Proof that Asrock (and SM, others) are/were gatekeeping Milan on cheaper boards for no reason

[gsrcrxsi]

I asked about this here and to Asrock directly in the past for Milan support on Asrock EPYC series motherboards (and H11 SM boards), which are much cheaper than AR ROME or SM H12 boards, especially if you don't care about PCIe gen 4.0. I was met with excuses and nonsensical arguments and an overall refusal to engage. they just want to force users to upgrade to more expensive boards. There's no technical reason to limit support, only corporate greed.

Them: "we can't support Milan because the BIOS only holds enough for two CPU series"
Me: So? you support 7001/7002 and 7002/7003 on your ROME series boards with two separate BIOS versions. Why can't you do that?
Them: *crickets*

Them: "we can't support Milan because that board doesn't support PCIe gen 4.0"
Me: So? Romeis supported which also has gen 4.0. so it still works at 3.0 speeds, and most users wanting to do this will accept this caveat since they already do for Rome. so why is this any different for Milan?
Them: *crickets*

Them: we can't support Milan because the TDP is higher at 280W"
Me: So? you support Rome 7H12, which also has 280W. even if you have to limit it to 240W, that's OK.
Them: *crickets*

I was holding out for someone to modify the AR EPYC or SM H11 boards' BIOS to add in Milan support, as I saw some BIOS mods happening with at least the H11 boards for certain CPU support (like adding 7002 to rev1 boards), but it never happened. not sure if there just wasnt any interest or they weren't able or what.

but... well well well...

*I was given this BIOS freely by Asrock support, probably accidentally, but given freely nonetheless. they ghosted me when I started asking dangerous questions like "Why is this BIOS versioned at 3.x?" and "Why did you remove Naples support for a BIOS for the EPYC-series board?" and "Does this BIOS support Milan?"

EPYCD8L3.02.zip

Shared with Dropbox

www.dropbox.com

 

[Jorge Perez]

You do understand it does cost them money to add support right?

 

[RolloZ170]

it is also a calculation. the motherboards support usual only two generations.
if they expect three the prices must rise to make (just a little) profit (that is why they build motherboards)
if we avoid SM in some time only the vendors remain which do what you want, prices rise (imagen gigabyte's are not the cheapest)
shortly afterwards these manufacturers goes bankrupt.

 

[gsrcrxsi]

You do understand it does cost them money to add support right?

You do understand that they already have and created this bios and intentionally chose to hold it back right? Releasing this costs them nothing.

lots of bootlicking going on here it seems

 

[gsrcrxsi]

it is also a calculation. the motherboards support usual only two generations.
if they expect three the prices must rise to make (just a little) profit (that is why they build motherboards)
if we avoid SM in some time only the vendors remain which do what you want, prices rise (imagen gigabyte's are not the cheapest)
shortly afterwards these manufacturers goes bankrupt.

This is fallacy #1 from my post. They already support all three versions on the ROME boards with different BIOS. They don’t only support 2 there.

 

[RolloZ170]

You do understand that they already have and created this bios and intentionally chose to hold it back right?

what is wrong with this ? what would you do if you where Mr. ASRock Rack ?

 

[RolloZ170]

This is fallacy #1 from my post. They already support all three versions on the ROME boards with different BIOS

"usualy"

 

[gsrcrxsi]

what is wrong with this ? what would you do if you where Mr. ASRock Rack ?

Support all three generations like AMD said is possible/intended. SP3 was always supposed to bridge all three 7001/7002/7003 and their official platform does. Just the various manufacturers decided to cut off support for Milan for no reason to fleece the consumer. People who want/need gen4 would still buy the better ROME/H12 boards. And consumers who just want the better IPC and cache structure of Milan for HPC but don’t care about gen 4 would be happier.

 

[RolloZ170]

And consumers who just want the better IPC and cache structure of Milan for HPC but don’t care about gen 4 would be happier.

they buy used boards. ASRock don't want this, ASRock want to make profit, not get bankrupt.
you would do same.

