Cryptocurrencies, fad or future?

Nov 14, 2018 49 Replies

Martin Brown wrote in news:psu1rq$1flg$1 @gioia.aioe.org:

The cell processor was a very good idea and Sony and the other players should have continued with it. It had one core specifically set to that task, in hardware, so no user choice mattered. Other than being clocked too slow, it was a very robust early multi-core processor. A lot of home/college brew mini-super clusters were made from playstations, as it was good for that too.

A nice PPC RISC core based multi-core player to compete with all this x86 CISC stuff would be cool.

OK, it is not cheap, but it is not /that/ expensive. And it is Power9, rather than PowerPC. Apparently it is particularly good at bitcoin calculations.

Well, a colleague of mine wrote some software that buys and sells BitCoin riding the wave of crypto-currencies. He's absolutely convinced that he'll make his first million in just under three years.

He's 27, so half my age. I just politely nodded my head (thinking to myself, what if he's right?!)

David Brown wrote in news:psu94r$610$ snipped-for-privacy@dont-email.me:

Not cheap at all. And the mobo has not much local modern power management on it, seemingly. For that price. All because they tout PCIe 4 apparently.

Yes, it is overtly too high a price point.

I am quite sure that IBM has made great strides in the last two decades, even if less than in the spotlight.

Those RISC architecture processors are very good at such things, and do so with a bit less overhead typically.

However, the difference is little these days, it seems. A good example is say the battle for the hill on the difference between Direct

3D and OpenGL on Windows PCs. On my Quadro GPU, it is likely a bit faster to go native OpenGL, but in the Xeon CPU integrated Intel GPU (also in this machine), it likely 'embraces' the Windows D3D accelleration 'elements' a bit better.

I think all GPUs are a very reduced instruction set devices. But I haven't been keeping up so maybe not any more.

It is usually compared to Xeons and other server systems, and paired with large amounts of memory - then it is cheap in comparison.

They have. For some purposes at least, the Power9 is the fastest chip around.

GPUs are a very different architecture. While you can do some tasks on either a GPU or a CPU, you can't compare them in general.

Nvidia stock has fallen from $280 to $154 in two months.

Watch ebay for pre-owned Nvidia gear.

John Larkin Highland Technology, Inc lunatic fringe electronics

LT saves all the voltages and currents to disk by default. You can really speed up some sims by selectively logging to the .raw file.

John Larkin Highland Technology, Inc lunatic fringe electronics

John Larkin wrote in news: snipped-for-privacy@4ax.com:

bitcoi

Their slack will get taken up by our global participation in the supercomputer race, which we currently lead with their help.

So, try again, future boy.

Shiny supercomputers generally aren't made of stuff from eBay. ;)

I'm a bit of an exception myself--we have about about 1 TFLOPS aggregate here, 75% of which is from eBay. There was a time when that would have been quite impressive.

Cheers

Phil Hobbs

Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com

Phil Hobbs wrote in news:psve6h $4oj$ snipped-for-privacy@dont-email.me:

I think my Xeon / Quadro laptop does that.

It certainly outperforms the first few Cray series machines.

Lenovo doesn't mess around when you pipe out the bucks.

Phil Hobbs wrote in news:psve6h $4oj$ snipped-for-privacy@dont-email.me:

Nvidia is not an ebay company, dingbat.

They make GPUs and plenty of products used by the makers of supercomputers to accellerate the activity of an individual node.

Multiply by several hundred or several thousand and you get a supercomputer. There are very few that do not use GPUs to speed certain processes in each node they integrate into the cluster.

So, yeah... they ARE made up of COTS devices. The XEONs are available to you and me, just like they are them. You and me simply do not have the $22k+ per processor.

Machine Learning will take up any slack in the GPU market. These days, an ML workstation that doesn't have $10K in GPUs is not considered a contender.

There are plenty of ML wannabes who would build one from EBay parts.

Clifford Heath.

Clifford Heath wrote in news:8%HID.166453$Ol2.131039 @fx17.iad:

Hardly.

I thought China was the world winner on supercomputers.

NT

Do they have their version of Deep Thought up and running as yet? I believe they are still working on it. No?

Rick C.

Tesla referral code -

formatting link

I do not know how to fix this in pan.

But here is another article...

I recall a Computer Science professor who worked with similar style machines back in the 70's. He had one with a lot of Z80s called Zmob. He would talk about them being "city dimmers" as a euphemism. I guess they really are that now.

Rick C.

Tesla referral code --

formatting link

Or putting the log file on a RAID0 striped array of identical SSDs chosen for fastest possible writes - certainly not on spinning rust.

Regards, Martin Brown

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required