It's almost like technology doesn't exist for the board to detect when a second CPU is present, and if so, switch some of the PCIe slots to use the lanes from that CPU instead. Since Supermicro apparently doesn't have access to this holy grail, they could have opted for a less advanced piece of manual technology known as "jumpers" and/or "DIP switches".
This incredible lack of basic functionality on SM's part, coupled with the lack of PCIe 4, makes this board DOA. Yeah, it's the only option if you want dual-socket EPYC, but it's not a good option by any stretch.
The complexity of using jumpers to reallocate entire PCIe lanes would be insane. You'd probably need a bridge chip to negotiate the transition, which would remove the need for jumpers anyway since it could be digitally enabled. But this would add latency - even if it wasn't in use since all lanes would need to be routed through it. Gone are the days of busmastering as everything is so complex now through serialization.
Have you got an example of a motherboard that implements your idea with PCIe? I've never seen it and as bryanlarsen said this type of layout where everything essential is connected to the 1st CPU is very standard in server and workstation boards. It allows the board to boot with just one CPU, adding the second CPU enables additional PCIe sockets usually.
At the very least they could have placed a bunch of M.2 connectors on the motherboard, even double stacked... or make a custom (dense) connector that would allow you to connect a pci-e x16 riser cable to a 4 x m.2 card.
you have to share with us lots of informative point here. I am totally agree with you what you said. I hope people will read this article. http://www.bestvpshostings.com/
This would not be the first time i've seen that. I remember playing with the first gen opterons, the nightmares of the pro-sumer motherboard design shortcomings were numerous.
1gb is pennies. 10gg costs a bit more. If you plan on using a different solution you have the option to get the cheaper board and install it. Save the 1gb for management duties or not at all.
Exactly. This way the user is free to choose from 10/25/100 GbE or even Infiniband or something more exotic if they wish. I would personally go for a 25 GbE card (about about $100 used).
There's one model with gigabit NICs, and one with 10 gigabit NICs. That covers what most people would want, and PCIe NICs for SPF+, and/or 25/40/100 gigabit covers most everyone else.
I can see this with the 1 gigabit NICs for monitoring/management and a 25 gigabit PCIe card for the VMs to use, for example.
I wish AMD would restructure their lineup a bit next gen.
- Their HEDT offerings are decently priced, but the boards are not. - All of the HEDT boards I’ve seen are gimmicky, not supporting features like ECC, and are focused on gaming and the like. - HEDT does not support a dual socket config, so you would naturally want to step up to EPYC. However, EPYC is honestly complete overkill, and the boards are typically cut down server variants. - For those that don’t need HEDT, but need more IO, they don’t have an offering at all.
I would love to see future iterations of Zen support an optional quad channel mode or higher, ECC standardized across the board (though if people realized how little ECC matters in modern systems...), and more PCIE lanes for everything.
All sTRX4 (Threadripper 3000 socket) boards listed on https://geizhals.at/?cat=mbstrx4 are listed as supporting ECC. Even a number of mainstream desktop boards (primarily from Asrock and ASUS) support ECC and we have built servers with them, and we have tested that ECC works.
Long time lurker here, love Anandtech. I had to register an account for this article in order to join the disappointment gang: there is a ridiculous low number of ATX formats options for AMD EPYC, and when talking Rome and PCI gen4, even lower. Found this the hard way: I hardly managed to get a Gigabyte MZ32-AR0 after months of waiting in vain for Supermicro to release some standard ATX/eATX with PCI gen4. Nowadays they seem to have some H12 models out, but of course those are nowhere to be found. If you want to buy one that is.
Problem is: now there is no new dual socket option! Exactly when I bought a 7402 Epyc, not a P part as usual... Supermicro is simply missing the server CPU party of the decade. We hope that the press can push things in the right direction, as it is not the time for big manufacturers to arbor the Intel fanboy flags. They did it for a very boringly long time, I just can't believe there isn't the slightest interest in the Rome platform. Because that's what it is: an astounding lack of interest and a very obtuse technology angle, that awful "partnership" inertia that plagues the server/workstations market. It seems that being future proof is a crime in this area.
It speaks for itself: we are reading here a review of a board that is 2 years old and got a revision half a year ago... Yeap, the cheapo BIOS revision, that one.
One thing about the article: the BIOS can't usually be updated via IPMI for Supermicro boards without a license. Only the firmware. Is this still the case or not?
Thank you Anandtech for reviewing enterprise, but standard formats (there shouldn't exist anything else, but that's just me)! Supermicro makes cool tech and I own a bunch, but sometimes, man...
