Power11 hits the market this month
It's notable that the "meet the family" document IBM links from the press release — so we can assume it's officially blessed — says nothing about OMI, only DDR5 RAM. However, IBM has made it clear that Power11 will continue to have OMI, since enterprise Power10 customers would certainly have had an investment in it, and the upper tier datasheets reference OMI channel capacity. We don't know if the OMI firmware for Power11 is open and libre (it was not in Power10), nor if the Synopsys IP blocks reportedly used in Power10's I/O are still present, because IBM didn't say, or if the "low-end" binned CPUs are different in this regard.
If there are going to be third-party Power11 systems, and IBM didn't say anything in the press release about them either, they generally follow six to twelve months after. We have heard little from Raptor since about the SolidSilicon S1 and X1, and because all indications suggest the S1 is a Power10 implementation without the crap, this already puts them behind the curve. That said, adapting Power11 should be possible to any next-generation Power ISA workstation: the Talos II and T2 Lite are fairly straightforward reworks of the reference POWER9 Romulus design, and Blackbird is still Romulus, just in a much smaller form factor. These first-generation P11 boxes, as presumably performant as they are, wouldn't be nice to have in an office and IBM just doesn't do end user sales. Creating a T3 based on Blueridge would seem to be the best way forward for Raptor to regain the top slot in OpenPOWER workstations — assuming the architecture is still open.
Honestly, RCS has been very quiet as of late, and I'm getting worried that Power11 might be another missed opportunity for them and IBM. Plus, with post COVID price spikes on all their hardware, it's going to be difficult I think for Raptor to get something new out the door that will sell anywhere close to decently well. Seeing the prices on Talos and Blackbird bundles in more recent times made my eyes water.
ReplyDeleteI really want Power11 to succeed, but if it's more of the same compared to Power10, my interest in the platform will have gone down even further, which is unfortunate. That said, I appreciate you keeping up with all these developments Cameron, as you're kind of our champion for the IBM Power platform cause.
I'd like to buy a Raptor Power11 too. I hope they produce one because right now the libre alternatives are any number of terribly underpowered RISC-V boards. I'm toying with an Ampere Altra but I'd much rather stay with OpenPOWER.
DeleteAs a side bar... given the current climate of you-know-what, I could see Raptor really lean into the "Made in America" slogan and it legitimately mean something here for a change. I'm hoping Tim has an ace up his sleeve. :)
Delete@ClassicHasClass To be honest, those RISC-V systems aren't any more open than POWER10 was. Many of them use the same DDR4 IP block with the same firmware. Likewise, Ampere has a rather large closed-source blob that's compiled into their BIOS equivalent which does chipset init and ram training, in the style of Intel's FSP binaries. I suppose hypothetically if coreboot ever gets ported to Ampere they might make an attempt to replace it some day, but I'd wish for a million wishes first. Right now POWER9 is still the only open DDR4 training implementation afaik.
DeleteDoing some guesswork based on ServeTheHome's article and the pictures they include, it looks like the way these systems work is they have an OMI interface coming off the CPUs to very POWER10-style boards, with a central OMI-to-Differential-DDR5 bridge chip surrounded by Differential-DDR5 chips. Hypothetically, assuming there's no use of the same locked down PCIe block in POWER11, then the moment ANYONE makes an open OMI-to-anything bridge chip then a POWER11 Raptor system should in principle become possible.
DeleteIt would also be hypothetically possible for Raptor to reverse engineer the firmware in any of the bridge chips and reimplement it, and OMI compatibility may mean that doing so with ANY of the bridge chips (DDR4 or D-DDR5) would unlock POWER11 in some capacity. Maybe they could also implement their own bridge chip on a very expensive and high-end FPGA. Such an FPGA would almost certainly require closed source tools given the maturity of the tools used for Arctic Tern, but maybe that is an acceptable compromize (I doubt kicad has been mature enough to support design of their existing boards). This would probably require a similar reverse engineering effort for the FPGA's DDR4/5 controller. Theoretically this could all have been done any time in the last 5 years on POWER10, though perhaps the other issues with POWER10's PCIe prevented it from being worthwhile.
Addendum: I can't say FOR SURE though that the CPUs don't have onboard D-DDR5 itself, and IBM don't have an open implementation of it (in the style of POWER9), but nothing I've seen suggests that.
DeleteSince Power 11 is still on a 7nm Samsung node--though improved--this is hardly a generational leap for IBM. As a result if Solid Silicon ever produces a workstation-class processor, then it may not be so far behind. On the other hand, it seems unlikely to me that such a processor will ever appear.
ReplyDeleteMy opinion is the health of the whole IBM Power ecosystem depends on having cost-effective workstations available to independent developers and small software companies. I'm just a random voice on the internet but am still hoping someone listens.