: Already now on-die ECC is mandatory for DDR5. Why? Because you wouldn't be able to manufacture any working DDR5 modules without it. At some point I'm pretty sure side-band ECC will also be more or less mandatory, for the reasons stated above. It only takes some background radiation to flip a bit. Doesn't even have to be a faulty module.

 

[A Computer Guy]

To play devils advocate if you get the 5600MT/s you can put that saved $100 to get JONSBO NF-1 ram cooler and Thermal Grizzly AM5 offset kit. Caution be warned that staring at the RGB might increase your decision making latency.

 

[ir_cow]

Tell that to v-color...

1.1 and high ass CAS. I smell JEDEC. Unless you need ECC stuff for a workstation, 4800 CL40 is going to perform the same.

Edit: Not really, but saving 10 second off a 30 minute render isn't going to matter.

 

[_roman_]

Already now on-die ECC is mandatory for DDR5

No it is not.

In my own words. Current Dims have some sort of FAKE ECC. Those ECC modules have some other form of ECC.
I'm happy we got a bit of error correction.

 

[ir_cow]

I must agree with @_roman_ on this one.

Also going back and reading this thread, I'm not sure the op even know what they want.

 

[ShrimpBrime]

Tell that to v-color...

I'm fully aware of that. Actually, if there's something I'm familiar with it's ECC. My proposition stands: Already now on-die ECC is mandatory for DDR5. Why? Because you wouldn't be able to manufacture any working DDR5 modules without it. At some point I'm pretty sure side-band ECC will also be more or less mandatory, for the reasons stated above. It only takes some background radiation to flip a bit. Doesn't even have to be a faulty module.

Bro. Did you even read that page?

It is unregistered memory. So it won't error correct single flip bit.

You want to look for Registered dimms.

And good luck finding one with an "XMP" profile.

Unless you intend to max the board capacity (128gb?) And actually utilize such capacity, then maybe consider registered dimms.

Other than that, no. There is no such thing as overclock profiles on registered memory.

 

[freeagent]

How effective is ECC when they are run out of spec?

 

[ShrimpBrime]

How effective is ECC when they are run out of spec?

You go memtest it and find out.

I'm gonna fire up 7000mt/s and maybe bench a video card, or play some games.

 

[A Computer Guy]

How effective is ECC when they are run out of spec?

That's an interesting question.

The obvious thing of course is if you do OC the ECC RAM essentially you are throwing away any guarantee of it's error correction capabilities that otherwise would have come with running the part at spec.

Just speaking from my own limited experiences...

When I OC'ed my DDR4 2667MT/s ECC to 3200MT/s ( UDIMM's ) it worked just fine for several years until I could get a higher capacity certified 3200MT/s kit. Forcing timings out of wack to force the system to correct errors still worked at the higher OC speed and corrections were logged by Windows. Now I'm not saying one should OC their ECC RAM but it can still function as far as I know since like normal RAM it's still just a complicated yet dumb part. ECC is more restrictive and unpredictable in OC ability because more circuitry and the PCB is in no way geared for it, not to mention the chips aren't binned for OC either. Was it worth OC'ing - probably not - I was only able to tell the difference in performance with synthetic benchmarks.

Way before my AM4 experience my P5E-WS with Core-2 Quad Q6600 used to auto overclock my DDR2 ECC and I didn't even realize it at the time until many years later - thanks Asus! It worked for about 10yrs overclocked from 667MHz to something like 1066MHz (I can't remember exactly what the number was) until one of the ECC sticks starting dying.

Now on AM5 even though I choose my ECC RAMS built-in 5600MT/s profile the UEFI decided to do a little OC of it's own and reduced my CL from 46 to 44 because memory training I guess. Technically it's an overclock but I kind of hate the idea it does this on it's own and didn't strictly follow the modules defined setting which is what I expected. So far I have been 100% stable so I am rolling with it for now.

 

[trparky]

I think their focus is the OEM market more now so not a big priority for then.

As disappointing as that might be, you might be right.

 

[ShrimpBrime]

That's an interesting question.

The obvious thing of course is if you do OC the ECC RAM essentially you are throwing away any guarantee of it's error correction capabilities that otherwise would have come with running the part at spec.

Just speaking from my own limited experiences...

When I OC'ed my DDR4 2667MT/s ECC to 3200MT/s ( UDIMM's ) it worked just fine for several years until I could get a higher capacity certified 3200MT/s kit. Forcing timings out of wack to force the system to correct errors still worked at the higher OC speed and corrections were logged by Windows. Now I'm not saying one should OC their ECC RAM but it can still function as far as I know since like normal RAM it's still just a complicated yet dumb part. ECC is more restrictive and unpredictable in OC ability because more circuitry and the PCB is in no way geared for it, not to mention the chips aren't binned for OC either. Was it worth OC'ing - probably not - I was only able to tell the difference in performance with synthetic benchmarks.