PCIe Gen 4. Supporting the feature set of the expensive CPUs you plan on utilizing is a necessity. Support a competitor to SuperMicro to get the board manufacturers to provide more support, AMD. We all know competition engenders motivation to get off one's arse . . .
Even if all the slots were available - you still would not have a use for 128+ lanes. I know the big number is enticing - but in reality - like the 16C desktop CPUs - it's just a marketing gimmick. It's like having a car with 3000HP and it only gets driven in Manhattan - Cool that you have 3000HP, but in reality, not much use.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
36 Comments
Back to Article
bryanlarsen - Wednesday, May 13, 2020 - link
> the second CPU is underutilized.This is common in server boards. It means that if you don't populate the second CPU, most of your peripherals and slots are still usable.
The_Assimilator - Wednesday, May 13, 2020 - link
It's almost like technology doesn't exist for the board to detect when a second CPU is present, and if so, switch some of the PCIe slots to use the lanes from that CPU instead. Since Supermicro apparently doesn't have access to this holy grail, they could have opted for a less advanced piece of manual technology known as "jumpers" and/or "DIP switches".This incredible lack of basic functionality on SM's part, coupled with the lack of PCIe 4, makes this board DOA. Yeah, it's the only option if you want dual-socket EPYC, but it's not a good option by any stretch.
jeremyshaw - Wednesday, May 13, 2020 - link
For Epyc, the only gain of dual socket is more CPU threads/cores. If you wanted 128 PCIe 4.0 lanes, single socket Epyc can already deliver that.Samus - Thursday, May 14, 2020 - link
The complexity of using jumpers to reallocate entire PCIe lanes would be insane. You'd probably need a bridge chip to negotiate the transition, which would remove the need for jumpers anyway since it could be digitally enabled. But this would add latency - even if it wasn't in use since all lanes would need to be routed through it. Gone are the days of busmastering as everything is so complex now through serialization.bryanlarsen - Friday, May 15, 2020 - link
Jumpers and DIP switches turn into giant antennas at the 1GHz signalling rate of PCIe3.kingpotnoodle - Monday, May 18, 2020 - link
Have you got an example of a motherboard that implements your idea with PCIe? I've never seen it and as bryanlarsen said this type of layout where everything essential is connected to the 1st CPU is very standard in server and workstation boards. It allows the board to boot with just one CPU, adding the second CPU enables additional PCIe sockets usually.mariush - Wednesday, May 27, 2020 - link
At the very least they could have placed a bunch of M.2 connectors on the motherboard, even double stacked... or make a custom (dense) connector that would allow you to connect a pci-e x16 riser cable to a 4 x m.2 card.johnwick - Monday, June 8, 2020 - link
you have to share with us lots of informative point here. I am totally agree with you what you said. I hope people will read this article.http://www.bestvpshostings.com/
Pyxar - Wednesday, December 23, 2020 - link
This would not be the first time i've seen that. I remember playing with the first gen opterons, the nightmares of the pro-sumer motherboard design shortcomings were numerous.Sivar - Wednesday, May 13, 2020 - link
This is a great article from Ian as always. Quick correction though, second paragraph:"are fairly numerate". Not really numerate means.
1_rick - Wednesday, May 13, 2020 - link
Yeah, "numerous" was the correct word here.peevee - Thursday, May 14, 2020 - link
Nope. 1 is not numerous.heavysoil - Friday, May 15, 2020 - link
He's talking about the options for single socket, and lists three - numerous compared to the single available option for dual socket.Guspaz - Wednesday, May 13, 2020 - link
$600 enterprise board supporting up to 256 threads, and it's still just using one gigabit NICs?Sivar - Wednesday, May 13, 2020 - link
"Don't worry, widespread 10-gigabit is just around the corner." --2006Holliday75 - Wednesday, May 13, 2020 - link
1gb is pennies. 10gg costs a bit more. If you plan on using a different solution you have the option to get the cheaper board and install it. Save the 1gb for management duties or not at all.DigitalFreak - Wednesday, May 13, 2020 - link
Why waste the money on onboard 10 gig NICs when most buyers are going to throw in their own NIC anyway?AdditionalPylons - Thursday, May 14, 2020 - link
Exactly. This way the user is free to choose from 10/25/100 GbE or even Infiniband or something more exotic if they wish. I would personally go for a 25 GbE card (about about $100 used).heavysoil - Friday, May 15, 2020 - link
There's one model with gigabit NICs, and one with 10 gigabit NICs. That covers what most people would want, and PCIe NICs for SPF+, and/or 25/40/100 gigabit covers most everyone else.I can see this with the 1 gigabit NICs for monitoring/management and a 25 gigabit PCIe card for the VMs to use, for example.
eek2121 - Wednesday, May 13, 2020 - link
I wish AMD would restructure their lineup a bit next gen.- Their HEDT offerings are decently priced, but the boards are not.
