Forum: GpuOwl
2019-01-09, 00:01
|
|
Replies: 2,760
Views: 367,061
|
|
Forum: GPU Computing
2018-06-05, 15:06
|
|
Replies: 33
Views: 10,505
I’m still interested in this for FPGAs
I...
I’m still interested in this for FPGAs
I happen to have ended up with exceptional access to a very large number of High end Xilinx FPGAs, and I would be more than happy to provide one or two on...
|
|
Forum: Miscellaneous Math
2018-05-28, 15:38
|
|
Replies: 3
Views: 593
|
|
Forum: GPU Computing
2018-05-23, 04:11
|
|
Replies: 7
Views: 3,673
|
|
Forum: Hardware
2018-04-21, 18:59
|
|
Replies: 77
Views: 13,519
|
|
Forum: Hardware
2018-04-21, 13:22
|
|
Replies: 77
Views: 13,519
You are correct that the routing kills...
You are correct that the routing kills utilization. In smaller chips it can be as bad as 50% due to routing overhead.
The LUT counts actually came from estimating/interpolating off of around page...
|
|
Forum: Hardware
2018-04-21, 03:10
|
|
Replies: 77
Views: 13,519
Looking at the TF side...
I had some down time tonight and started looking realistically at what kind of trial factoring performance I could expect get from some of the FPGAs I have in the lab. I want to make sure my starting...
|
|
Forum: Linux
2018-04-11, 13:44
|
|
Replies: 4
Views: 4,709
|
|
Forum: Linux
2018-04-11, 02:46
|
|
Replies: 4
Views: 4,709
|
|
Forum: GPU Computing
2018-04-03, 18:44
|
|
Replies: 32
Views: 7,648
1. Was my biggest concern. Regarding 3, at a...
1. Was my biggest concern. Regarding 3, at a large enough scale don't statistics take care of this?
I don't think you can prevent pre-factoring, you just have to ensure that you would have to...
|
|
Forum: GPU Computing
2018-04-03, 01:18
|
|
Replies: 32
Views: 7,648
|
|
Forum: GPU Computing
2018-04-01, 12:41
|
|
Replies: 32
Views: 7,648
|
|
Forum: GPU Computing
2018-04-01, 00:51
|
|
Replies: 32
Views: 7,648
Pump up GIMPS power? (Wishful thinking)
https://github.com/ethereum/EIPs/issues/958#comment-377734068
The cryptocurrency space is going through a bit of a crisis, maybe we can convince them that elliptic curve factoring candidates for...
|
|
Forum: Hardware
2018-03-30, 03:20
|
|
Replies: 52
Views: 7,621
I’ve always liked the idea that the only reason...
I’ve always liked the idea that the only reason prime patterns seem so random to us is our obsession with representing numbers with truly arbitrary bases.
I propose a base of prime factorizations,...
|
|
Forum: Hardware
2018-03-25, 01:26
|
|
Replies: 52
Views: 7,621
|
|
Forum: Hardware
2018-03-24, 14:03
|
|
Replies: 52
Views: 7,621
|
|
Forum: Hardware
2018-03-23, 05:54
|
|
Replies: 52
Views: 7,621
I’m totally willing to agree that it is likely...
I’m totally willing to agree that it is likely any reasonably sized NN probably can’t predict Mersenne sized primes, but I could probably construct a proof that an impossible one trained perfectly ...
|
|
Forum: Hardware
2018-03-23, 05:23
|
|
Replies: 52
Views: 7,621
I might say “of unknown value”, but unless you...
I might say “of unknown value”, but unless you know of some proof of truly random distribution of primes I don’t know about, I don’t think you can say for certain that a sufficiently large neural...
|
|
Forum: GPU Computing
2018-03-23, 04:59
|
|
Replies: 28
Views: 5,772
|
|
Forum: Hardware
2018-03-23, 04:53
|
|
Replies: 52
Views: 7,621
|
|
Forum: Hardware
2018-03-23, 04:03
|
|
Replies: 52
Views: 7,621
I suppose I am searching for a proof that a...
I suppose I am searching for a proof that a neural network of X complexity (node count) and Y bits of precision per node can effectively model any problem that can be expressed in Z gates of a Base-Y...
|
|
Forum: Hardware
2018-03-23, 03:22
|
|
Replies: 52
Views: 7,621
|
|
Forum: Hardware
2018-03-22, 04:34
|
|
Replies: 52
Views: 7,621
I should be clear I only mentioned this as a...
I should be clear I only mentioned this as a thought exercise not a serious pursuit.
That said I also figured you’d try to build it as a general prime generator. Let’s say we give it 128 bits of...
|
|
Forum: Hardware
2018-03-22, 02:56
|
|
Replies: 52
Views: 7,621
The concept was that running prime95 on an...
The concept was that running prime95 on an exponent is essentially a black box that takes < 32 bits as its input and outputs a Boolean true or false. You could generate a theoretical Boolean circuit...
|
|
Forum: GPU Computing
2018-03-22, 02:38
|
|
Replies: 33
Views: 10,505
You’re right - I was only accounting for one...
You’re right - I was only accounting for one pass. Of course I’m not sure how all the constants translate into integer based calculations, but constant in hardware are significantly faster than...
|
