IBM unveils their Power9 chip

[SpaceBeer]

Since 14nm HP was part of the IBM/GF deal it s likely that IBM agreement is necessary if they want to use it for another customer, and i m not sure that IBM would agree to give some help to the concurrent ISA, be it AMD..

It is stated/rumoured Zen+ will go to 14nm HP, Zen will sure stay on 14nm LPP. I doubt difference between HP and LPP is that big, and I think it won't be too hard to switch from one to other

 

[Arachnotronic]

It is stated/rumoured Zen+ will go to 14nm HP, Zen will sure stay on 14nm LPP. I doubt difference between HP and LPP is that big, and I think it won't be too hard to switch from one to other

Zen+ won't use the IBM 14nm HP process.

 

[superstition]

Zen+ won't use the IBM 14nm HP process.

AMD said this?

 

[Arachnotronic]

AMD said this?

Just watch, IBM's processes are very complex, very specialized, and usually have poor yields. They're suitable for big expensive chips like POWER9 that IBM sells wrapped in a multi-thousand dollar server, but if you're AMD trying to sell chips in the consumer market and in the bulk of the server market, it's just not a practical process. There's a reason that GloFo went begging to Samsung to license 14LPP rather than use IBM's 14HP.

 

[The Stilt]

Personally I find TSMC 16nm FF+ to be much more likely as the process for Zen+ than IBM 14nm HP.

 

[superstition]

If AMD is selling Zen+ and Zen simultaneously it could make sense to use the better process for Zen+. Yields and a lack of lowered pricing from competition/volume (e.g. just one supplier capable of doing the microbumps) didn't stop them from adopting HBM1.

 

[DrMrLordX]

https://www.computerbase.de/2016-08/ibm-power9-prozessor/

The chip is manufactured at GF....

Of course. GF acquired all those IBM fabs, after all. And the engineers to boot.

According to BitsAndChips all of them are produced in the same GloFo fab nr 8.

Doesn't make them the same process though.

No idea.
Wouldn't matter much for existing designs anyway, since they are all designed for 14nm LPP

Zen+ maybe?

Just watch, IBM's processes are very complex, very specialized, and usually have poor yields. They're suitable for big expensive chips like POWER9 that IBM sells wrapped in a multi-thousand dollar server, but if you're AMD trying to sell chips in the consumer market and in the bulk of the server market, it's just not a practical process. There's a reason that GloFo went begging to Samsung to license 14LPP rather than use IBM's 14HP.

Eh, not too long ago, the word was that 22nm FDSOI was already cheaper per transistor than 14nm LPP. Granted it's been in use since at least 2014 when POWER8 launched, but it's not like 14nm HP is going to have yield problems forever.

 

[NostaSeronx]

Eh, not too long ago, the word was that 22nm FDSOI was already cheaper per transistor than 14nm LPP. Granted it's been in use since at least 2014 when POWER8 launched, but it's not like 14nm HP is going to have yield problems forever.

FDSOI and SOI FinFETs don't have "yield problems." Especially, if the foundry has done a bulk version of that node.

Samsung has given us a slide for this;

Because GlobalFoundries has done 20LPM & 14LPP. Thus, we should see the above with 22FDX & 14HP.

Stuff I have found;
22FDX volume starts March 2017. [1.0 PDK and production*]
14HP volume states July 2017. [1.0 PDK and production*]
*For the past six months.

 

[DrMrLordX]

All nodes have a breaking in period. Arachnotronic seemed to imply that 14nm HP is going to be yield hell for awhile, but based at least on the last boutique node used for a POWER chip, I don't think that's the case.

 

[Phynaz]

Just so you know i spent the last year working in an environment with more redhat on power and aix on power then x86. They have power8, as well as z13. The power decision was purely political the actual enterprise architects wanted to move it all to redhat on x86.

IBM does a good job of getting open *ix stuff cross compiled and it mostly works. But anything Microsoft (obviously) , any cloud services automation (azure stack, Vrealize, etc), there are lots of issues with middle-ware even for stuff that is well supported on power like Hana. Your standard document archive solutions (objective, trim, etc), unified communications,etc the list just goes on and on, if it wasn't midnight i might indulge you more.

Honestly in an average enterprise workplace how much as a percentage of x86 do you think you could drop in replace with power? Now how much of a standard enterprise could you drop in replace with Zen?

Our big databases run on either Power or Z. Our big middleware runs on Power, while small middleware runs on x86. Our Hadoop runs on Power. Actually all our HA platforms are Power based.

 

[Lepton87]

Our big databases run on either Power or Z. Our big middleware runs on Power, while small middleware runs on x86. Our Hadoop runs on Power. Actually all our HA platforms are Power based.

How does the Z13 compare to the Power8? I couldn't find any direct comparisons and I'm really curious.

 

[thecoolnessrune]

Our big databases run on either Power or Z. Our big middleware runs on Power, while small middleware runs on x86. Our Hadoop runs on Power. Actually all our HA platforms are Power based.

It's too bad that PowerHA licensing is so expensive, because everything else is so darn bad. HA in AIX and especially IBM i when not using SystemMirror is horridly messy, and far too hands-on to maintain. Cisco UC HA is also terrible, but it's great compared to Maxama, Vision Solutions (ugh), RSF, and those other non-IBM HA Solutions.

I think part of that is just a consequence of building an architecture designed to be compatible with 30 year old code.