- All of the HEDT boards I’ve seen are gimmicky, not supporting features like ECC, and are focused on gaming and the like.
- HEDT does not support a dual socket config, so you would naturally want to step up to EPYC. However, EPYC is honestly complete overkill, and the boards are typically cut down server variants.
- For those that don’t need HEDT, but need more IO, they don’t have an offering at all.
I would love to see future iterations of Zen support an optional quad channel mode or higher, ECC standardized across the board (though if people realized how little ECC matters in modern systems...), and more PCIE lanes for everything.
eek2121 - Wednesday, May 13, 2020 - link
A though I had: It would be nice if PCIE latency could be measured going forward.headeffects - Wednesday, May 13, 2020 - link
Can you explain a bit why ECC is less useful now than on the past? I’m curious.AntonErtl - Thursday, May 14, 2020 - link
All sTRX4 (Threadripper 3000 socket) boards listed on https://geizhals.at/?cat=mbstrx4 are listed as supporting ECC. Even a number of mainstream desktop boards (primarily from Asrock and ASUS) support ECC and we have built servers with them, and we have tested that ECC works.Micronsystems - Friday, May 15, 2020 - link
Much informative blog.Reach Micronsystems for the best<a href = "https://www.micronsystems.co.in/">desktop rental in chennai</a>
fakemoth - Friday, May 15, 2020 - link
Long time lurker here, love Anandtech. I had to register an account for this article in order to join the disappointment gang: there is a ridiculous low number of ATX formats options for AMD EPYC, and when talking Rome and PCI gen4, even lower. Found this the hard way: I hardly managed to get a Gigabyte MZ32-AR0 after months of waiting in vain for Supermicro to release some standard ATX/eATX with PCI gen4. Nowadays they seem to have some H12 models out, but of course those are nowhere to be found. If you want to buy one that is.Problem is: now there is no new dual socket option! Exactly when I bought a 7402 Epyc, not a P part as usual... Supermicro is simply missing the server CPU party of the decade. We hope that the press can push things in the right direction, as it is not the time for big manufacturers to arbor the Intel fanboy flags. They did it for a very boringly long time, I just can't believe there isn't the slightest interest in the Rome platform. Because that's what it is: an astounding lack of interest and a very obtuse technology angle, that awful "partnership" inertia that plagues the server/workstations market. It seems that being future proof is a crime in this area.
It speaks for itself: we are reading here a review of a board that is 2 years old and got a revision half a year ago... Yeap, the cheapo BIOS revision, that one.
One thing about the article: the BIOS can't usually be updated via IPMI for Supermicro boards without a license. Only the firmware. Is this still the case or not?
Thank you Anandtech for reviewing enterprise, but standard formats (there shouldn't exist anything else, but that's just me)! Supermicro makes cool tech and I own a bunch, but sometimes, man...
JustTheInductions - Friday, May 15, 2020 - link
PCIe Gen 4. Supporting the feature set of the expensive CPUs you plan on utilizing is a necessity. Support a competitor to SuperMicro to get the board manufacturers to provide more support, AMD. We all know competition engenders motivation to get off one's arse . . .Deicidium369 - Monday, May 18, 2020 - link
Which competitor is that?JustTheInductions - Friday, May 29, 2020 - link
Probably ASRock Rack.kwinz - Saturday, May 16, 2020 - link
E-ATX is just a painful form factor for a dual socket EPYC. Think of all the PCIe lanes that you can never use.Deicidium369 - Monday, May 18, 2020 - link
Even if all the slots were available - you still would not have a use for 128+ lanes. I know the big number is enticing - but in reality - like the 16C desktop CPUs - it's just a marketing gimmick. It's like having a car with 3000HP and it only gets driven in Manhattan - Cool that you have 3000HP, but in reality, not much use.johnwick - Monday, June 8, 2020 - link
I am glad to be here and read this wonderful article. all the info you provide to us, it was very informative to us! http://www.bestvpshostings.com/Pyxar - Wednesday, December 23, 2020 - link
Dear AMD, 1994 wants their comport and VGA connectors back..Thx, sincerely,
All the geeks in 2021.
boiledeggdiet - Friday, June 11, 2021 - link
Actually, I was searching for the best motherboard but I got little confuse here. https://fashionhealthbuzz.wordpress.com/2021/06/02...