With revenue down quarter after quarter, year over year, Power group running out of cash flow, has to eliminate the POWER core design team to survive. Chief Architects were let go during Thanksgiving holiday. Expect 60% - 80% cut in Core design team beginning next year.
26 replies (most recent on top)
Moving to another risc architecture like arm or risc v will require a large software migration effort and likely compiler changes. This is likely to be the real customer pain point.
Ibm will only likely move to new architecture if there is significant customer interest to develop it. Would likely need a hyperscaler partner. .. more than one.
Otherwise emulation is the cheapest last resort option to preserve the legacy Emulate on system z. emulate on xeon. Historically alpha pa risc sparc ended up being emulated Unisys also emulates legacy architectures. Apple switched macOS two times successfully with emulation.
It would be ibm ibm best interest to combine power with systemz assests, otherwise emulation or arm or risc v will be the road ahead.
Good luck!
Anything new on this, we know Ibm is aggressively going after competitor’s legally in Austin because of how many of the power engineers and programmers are leaving to companies like amd, nvidia, intel,Apple, Facebook and Google, best of the best have left or planning their exit
This is not good news for power. ISV’s are the life blood in the power division, and having large ISV’s get sucked up into the Intel world is just a death knell
https://finance.yahoo.com/m/5dd2e706-c5eb-3929-a334-90588eecca7f/cerner-stock-soars-as-oracle.html?siteid=yhoof2&yptr=yahoo
This reminds of DEC and Alpha revisited 20 years later.
@7xtu+1e8BVRj4 bruh don't fool around here.
All power architects who can create anything new are gone. If the picture is still not clear to you/anyone then you deserve to be out of the job.
Again, if you're in power, find a fu----g job right now!!
Powers future will be determined by 2 main factors
- ISV’s interest in future Power capabilities vs other chips capabilities (Power succeeds due to unique capabilities other chips can’t/will not deliver)
- IBM’s decision to invest in Power vs other opportunities (last quarter Power lost approx 200 million). The question is does Power generate enough intangible revenue (services/SW) to justify keeping it. Certainly Powers AI capabilities need to be taken into consideration in addition to the existing ISV install base
And we still need a fab to physically make Power 10 ... probably bean counters are gagging on fab contract prices with all the chip shortages and out-bidding by companies with deeper pockets. The ongoing Globalfoundries lawsuit is not going to solve this, and even if (years from now) it is determined to be GFS fault, Wall Street does not care - if we cannot deliver chips, we have lost more customers forever, and the execs won't get their bonuses unless they can lay off more of us.
The best and brightest from ibm power austin moved to amd, there is no way power 11 happens. Power10 is a train wreck, if power survives it will leave Austin and will be part of z in BOE and POK
It is a no-win situation for IBM, and it began way back when MS-DOS and PC clones happened .. just took a long time to fully play out.
I am not op but I can tell you that executives (4th line above) and distinguished engineers are actively discussing the options (including but not limited to ARM, AMD, intel) to license the cores. ARM seems the most viable but may not achieve the original Power performance. AMD and Intel can achieve or surpass the POWER core, but they are unlikely to license them. All three cores have a better software ecosystem and larger customer base than Power.
The points are:
cut the cost of power core design, replace it with cheaper (maybe) licensing
keep SoC team
leverage ARM and x86 software ecosystem and customer base
minimize the software team
Power is morphing towards a performance based solution only. (server, IO, and storage applications) The low end (think scale out) has dramatically moved towards Intel in the last few years due to ISV’s only wanting to support 1 code stack. LINUX SW development has only accelerated that migration, as Power Linux HW has been a very large disappointment. ISV’s have convinced their customers that “Intel is good enough” I do not see IBM moving towards an AMD or ARM solution due to a massive investment to support the code stack for a very small customer set, but rather IBM licensing/partnering a “high end” power solution when performance is the buying criteria, via an IP offering. Remember Power is one heck of a technology (especially P10), but the customer set “numbers wise” just isn’t there. IBM labs could offer up the performance design at a much smaller cost vs the current Power portfolio offering. The bean counters have come to the same conclusion, and the most probable solution is a NICHE High end Power cloud offering from one of the hyperscalers. Perhaps Kyndryl may be able to cut that deal, now that they are free to market outside of the traditional IBM walls.
It may be worth providing emulation of Power on X86 or System Z, as a transition means to protect IBM customers' software investments, sort of like what Apple is doing with Rosetta 2.
Or exactly what Apple did in 2006 with (the first) Rosetta when they switched from Power to x86.