 

[gsrcrxsi]

I disagree. I bought several boards new before the ROME boards were even released. And the messaging from AMD was that all SP3 boards would/could support Milan. So many consumers, myself included, had an appropriate assumption that they would be grandfathered in, only to be shafted by the board manufacturers when Milan actually released. I have three of the EPYCD8 boards, only one of which was bought used (recently) when I needed to build another setup in the cheap.

closer to when Milan was released, I did buy a single ROMED8-2T board new for use with a 7443P as a nice workstation/toy build to see how much benefit Milan had and generally play with things. It’s a great board, but I don’t have a need for gen4 and it’s much better to use a cheaper board. I’m certain I’m not the only one in this situation. I would have upgraded the systems to Milan by now if not for the huge price difference in their EPYC/ROME lineups. The EPYCD8 board was like $330 new. I never saw new pricing for ROMED8-2T less than like $650 or $700 around launch (they didn’t have cheaper options that provided 7x PCIe in ATX FF). Paying twice the price just for the privilege of using Milan didn’t sit right with me when the board I already had/have works fine for all my other connectivity needs.

I’m not surprised by this behavior from SM, but kind of am from AsRock since they generally have been more consumer friendly with not gatekeeping features behind paywalls and licenses (like OOB management features) and planned obsolescence.

but now at least anyone who has an EPYCD8 can be vindicated and finally use Milan thanks to this mistake

 

[Advin]

Lots of things are unfair. Tesla and BMW charge for rear heated seats, even though they are already integrated into the car. Tesla charges extra for FSD, even though it's already integrated into the car. Intel could've supported ECC long beforehand on their Intel Core lineup's, but chose to intentionally gatekeep it.

We're seeing the same here.

 

[gsrcrxsi]

Lots of things are unfair. Tesla and BMW charge for rear heated seats, even though they are already integrated into the car. Tesla charges extra for FSD, even though it's already integrated into the car. Intel could've supported ECC long beforehand on their Intel Core lineup's, but chose to intentionally gatekeep it.

We're seeing the same here.

Everything you mentioned is 1st party decisions of products they directly control.

what happened here was several 3rd party manufacturers collectively deciding to go against AMD’s plan for the platform. there is a difference, even if it seems similar.

But if anyone wants this BIOS and not have to upgrade their perfectly good EPYCD8 board, they can find the PM button

 

[Jorge Perez]

I mean, there's nothing stopping you from making your own bios?

 

[gsrcrxsi]

I mean, there's nothing stopping you from making your own bios?

I think if it were trivial, someone would have done it by now.

but at the same time, no need now that AsRock already did it and leaked it by accident. We finally have what they should have given in the first place.

 

[Advin]

My other guess would be that maybe they just don't want to provide support services for it or maybe they don't want to spend the time to properly validate Milan CPU's for the BIOS (and redo their QVL), but just "unofficially" support it through these beta BIOS revisions (use at your own risk style).

It's the same with Gigabyte I believe, on the website they don't mention the Rev 1.0 boards supporting Milan, but they do have a BIOS for it.

It is a little weird though, most of the boards released for EPYC Naples only got updates for Rome support, and then when Milan rolled around only boards that released for EPYC Rome (sometimes) got BIOS updates for Milan.

 

[Advin]

Everything you mentioned is 1st party decisions of products they directly control.

what happened here was several 3rd party manufacturers collectively deciding to go against AMD’s plan for the platform. there is a difference, even if it seems similar.

But if anyone wants this BIOS and not have to upgrade their perfectly good EPYCD8 board, they can find the PM button

If it was AMD's plan, then AMD would force vendors to release BIOS revisions. But they don't really care

 

[RolloZ170]

Tesla and BMW charge for rear heated seats, even though they are already integrated into the car.

and I pay for eggs, but they are already laid !

 

[bayleyw]

I wonder how stable it is. The 1st gen Epyc systems uniformly lack official Milan support, maybe there is some electrical gremlin that rears its head under heavy I/O load or something.

 

[gsrcrxsi]

I wonder how stable it is. The 1st gen Epyc systems uniformly lack official Milan support, maybe there is some electrical gremlin that rears its head under heavy I/O load or something.

I’ll let you know.

so far been running 24hr under full load on Power determinism, Seems perfectly stable. This is only a 200W CPU though and not very stressful. This is a beta “lab” BIOS however, denoted with the L prefix.

I might have concerns about a 280W CPU, as that might overwhelm the VRM. But my other boards with 240W CPU (on Power Determinism) and barely any VRM cooling have been running for months (even years) 24/7, so even if you have to limit to 240W, I’m sure it’s fine and is a better situation than not being able to run it at all.

it’s clear that they all lack Milan support as a “business decision” and not a technical one. The official AMD Daytona platform supported 1st gen EPYC all the way through Milan.

 

[gsrcrxsi]

so far so good. the BIOS seems to work with ES processors also. a "7V13 ES" (100-000000425_37/24_N) works fine also.