arnon

25 ms·

331.
by arnon 8y ago
I'm just happy the gorgeous building is being used again, not being demolished.
332.
by arnon 8y ago
That's exactly the thing. There is no data to the contrary. You can't prove that the FAA is wrong, because we don't know why both planes crashed. We have assumptions, and there's a public perception - but it may not be c
333.
by arnon 8y ago
I don't like the PaxEx in the 787. It feels narrower and more cramped. A380 -> A330/A340 (on account of the 2-4-2 arrangement) -> 767 -> 777 for me.
334.
by arnon 8y ago
In American carriers, "Economy Plus" or similar is just a little bit of more leg room, in the standard 3-4-3 or 3-3-3 configuration. For companies like Lufthansa, SAS and others - premium economy is closer to what Business Class u
335.
by arnon 8y ago
It's interesting that it was actually AMD that kept the Intel x86-64 architecture alive. Intel knew that the x86 architecture was limited in time, and tried to kill it off with the the 64-bit Itanium. AMD had a different plan, and rele
336.
by arnon 8y ago
NVLink, which works between IBM Power CPUs and NVIDIA GPUs is already 9.5x faster than PCIe 3.0... Not all queries or database operations are bottlenecked by PCIe. Some are compute bound, others are network bound, etc.
337.
by arnon 8y ago
SQream DB - around since 2010 MapD (now OmniSci) - around since 2013 Kinetica - around since 2009 ...
338.
by arnon 8y ago
It's a bit of a bummer they didn't choose to use a common language like some flavour of SQL, but opted to create another proprietary API
339.
by arnon 8y ago
Adding GPU support to an existing database is incredibly hard, and usually doesn't pay off very well. We were considering doing that some years ago, but opted for writing our own (GPU accelerated) database too. We talked about it recen
340.
by arnon 8y ago
At SQream (a GPU-accelerated data warehouse) we use the GPU for many operations, including sorting, aggregating, joining, transformations, projections, etc. We augment the GPU algorithms with external CPU algorithms when the GPU implementat
341.
by arnon 8y ago
Also worth noting this is a dataset that fits comfortably in-memory
342.
by arnon 8y ago
GPUs have a very high memory bandwidth, and can be used to perform memory-intensive operations (think decompression, for example). You can load compressed data up to the GPU, decompress it there, and run very complicated mathematical functi
343.
by arnon 8y ago
Yet another SQL-on-Hadoop product we'll have to deal with
344.
by arnon 8y ago
I took CSP as a basis for learning how distributed and concurrent protocols work, and then did Erlang to see that into reality. Great thing to learn
345.
by arnon 8y ago
RDMA may bypass the CPU but it doesn't bypass PCIe. It still goes through it, which is slower than NVLink on Power9 (~16GB/s on 16x PCIe 3.0 vs ~150GB/s on NVLink). Data has to be copied up to the GPU for processing. It has t
346.
by arnon 8y ago
NVLink between CPU and GPU is the key, and only IBM has implemented that with Power8 and Power9. Everything NVSwitch does is between GPUs, and doesn't really help with getting data up to the GPUs.
347.
by arnon 8y ago
This AMD card can compete with NVIDIA's high end Tesla V100 accelerator. At 7.4 TFlops of double-precision, it is smack in the middle between the PCIe version of the V100 at 7.0 and the NVLink version at 7.8. Memory bandwidth for the M
348.
by arnon 8y ago
Herein lies the problem. Notice you said "I believe that... but could be interpreted". I've worked with some copyright lawyers that basically said "it's impossible to know, and we'd rather just avoid the questi
349.
by arnon 8y ago
He seems a bit sour. Not sure why this statement is necessary to begin with
350.
Westinghouse Total Electric Home (1959)(youtube.com)
1 points by arnon 8y ago|0 comments
351.
by arnon 8y ago
There is no such plan that I'm aware of
352.
by arnon 8y ago
I've had larger excel sheets :P
353.
by arnon 8y ago
You're assuming the data is split among the K80 GPUs (12GB per GPU instance), which may not be the case here. Who's doing the split? How is the data partitioned along the GPUs?
354.
by arnon 8y ago
You're right, I got confused with another Postgres-GPU plugin type product. However, Postgres isn't columnar. How does PG-Strom (now HeteroDB) arrange data for the GPU properly?
355.
by arnon 8y ago
32GB is big?
356.
by arnon 8y ago
It is true though. A data lake is what companies end up with by dumping everything into HDFS, regardless of it's importance and relevance. So, it is just a hoarding database. I'll let the BlazingSQL guys answer the workloads quest
357.
by arnon 8y ago
NVIDIA invests in both OmniSci and BlazingDB so don't worry about it. They have it covered.
358.
by arnon 8y ago
It appears to be, more than anything else, some CUDA code wrapped in UDFs. Not sure I'd want that anywhere near my production stuff in that format.
359.
by arnon 8y ago
Can you explain this "industry" then? What industry? In-memory databases? GPU accelerated visualization? Which one are you referring to? Comparing MapD to SQream DB is akin to comparing VoltDB to Vertica. Disclaimer: I was intervi
360.
by arnon 8y ago
TPC focuses on very specific scenarios. That's why there's more than one. There are also a variety of scale factors. A database that performs well on TPC-H SF 100 will not necessarily perform well on TPC-H SF 10000. One that perfo

More ›