Way before my AM4 experience my P5E-WS with Core-2 Quad Q6600 used to auto overclock my DDR2 ECC and I didn't even realize it at the time until many years later. It worked for about 10yrs overclocked from 667MHz to something like 1066MHz (I can't remember exactly what the number was) until one of the ECC sticks starting dying.

Now on AM5 even though I choose my ECC RAMS built-in 5600MT/s profile the UEFI decided to do a little OC of it's own and reduced my CL from 46 to 44. Technically it's an overclock but I kind of hate the idea it does this on it's own and didn't strictly follow the modules defined setting which is what I expected. So far I have been 100% stable so I am rolling with it for now.

You havent passed any Jedec frequencies yet.

And Q6600 doesn't support ECC.

 

[A Computer Guy]

And Q6600 doesn't support ECC.

I remember it did, paired with X38 chipset. That went into my purchasing decision at the time.

 

[ShrimpBrime]

I remember it did, paired with X38 chipset. That went into my purchasing decision at the time.

I thought, back then, had to be a xeon. Same thing on 1366, which I have, the cpu i7 980 and 980x do not support ECC.

The problem from what I can remember, is that there's really no way to prove it's working. It's either enabled in bios with appropriate hardware or disabled.

So you may have had the board and memory with support, enabled in bios, but if the cpu doesn't support it, it probably wasn't doing error correction.

I had K8N-DRE (couple few of them through the years) which did not require ECC memory and was one of a couple of socket 940 boards that was over-clockable. So did not bother because wanted performance. Obviously not using the hardware for it's purposes...

 

[A Computer Guy]

I thought, back then, had to be a xeon. Same thing on 1366, which I have, the cpu i7 980 and 980x do not support ECC.

Just like today select non-Xeon chips support ECC however back then the memory controller was not on the CPU so chipset selection was important to having ECC capability. In this era to the best of my knowledge the CPU needed to support the correct memory interface the motherboard implemented. In my case with the Q6600 the Asus P5E-WS required DDR2 and supported DDR2 ECC memory. There may have also been some other consideration specifically with the Q6600 that I have forgotten since I had assembled it in 2008/2009.

The problem from what I can remember, is that there's really no way to prove it's working. It's either enabled in bios with appropriate hardware or disabled.

I think there are several things you can look at in general when it comes to ECC memory support.

Prerequisites
1) Properly researched part selection for ECC support ( CPU, Motherboard , RAM )

Your vendor parts should indicate if they support ECC and in what conditions. For example if you buy a Xeon, and Server/Workstation motherboard, and ECC RAM there is a high degree of trust they will work together as advertised provided you match the parts correctly.

AM4 is a bit dubious as for their non-PRO CPU's the ECC hardware is "present" but not validated. On the PRO CPU's ECC is properly validated. The exception being non-PRO G series and laptop CPU's don't support ECC at all. Some early embedded CPU's models V1000 and V2000 series I think also supported ECC memory.

AM5 as far as I know ECC is validated for non-PRO CPU's and PRO CPU's alike. The exception being non-PRO G series and laptop CPU's don't support ECC at all.

Indicators
2) UEFI/BIOS Post screen and/or UEFI/BIOS indicating installed ECC memory
3) OS commands indicating detected ECC memory

Testing
4) The vendor should have done this for you through their validation programs
5) MemTest86 with error injection (requires UEFI/BIOS support and possibly CPU dependent) For example error injection works with AM4 PRO CPU's.
6) Overclocking to force error correction (see logging)
7) DIMM riser (hardware) to induce errors (see logging)
8) Pin shorting (I don't recall the specifics of what pins) (see logging)

Logging
9) UEFI/BIOS logged events of memory correction or failed correction
X) OS logged events of memory correction or failed correction

So you may have had the board and memory with support, enabled in bios, but if the cpu doesn't support it, it probably wasn't doing error correction.

I'm pretty sure it was working. When memory can't be corrected the system is supposed to halt or reboot. When my DDR2 module was going bad the system did reboot instead of corrupting all my stuff so I'm reasonably sure it was working.

 

[AusWolf]

Nobody knows much, if anything about Zen 6. As for current configs, any high speed RAM is fine (you don't even need high speed with X3D), unless you think a 0.1% difference matters.