Power has really been focussed on the high-power enterprise UNIX segment, it is unclear if trying to convert that over to ARM would be successful. The main problem for Power is that its market has shifted to more cloud based X86 server farms, as Intel and AMD compete more there with high-CPU count 7nm and next year 5nm chips, hard to compete (development costs are pretty high), IBM lacks a strategy to get the volumes. Power is also already weak on the I/O side of things, it got passed there by X86 some time back. ARM is generally more low-power focussed, which makes it good for things that run off of batteries like phones, tablets, laptops, some companies (like Apple) are investing in higher performance variants. ARM licensing costs are also not that cheap. There has been a shift from UNIX (AIX) type systems to Linux, IBM has been part of that with the Red Hat acquisition. It may be that IBM tries to curtail AIX investment (without Power, not much reason to do that anyway), shifting some of the headcount over to adding more Enterprise features to Red Hat, that then could be differentiating and produce more software revenue (which tends to be where the money is anyway), and it may be a way to help out with the Red Hat staffing some, lots of turnover. It may be worth providing emulation of Power on X86 or System Z, as a transition means to protect IBM customers' software investments, sort of like what Apple is doing with Rosetta 2. Same may happen with Power hardware development, shifting some of the headcount to System Z (lots of turnover and job openings there, and Z) or AI processors or Cloud Processors (though there, we pretty clearly would need a big partner). In some ways, the headcount and skills on Power and AIX is worth more elsewhere, hopefully IBM can find a way to use it to fix gaps and bolster the remaining efforts.
The partner program is a joke right now.
It’s all about Z going forward. Power will rely on the partner ecosystem
https://finance.yahoo.com/news/ibm-launches-ibm-z-cloud-130000060.html
Shagadelic baby
architects were let go during past month
That's OK, there are plenty of new collar and offshore replacements.
Power 11 is going to be great!
This is a speculation of Powers revenue listed out by quarter. Note they also include Power in storage. Both are shrinking quite aggressively. I suspect AK can read a graph and make a business decision. The trend is not Powers or storages friend given that technology changed yet the graphs kept shrinking. Even though margins on Power are good, the volumes doesn’t justify the investment
https://www.itjungle.com/2021/10/25/a-proper-accounting-of-the-power-business/
I am not op but i can tell you that architects were let go during past month, saw that in my network.
so that part is correct.
Not sure about arm licensing (which ibm might be saying to keep people around until they decide to lay them off.
If you're in power, you should have already been looking for a job! The writing was on the wall long ago.
Fake or real news ? The devil is in the details
but if IBM throws in enough incentives .. perk a potential partners ears up
IBM is fill of promises that never go as planned. A wise partner will simply stay away. Sadly, Power is dead.
This is a big deal if true and we all know items posted here are a mix of reality and fake news,
can you provide additional comments and proof on the architects let go and new direction for power, there are a lot of skilled people that need to start looking if this is true and no one wants to be surprised and scrambling for a job
Switching to AMR or AMD is quite the heavy lift. I can’t see IBM risking the ISV install base. Instead IBM will partner off Power. Yes there is that pesky issue of “who wants it”, but if IBM throws in enough incentives (research for P11, a guarantee of using P10 in their cloud, a customer list of p9 and before, a built in channel, a Fully functional TSS team, and finally a commitment to only play in the enterprise space) That would perk a potential partners ears up
Wait, what do you mean by switching to arm licensing?
So they'll still make the power boxes but use Arm's ip?
IBM doesn’t want to be in the HW business outside of “enterprise”. They abandoned Intel, thinking Power would pick up the slack. How did that work out, given that Intel was a commodity? (10-25% shrink per year) Power is hemorrhaging money, and most of the legacy IBM talent has moved to AMD. IBM’s only interest in Power is the large ISV’s (SAP, Oracle, Epic, etc etc), think Enterprise. Expect IBM to partner off Power and the channel to stop the bleeding. Ginni thought Power was far more valuable than it was due to its superior AI and memory capabilities. Guess what Intel caught up and Power stagnated due to lack of investment. That’s not to say Power isn’t a superior product, it is. The fundamental problem is there are not enough customers who need the performance to make the investment in Power worth it. IBM was hoping mainframe and storage would spur demand for Power chips, but that has not come to pass. IBM will farm the IP via a partnership sale in Power, and exit stage right. The bean counters have already run the numbers and AK doesn’t need another mill stone legacy investment when Power just doesn’t bring in that many customers. He would rather pay double for the chip sets he needs, farm the legacy IP including OS’s, dump the channel, and partner off the non-strategic HW. It’s a global foundries replay but at a much smaller scale.
I don't know if it's true, but it would make a lot of business sense. (Come of think of it, that's a good reason to not believe it!).
The most conservative move for IBM right now would be ARM licensing - it would still be a follower not a leader, but then so are many of its customers. Second would be to roll the dice on RISC-V - a bit more leaderish, and a differentiator against the cloud heavyweights like AWS who are investing in ARM and will rapidly commoditize it.
Third, and by far the boldest, would be to try to acquire ARM as the Nvidia bid is struggling with regulators. Possibly in concert with a separation of IBM into software and hardware companies. No, I don't think that's very likely; but it's the kind of radical move required if IBM wants to be on any path other than a slow ride off into the sunset.