Hacker News Reader

Comment by DigitalJack | Hacker News Reader

They must not be expecting to produce these at scale if they are loading them with FPGAs. Or the markup is enormous. Being IBM, the latter might be it.

edit: looking at the picture, they are XC7K410T's. 500,000 flops... almost 2000 multipliers. Looking at the specs on FPGAs these days makes me giddy.

http://www.xilinx.com/support/documentation/data_sheets/ds18...

View on HN

Your comment made me curious, so I went looking and found a price sheet. I see two SKU's, one at $32K and one at $39K, I think probably this is just for the chassis. The storage modules list at $22K, $46K and $92K for the 1.2TB, 2.9TB and 5.7TB models, respectively.

I don't think too many people really end up paying list, but I have to imagine the markups are very healthy.

This thing looks rad af. I work exclusively with IAAS stuff now, but in a past life I used to spend a lot of time with enterprise hardware like this. I bet those modules feel real good when you snap em in.

http://www-01.ibm.com/common/ssi/rep_ca/3/897/ENUS115-003/EN...

The SM950 can be had at $300 for 512GB with 2.5/1.5 GB/s sequential read/write and 300K IOPS. RAID together 6 of them in parallel and you'll have something with better specs and storage than the 2.9TB module ($46K) for just $1.8K (albeit with less endurance, though not sure how much less).

IBM's margins are heavenly, no argument here. Yet what you propose is apple to oranges. The SM950 would have endurance of 1K to may be 3K writes. The IBM's flash is 10K. This is enterprise/database, not desktop, ie. you need the write endurance. It is one of the main reasons why we don't see many database servers with SSD arrays like you proposed, and instead we see huge RAM servers with RAID-ed HDDs.

That is bad engineering, HDDs have about the same drive writes per day as SDDs once you get into the TB sized SSDs.

Reply on news.ycombinator.com