 

[ShrimpBrime]

Just like today select non-Xeon chips support ECC however back then the memory controller was not on the CPU so chipset selection was important to having ECC capability. In this era to the best of my knowledge the CPU needed to support the correct memory interface the motherboard implemented. In my case with the Q6600 the Asus P5E-WS required DDR2 and supported DDR2 ECC memory. There may have also been some other consideration specifically with the Q6600 that I have forgotten since I had assembled it in 2008/2009.



I think there are several things you can look at in general when it comes to ECC memory support.

Prerequisites
1) Properly researched part selection for ECC support ( CPU, Motherboard , RAM )

Your vendor parts should indicate if they support ECC and in what conditions. For example if you buy a Xeon, and Server/Workstation motherboard, and ECC RAM there is a high degree of trust they will work together as advertised provided you match the parts correctly.

AM4 is a bit dubious as for their non-PRO CPU's the ECC hardware is "present" but not validated. On the PRO CPU's ECC is properly validated. The exception being non-PRO G series and laptop CPU's don't support ECC at all. Some early embedded CPU's models V1000 and V2000 series I think also supported ECC memory.

AM5 as far as I know ECC is validated for non-PRO CPU's and PRO CPU's alike. The exception being non-PRO G series and laptop CPU's don't support ECC at all.

Indicators
2) UEFI/BIOS Post screen and/or UEFI/BIOS indicating installed ECC memory
3) OS commands indicating detected ECC memory

Testing
4) The vendor should have done this for you through their validation programs
5) MemTest86 with error injection (requires UEFI/BIOS support and possibly CPU dependent) For example error injection works with AM4 PRO CPU's.
6) Overclocking to force error correction (see logging)
7) DIMM riser (hardware) to induce errors (see logging)
8) Pin shorting (I don't recall the specifics of what pins) (see logging)

Logging
9) UEFI/BIOS logged events of memory correction or failed correction
X) OS logged events of memory correction or failed correction


I'm pretty sure it was working. When memory can't be corrected the system is supposed to halt or reboot. When my DDR2 module was going bad the system did reboot instead of corrupting all my stuff so I'm reasonably sure it was working.

Here is Intels list. You can utilize this to see exactly what cpus support ECC.

Intel product specifications

Intel® product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.

www.intel.com

Unfortunately, it doesn't look to go as far back and cover socket 775.

Keep clicking show more.

It ends at 2007. All xeons only. No i5s. No i7s. No extreme chips. (For that time period)

CPU must have ECC support code embedded. Back then, it wasn't on desktop. Doesn't matter where the memory controller is loacted.

Today, they do offer registered ECC support on just about everything.

The claims where, to point out...

You didn't really overclock past ECC JEDEC.
Q6600 doesn't support ECC.

Maybe OCCT burn in shows memory errors, but the system doesn't halt or crash, but doesn't mean it's showing single bit errors either. Cause I am not sure that part.

But you do make good points that I don't have the data to disprove.

 

[azrael]

Bro. Did you even read that page?

It is unregistered memory. So it won't error correct single flip bit.

You want to look for Registered dimms.

And good luck finding one with an "XMP" profile.

Unless you intend to max the board capacity (128gb?) And actually utilize such capacity, then maybe consider registered dimms.

Other than that, no. There is no such thing as overclock profiles on registered memory.

It seems you're confusing things. ECC and registered memory are two separate things. Registered memory allows for huge amounts of memory because it is buffered. For that reason it is also quite a bit slower. ECC works on registered and unbuffered memory alike *if* the CPU, the motherboard and the memory modules support it. Again, I've used ECC memory for 30 years, so you would think I know a bit about this stuff. And yes, I'm that old.

I must agree with @_roman_ on this one.

Also going back and reading this thread, I'm not sure the op even know what they want.

Really? It seems you don't really know much about ECC, particularly with regard to DDR5. Try looking up "on-die ECC" (mandatory on DDR5) and "side-band ECC". And let me assure you, I definitely know what I want.

I'm sorry this thread got derailed, because I mentioned ECC (or rather, someone else did). As I've stated before, that was the exact reason I did not mention it in my original post. I get it, most people here aren't concerned with their data, but only performance. I like to have both, even if it means giving up some of the performance.

 

[_roman_]

It's so hard to read, comprehend, think and than write.

