Comments on: With Cray Deal, HPE Finally Enters The HPC Big Leagues

By: kiers

kiers — Mon, 20 May 2019 15:52:56 +0000

I think this article is shilling a bit much for the deal. I don’t think these statements are entirely true:

“Hewlett Packard Enterprise has reached the top of the HPC sector”
“there has never been a better time to buy Cray”
“With those three big wins at US national labs, the payoff was visible on the horizon”
“Both HPE and Cray have been chasing the same markets”

By: luis river

luis river — Mon, 20 May 2019 10:30:38 +0000

I suspect that the Cray brand may live on in high-end machines, but perhaps not – and if that happens, that will be a pity. Cray means, and has always meant, supercomputing.

By: Steven Rapacki

Steven Rapacki — Sun, 19 May 2019 19:00:18 +0000

I believe the premium would have been higher had the stock price had some more time to reflect current government wins. I won’t be voting my shares for this sale at this time.

By: Paul Berry

Paul Berry — Fri, 17 May 2019 19:27:03 +0000

Doesn’t seem like a terribly good stock price premium, given all the good news Cray has been having lately.

By: Johannes

Johannes — Fri, 17 May 2019 18:43:03 +0000

It’s ironic that HP has to fight its way back into the HPC market after is dominated the Top500 list in 2002, 2003, 2007 and 2008. In the November 2003 list they held both place 2 (ASCI Q) and 5, while in the November 2008 list Cray already dominated the Top10 (5 out of 10) and HPE didn’t have a single Top 10 machine anymore.

By: Big_Fish_Wine_And_Dine_On_U_Sams_Dime

Big_Fish_Wine_And_Dine_On_U_Sams_Dime — Fri, 17 May 2019 17:38:34 +0000

I hope that this deal will not put AMD’s x86 based HPC/server offerings at a disadvantage over the longer run. So after this HP acquisition of Cray recieves it’s proper US Justice Dapartment review, the x86 market needs to be monitored even after approvial.

The entire PC/Laptop and HPC/Server systems market is dominated by the x86 ISA/ISA ecosystem to such a degree currently so any one x86 CPU maker’s market history with regards to market tactics and OEMs(big and small) needs to be kept in mind and continously monitored.

Also I do not see why HP/Others can not make use of some OpenPower offerings also as that’s even better for some non x86 based HPC options as well. ARM designs also but HP is already offering “Moonshot”/others options.

Maybe Cray can now get some cash infusion and continue to be in competition in that federal systems market as well as server/cloud systems market. AI is really starting to grow at a more rapid pace compared to the Traditional server market’s growth rate.

There are rumors that the Zen3 based Epyc/Milan will support SMT4 and that probably means that things are going to get even wider order superscalar from AMD on its Zen3 based server designs, if that is in fact the case about Epyc/Milan and SMT4. IBM’s power8/power9(SMT4 and SMT8 variants) for example has that SMT8 and really the widest order superscalar instruction issue rates and total numbers of execution ports as well. AMD must really be having to consider some L4 cache on its I/O die for Epyc/Milan also in order to support SMT4 if true.

So OpenPower’s/IBM’s centaur I/O dies each have 16 MiB of eDRAM to avoid having to suffer for the incresaed latency that comes with having to access System DRAM. The more levels of cache on a system the more chances to hide latency issues by having the code/data already residing on some cache level as opposed to being out on system DRAM and requiring the most added latency to get at.

The Exascale ERA is almost here!