PrimeGrid
Please visit donation page to help the project cover running costs for this month

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Aggie The Pew message board (Message 111752)
Posted 2 days ago by Iain BethuneProject donor
I made a bit of a horlicks of this one. For some reason, BOINC wasn't running on my KL machine all weekend. Hey ho...
2) Message boards : Aggie The Pew message board (Message 111437)
Posted 8 days ago by Iain BethuneProject donor
I visited the Science Museum in London today. Among the exhibits were a Cray 1A which in the late 70s cost some millions of USD/GBP and a PDP8 which my academic department aspired to but could not afford.

Could someone put into context the power and capacity of these machines as compared with modern and contemporary devices.

I understand this is very much like asking what the value of GBP1 in 1700 is today.


In addition to what others already posted, there are two main metrics which are used to compare supercomputers:peak double-precision FLOPs (RPeak) which is the theoretical 'marketing' number taken by multiplying the number of cores, the clock speed and the number of FLOPs per cycle (i.e. the vector length); and measured FLOPS (RMax), typically measured by running a benchmark called LINPACK which does dense matrix inversion, and is typically about 80-90% of the peak (for CPUs), 50-70% for GPU-based systems or as low as 25% of peak for accelerator-based machines like Intel Xeon Phi. RPeak is supposed to representative of 'real-world application' performance. However, LINPACK is almost an ideal case for vectorisation and cache efficiency, so most real applications will go somewhat slower. For example, on my Haswell CPU, Genefer achieves 40% of peak FLOPs.

The ranking of the fastest computers in the world was updated yesterday at the Supercomputing conference in the US. See https://www.top500.org/list/2017/11/ for the full list. The fastest machine achieves has a peak of 125 PFLOPs but the list tails off quickly. The fastest in the UK is only 8 PFLOPs at number 15 on the list.

By comparison, a recent Intel processor (Haswell or later), achieves 16 FLOPs per clock cycle, so a quad-core running at 2 GHz has a peak of 16*4*2=128 GFLOPs - roughly a million times slower than the faster supercomputer in the world today.

A modern GPU like the GTX 1080 Ti that Mike posted, achieves 355 GFLOPs (which is only 1/32 the single-precision performance).

When the top500 was first published, in 1993, the fastest machine achieved RMax of ~60 GFLOPs (https://www.top500.org/resources/top-systems/cm-5-los-alamos-national-lab/) The peak performance is not stated, but roughly it looks like the equivalent of a modern-day i3 CPU. So it takes about 25 years for entry-level consumer kit to achieve the performance of the world's fastest computer.

Going back further to the Cray-1A (1982), it achieved only 160 MFLOPs! So roughly 1000x slower than the quad-core Haswell I quoted above.

Interestingly, there are some charts on the top500 site which indicate that Moore's law (exponential growth, doubling every 18-24 months), is alive and well over a period of 20+ years (https://www.top500.org/statistics/perfdevel/). It does appear to be slowing up a little, however. The first GFLOP machine was in 1988, the first TFLOP in 1998, the first PFLOP in 2008, but I would bet against there being an ExaFLOP machine by next year - but we will see!

- Iain
3) Message boards : Aggie The Pew message board (Message 111119)
Posted 20 days ago by Iain BethuneProject donor
Who would true valour see,
Let him come hither;
One here will constant be,
Come wind, come weather
There’s no discouragement
Shall make him once relent
His first avowed intent
To be a vermin.



Hobgoblin, nor foul Fiend,
Can daunt his Spirit:
He knows, he at the end,
Shall Life Inherit.
Then Fancies fly away,
He'l fear not what men say,
He'l labour Night and Day,
To be a vermin.



Whoso beset him round
With dismal stories
Do but themselves confound
No Cow can him fright
He'll with a German fight
But he'll have a right
To be a vermin.


Oddly enough, my mother spent much of her childhood under the mistaken impression that the last line of this hymn that was often sung in church was "To be a piglet". Or so she says!
4) Message boards : Generalized Fermat Prime Search : My little play with generalized fermat primes (Message 110917)
Posted 33 days ago by Iain BethuneProject donor
So the list of GFN n=13 primes is completed up to b=101M. 6223 primes are known, including 167 at very large b.

For b <= 101M, there are therefore 6056 known primes, which compares very well (< 1% error) to the estimate of 6012 from Dubner and Gallot's conjectured distribution

Onwards to 100M for n=14!

- Iain
5) Message boards : Generalized Fermat Prime Search : My little play with generalized fermat primes (Message 110647)
Posted 48 days ago by Iain BethuneProject donor
The list of GFN n=13 primes is now updated with September's finds. Look forward to seeing us finish up to 101M and make a start on n=14 again :)

- Iain
6) Message boards : Generalized Fermat Prime Search : GFN Prime Discoveries (Message 110642)
Posted 48 days ago by Iain BethuneProject donor
The list of GFN primes is now updated with all the finds at n=15,16,17,18 and the first n=20 prime!

Nice. Shouldn't the b value 919444 for n=20 be located inside the colored boxes? I guess it should be in the yellow box in this case. /JeppeSN


Thanks, fixed. HTML table trouble.

- Iain
7) Message boards : Generalized Fermat Prime Search : GFN Prime Discoveries (Message 110639)
Posted 49 days ago by Iain BethuneProject donor
The list of GFN primes is now updated with all the finds at n=15,16,17,18 and the first n=20 prime!

- Iain
8) Message boards : Generalized Fermat Prime Search : My little play with generalized fermat primes (Message 110234)
Posted 70 days ago by Iain BethuneProject donor
The list of GFN n=13 primes is now updated with the finds from August (up to b=80M).

Cheers

- Iain
9) Message boards : Aggie The Pew message board (Message 110220)
Posted 70 days ago by Iain BethuneProject donor
Heavy rain in an hour & 50mph winds overnight.

Hatton down the batches


Was it just me, or did 'Aileen' not amount to much? At least here in the NW it seemed to be not much wetter or windier than it has been for the last couple of weeks!

- Iain
10) Message boards : Aggie The Pew message board (Message 110198)
Posted 71 days ago by Iain BethuneProject donor
some humor


Are you sure, I thought everyone around here was pretty straight-laced?


Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2017 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.66, 1.63, 1.70
Generated 23 Nov 2017 | 7:34:28 UTC