The first search result.
that link comes from. -> https://superuser.com/questions/186...-between-ecc-and-ddr5s-built-in-data-checking

https://d2pgu9s4sfmw1s.cloudfront.net/UAM/Prod/Done/a062E00001fCp2QQAS/3549300d-62b2-208b-e140-9055ee5963d3?response-content-disposition=inline%3Bfilename*%3DUTF-8%27%27102021-ecc-explainer-sales-brief-final.pdf&Expires=2001034924&Key-Pair-Id=APKAJKRNIMMSNYXST6UA&Signature=gkRiaR9BIp7EOiJ2yKDb9rBKixTf7xN2tAGcK~C6k5BmJGfQZ3mkDMloucIa5W0f5lMBSqGEBplZGcrVgOx9FfxNliMCJu46QlK~wL3pMO4HIfH~YqlZUi2npkVBDRq9RDVN0VxoL99zrcRnvCmVzEYp2H6KYyXFrIQc~4fAVKrCFGClciZeituQnklB~6d69glP~XexEYg6uG62lOMzkaS4as3hXLu3umoc5PB2QRP2Ny4xK64yJjMrxVN5HsqC-q1tRpM7E93UuUukeaYbXw04Nz6HA5l0Igcwo3O-m3OCLqZIJi6RRq~NAWqZqLNSAf3Mx0BFGKLeYCjaYo~E2g__

a few starting points

What is on-die ECC?

According to Crucial, Designed with on-die ECC (ODECC)8 at the component level for long-term stability, Crucial DDR5 is engineered to maintain the same reliability as the previous generation, even...

superuser.com

--

You may take your time and search before you respond please.

Do not write. Bring me arguments that I am wrong. PDF or news articles from micron or such.

I bet one coffee I'm right on the statement. FAKE ECC vs REAL ECC.

You should also know that by only reading the usual tech sites like computerbase.de / hardwareluxx.de / pcgameshardware.de and the other sites. That information was written several times when we had the changes from DDR4 to DDR5.

--

The argument I used 30 years ECC and know a bit is not valid.
I also drive a car for 25 years. Still I'm a bad driver.
Some people used a screwdriver for 40 years and still damage those plastic clips or ruin mainboards by scratches the pcb of the mainboards. So what.
Age is not really an indicator for knowledge.

 

[azrael]

It's so hard to read, comprehend, think and than write.

The first search result.
that link comes from. -> https://superuser.com/questions/186...-between-ecc-and-ddr5s-built-in-data-checking

https://d2pgu9s4sfmw1s.cloudfront.net/UAM/Prod/Done/a062E00001fCp2QQAS/3549300d-62b2-208b-e140-9055ee5963d3?response-content-disposition=inline%3Bfilename*%3DUTF-8%27%27102021-ecc-explainer-sales-brief-final.pdf&Expires=2001034924&Key-Pair-Id=APKAJKRNIMMSNYXST6UA&Signature=gkRiaR9BIp7EOiJ2yKDb9rBKixTf7xN2tAGcK~C6k5BmJGfQZ3mkDMloucIa5W0f5lMBSqGEBplZGcrVgOx9FfxNliMCJu46QlK~wL3pMO4HIfH~YqlZUi2npkVBDRq9RDVN0VxoL99zrcRnvCmVzEYp2H6KYyXFrIQc~4fAVKrCFGClciZeituQnklB~6d69glP~XexEYg6uG62lOMzkaS4as3hXLu3umoc5PB2QRP2Ny4xK64yJjMrxVN5HsqC-q1tRpM7E93UuUukeaYbXw04Nz6HA5l0Igcwo3O-m3OCLqZIJi6RRq~NAWqZqLNSAf3Mx0BFGKLeYCjaYo~E2g__

a few starting points

What is on-die ECC?

According to Crucial, Designed with on-die ECC (ODECC)8 at the component level for long-term stability, Crucial DDR5 is engineered to maintain the same reliability as the previous generation, even...

superuser.com

--

You may take your time and search before you respond please.

Do not write. Bring me arguments that I am wrong. PDF or news articles from micron or such.

I bet one coffee I'm right on the statement. FAKE ECC vs REAL ECC.

You should also know that by only reading the usual tech sites like computerbase.de / hardwareluxx.de / pcgameshardware.de and the other sites. That information was written several times when we had the changes from DDR4 to DDR5.

--

The argument I used 30 years ECC and know a bit is not valid.
I also drive a car for 25 years. Still I'm a bad driver.
Some people used a screwdriver for 40 years and still damage those plastic clips or ruin mainboards by scratches the pcb of the mainboards. So what.
Age is not really an indicator for knowledge.

What are you on about? I've never claimed that "on-die ECC" is the same as "side-band ECC". Quite the contrary. The former is more akin to error correction on an SSD. Please stop putting words into my mouth. I know EXACTLY what ECC is and how it works. Why is this so difficult to comprehend? I think you should be the one to read up on ECC. Just google the terms "on-die ECC" and "side-band ECC".

As for you being a bad driver after 25 years of driving. Sounds like a "you" problem. Full disclosure, I don't have a driver's license.