Do you think AMD will have validated ECC for Zen 2 AM4 CPUs?

[cbn]

Do you think AMD will have validated ECC for Zen 2 AM4 CPUs?

(With core count increasing on 7nm I think it would be nice to have this feature at least on certain SKUs, if not all of them)

 

[plopke]

I think it will be kinda the same as now , yes you can use it , validation is up to other people like OEM's if they want to do it.

 

[ultimatebob]

No. They'll only validate it for the Epyc and the Ryzen Pro line.

 

[cbn]

No. They'll only validate it for the Epyc and the Ryzen Pro line.

Ryzen Pro is actually AM4. Maybe you mean Threadripper? (I've been a bit confused on whether that one is validated for ECC or not.)

 

[cbn]

I think it will be kinda the same as now , yes you can use it , validation is up to other people like OEM's if they want to do it.

One thing that surprised me was that neither HP, Dell or Lenovo chose Ryzen or Threadripper for their lower end desktop workstation lines. (ie, the workstation lines that use Intel LGA 1151 and Intel LGA 2066 processors)

And at the same time there is some confusion on the completeness of ECC support for the Ryzen processors among White box motherboard OEMs (eg, ASUS, Gigabyte, MSI, ASRock, Biostar, ECS, etc). Not sure about completeness of ECC Support for Threadripper on Whitebox notherboards as I haven't looked there very much.

 

[cbn]

Current Poll results:

Do you think AMD will have validated ECC for Zen 2 AM4 CPUs?

1. Yes
   0 vote(s)
   0.0%
2. No
   5 vote(s)
   100.0%

For those that voted, do you think AMD should validate Zen 2 CPUs on AM4?

 

[OTG]

Current Poll results:

Do you think AMD will have validated ECC for Zen 2 AM4 CPUs?

1. Yes
   0 vote(s)
   0.0%
2. No
   5 vote(s)
   100.0%

For those that voted, do you think AMD should validate Zen 2 CPUs on AM4?

I can't see any reason why Zen 2 would be different from Zen 1 in that way- the processors are able to use ECC memory, but since 99% of the target market doesn't have any use for ECC, there's no point in spending $[whatever it costs to validate] on doing that.
Almost certainly there won't be any issues, but what's the point when that money can go towards something with a better cost/benefit ratio.

 

[cbn]

I can't see any reason why Zen 2 would be different from Zen 1 in that way- the processors are able to use ECC memory, but since 99% of the target market doesn't have any use for ECC, there's no point in spending $[whatever it costs to validate] on doing that.
Almost certainly there won't be any issues, but what's the point when that money can go towards something with a better cost/benefit ratio.

What do you think about what I wrote in post #5?

One thing that surprised me was that neither HP, Dell or Lenovo chose Ryzen or Threadripper for their lower end desktop workstation lines. (ie, the workstation lines that use Intel LGA 1151 and Intel LGA 2066 processors)

And at the same time there is some confusion on the completeness of ECC support for the Ryzen processors among White box motherboard OEMs (eg, ASUS, Gigabyte, MSI, ASRock, Biostar, ECS, etc). Not sure about completeness of ECC Support for Threadripper on Whitebox notherboards as I haven't looked there very much.

Also factor in AMD is pretty weak in OEM consumer and business desktop. (Therefore it seems to me for one than one reason workstation is a better place for AMD's AM4 CPUs)

 

[sandorski]

Ryzen-no
Threadripper-yes
Epyc-yes

They could do all 3, but it doesn't make business sense to me. Even Threadripper seems like ECC isn't entirely necessary, but I suspect some usage/customers may prefer/demand it.

 

[CatMerc]

It will be like it is now with Ryzen most likely. It's there, but AMD doesn't promise anything.

 

[cbn]

ECC isn't entirely necessary, but I suspect some usage/customers may prefer/demand it.

I don't think HP, Dell or Lenovo would actually build ECC capable AM4 motherboards for AMD workstations (these positioned alongside LGA 1151 and LGA 2066 Workstations) if the AMD processors sold were not guaranteed ECC (ie, validated).

However, I do think HP, Dell or Lenovo would use a consumer AMD chipset with the workstations though. (Examples were the Intel 1P LGA 1366 workstations where a Xeon was combined with ECC UDIMMs using X58 chipset)

 

[OTG]

What do you think about what I wrote in post #5?

Also factor in AMD is pretty weak in OEM consumer and business desktop. (Therefore it seems to me for one than one reason workstation is a better place for AMD's AM4 CPUs)

Huh, I missed that, good point.
It does seem like they'll want to get into that market ASAP, given Ryzen's strengths.
I wonder if that's enough to make ECC worth it for the whole lineup.
Would Ryzen and Ryzen Pro need to be validated separately, considering they're the same processors?

 

[Mr Evil]

I would like to see it, I think it should be standard across all PCs, but I don't think there is enough demand. I'm just happy that AMD don't deliberately disable it like some other x86 CPU manufacturers do.

...99% of the target market doesn't have any use for ECC...

99% might think they have no use for it, but they still feel the pain of corrupted data when it happens.

 

[moinmoin]

Short answer: No. AMD didn't deactivate ECC on AM4 so as a feature it's fully usable but optional. Being optional means both support and validation is not up to AMD but up to the boards' manufacturers.

I guess the better question would be: Will AMD launch an AM4 Pro platform or some such that makes validated ECC an obligatory part of it (but is otherwise the same as AM4)? I don't expect that but AMD prolly should make it easier for people looking to combine AM4 with validated ECC.

 

[cbn]

Huh, I missed that, good point.
It does seem like they'll want to get into that market ASAP, given Ryzen's strengths.
I wonder if that's enough to make ECC worth it for the whole lineup.
Would Ryzen and Ryzen Pro need to be validated separately, considering they're the same processors?

What Intel does is have some of their chips (using the same processor die) validated for ECC while others have ECC disabled. (Eg, Xeon E3-1200 v6 vs. quad core Kabylake consumer processors like Core i5 7600 and Core i7 7700K, etc.)

 

[cbn]

Biggest question I have about the current situation of combining AM4 processors with ECC enabled motherboards is "How do I know the AM4 processor is really a bin capable of ECC"? Maybe it is or maybe it is not? How can I tell?

 

[moinmoin]

How do I know the AM4 processor is really a bin capable of ECC?

Are there any Zen based AM4 processors that are not capable of ECC? I thought it's all down to the boards, not the processors.

 

[cbn]

Are there any Zen based AM4 processors that are not capable of ECC? I thought it's all down to the boards, not the processors.

The memory controller is in the processor, so the processor needs to have the ECC feature verified working (just like other features on the chip). Otherwise it needs to be disabled.

Furthermore, there has been some issue with the completeness of ECC in AM4 motherboards:

http://www.hardwarecanucks.com/foru...ws/75030-ecc-memory-amds-ryzen-deep-dive.html

In conclusion, what is currently available on the AM4 platform is an incomplete implementation of ECC. This is very likely why motherboard manufacturers have been relatively hesitant about claiming that their products support ECC memory in ECC mode. Based on our findings, there is clearly some level of ECC functionality that is working right now, but it does not cover the full spectrum of memory error detection and correction. Having said that, the status quo is arguably better than nothing, especially since single-bit errors are much more likely than multi-bit errors (which are often caused by a failing memory module), so I suspect that many people will still want the extra protection that is available right now.

What I would like to see is AM4 processors that are checked/verified/validated by AMD to have guaranteed ECC working (ie, Workstation bin silicon) and then have a ECC standard for motherboard manufacturers to follow.

 

[moinmoin]

The memory controller is in the processor, so the processor needs to have the ECC feature verified working (just like other features on the chip). Otherwise it needs to be disabled.

I never have heard of it being disabled in any Zen processor (I wonder if the memory controller is even modularized to have such a feature hard gated, I'd expect not). I think nobody doing the full chain validation (since it's optional) is the "only" issue.

 

[cbn]

I never have heard of it being disabled in any Zen processor (I wonder if the memory controller is even modularized to have such a feature hard gated, I'd expect not). I think nobody doing the full chain validation (since it's optional) is the "only" issue.

If you were an OEM like HP, Dell or Lenovo would you make a workstation from chips that might have ECC working? Or would you want it guaranteed?

P.S. EPYC (which uses the same dies as AM4) has ECC RDIMM functionality (while AM4 only has ECC UDIMM). So various levels of ECC can be enabled/disabled just like various levels of memory speed or RAM limit can be enabled/disabled at the processor level.