Bitcoin Forum
March 28, 2024, 07:04:37 PM *
News: Latest Bitcoin Core release: 26.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 [18] 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 ... 71 »
  Print  
Author Topic: BitShares PTS (formerly ProtoShares) Mandatory Upgrade & Snapshot Announcement  (Read 218396 times)
This is a self-moderated topic. If you do not want to be moderated by the person who started this topic, create a new topic.
FreeTrade (OP)
Legendary
*
Offline Offline

Activity: 1428
Merit: 1030



View Profile
November 07, 2013, 06:51:18 AM
 #341

GPUs aren't limited to 1GB of memory, many have closer to 6 - 10GB now.

Yes - it's not about the amount of memory, GPU's can compare there with the current implementation.

The salient issue (unless I've misunderstood the algorithm) is that he asserts a CPU can use a hash table but a GPU can't, thus he asserts the GPU needs N ^ 2 memory. I don't see why the GPU can't use the same amount of memory as the CPU?

I think hashtables are a distraction from the point - which is . . .

The GPU can compute 1000s of lookups on that memory in parallel, thus obviating the random memory latency to be bound at the superior memory bandwidth of the GPU.

I'm not an expert in this area, but as far as I know the GPU has a wider bus, so can do more memory access (x4), but is slower so can do less (x0.25). So I think you could get comparable speeds out of a high end GPU as for a high end processor.

Membercoin - Layer 1 Coin used for the member.cash decentralized social network.
10% Interest On All Balances. Browser and Solo Mining. 100% Distributed to Users and Developers.
1711652677
Hero Member
*
Offline Offline

Posts: 1711652677

View Profile Personal Message (Offline)

Ignore
1711652677
Reply with quote  #2

1711652677
Report to moderator
1711652677
Hero Member
*
Offline Offline

Posts: 1711652677

View Profile Personal Message (Offline)

Ignore
1711652677
Reply with quote  #2

1711652677
Report to moderator
1711652677
Hero Member
*
Offline Offline

Posts: 1711652677

View Profile Personal Message (Offline)

Ignore
1711652677
Reply with quote  #2

1711652677
Report to moderator
BitcoinCleanup.com: Learn why Bitcoin isn't bad for the environment
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
1711652677
Hero Member
*
Offline Offline

Posts: 1711652677

View Profile Personal Message (Offline)

Ignore
1711652677
Reply with quote  #2

1711652677
Report to moderator
Luckybit
Hero Member
*****
Offline Offline

Activity: 714
Merit: 510



View Profile
November 07, 2013, 06:54:26 AM
 #342

I believe someone is already GPU mining it and planning to dump thousands of coins.
And your evidence for this is what?

People kept believing that Primecoin was being GPU mined, without any evidence. And still no-one has admitted to being able to mine it on GPUs, the two well publicised efforts both ending in failure.

Primecoin wasn't made by a corporate entity with funding. I'm not saying Bitshares is a bad idea, I like the idea. I'm saying this new algorithm seemed to be untested and their parameters I immediately knew were right.

But hey, if you manage to get some coins thats fantastic. I just don't see those coins going for a lot right now. I think the smart person will be the one who mines now and holds onto them until they convert them to Bitshares.

Do not sell a single coin. Each coin may be worth something decent but not for the next 6 months. Bitshares needs time to develop the decentralized exchange which will be powered by this coin.

AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 06:55:42 AM
 #343

The GPU can compute 1000s of lookups on that memory in parallel, thus obviating the random memory latency to be bound at the superior memory bandwidth of the GPU.

I'm not an expert in this area, but as far as I know the GPU has a wider bus, so can do more memory access (x4), but is slower so can do less (x0.25). So I think you could get comparable speeds out of a high end GPU as for a high end processor.

The limiting factor on the CPU is not memory bandwidth, rather memory latency which is less than 1 GB per second because the memory latency of main memory (if outside of L3) is several hundred clock cycles. With only 8 hardware hyperthreads, that latency isn't entirely masked away.

Whereas, the GPU can run 1000s of hardware threads (not software threads!) which masks away the latency and hits the memory bandwidth as the limit.

This is why the GPU blows away the CPU.

A hardware thread has its own copy of registers so there is nearly no cost to blocking the thread on memory access, so another thread can run which was blocked and is ready to proceed.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
testz
Legendary
*
Offline Offline

Activity: 1764
Merit: 1018


View Profile
November 07, 2013, 06:56:14 AM
 #344

Sorry, but I can't not to answer to @AnonyMint  Smiley:
If you experienced programmer you should know that "the devil's in the details" and you should trust only proven working solutions. If you don't have time, left it, someone will make the working GPU miner for this algo and get his a well-deserved award.

PS: If you so busy, probably you get salary and you don't need this reward. IMHO

            ▄▄████▄▄
        ▄▄██████████████▄▄
      ███████████████████████▄▄
      ▀▀█████████████████████████
██▄▄       ▀▀█████████████████████
██████▄▄        ▀█████████████████
███████████▄▄       ▀▀████████████
███████████████▄▄        ▀████████
████████████████████▄▄       ▀▀███
 ▀▀██████████████████████▄▄
     ▀▀██████████████████████▄▄
▄▄        ▀██████████████████████▄
████▄▄        ▀▀██████████████████
█████████▄▄        ▀▀█████████████
█████████████▄▄        ▀▀█████████
██████████████████▄▄        ▀▀████
▀██████████████████████▄▄
  ▀▀████████████████████████
      ▀▀█████████████████▀▀
           ▀▀███████▀▀



.SEMUX
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
  Semux uses .100% original codebase.
  Superfast with .30 seconds instant finality.
  Tested .5000 tx per block. on open network
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 06:59:10 AM
 #345

Sorry, but I can't not to answer to @AnonyMint  Smiley:
If you experienced programmer you should know that "the devil's in the details" and you should trust only proven working solutions. If you don't have time, left it, someone will make the working GPU miner for this algo and get his a well-deserved award.

PS: If you so busy, probably you get salary and you don't need this reward. IMHO

Someone can go implement what I have explained. If they want to send me a tip to be considerate that is good honor. Correct I don't need the money. I am looking to see the honor (or ethics) of the people involved.

Another factor is the requirements of the bounty. They don't allow for the GPU speedup I have explained. It isn't about reducing the memory on the GPU to less than that of the CPU, it is about running more threads on the same memory.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
testz
Legendary
*
Offline Offline

Activity: 1764
Merit: 1018


View Profile
November 07, 2013, 07:03:03 AM
 #346

Sorry, but I can't not to answer to @AnonyMint  Smiley:
If you experienced programmer you should know that "the devil's in the details" and you should trust only proven working solutions. If you don't have time, left it, someone will make the working GPU miner for this algo and get his a well-deserved award.

PS: If you so busy, probably you get salary and you don't need this reward. IMHO

Someone can go implement what I have explained. If they want to send me a tip to be considerate that is good honor. Correct I don't need the money. I am looking to see the honor (or ethics) of the people involved.

I'm sure that if some one will implement you idea, you will get the honor from bytemaster, but let's wait for proven solution.

            ▄▄████▄▄
        ▄▄██████████████▄▄
      ███████████████████████▄▄
      ▀▀█████████████████████████
██▄▄       ▀▀█████████████████████
██████▄▄        ▀█████████████████
███████████▄▄       ▀▀████████████
███████████████▄▄        ▀████████
████████████████████▄▄       ▀▀███
 ▀▀██████████████████████▄▄
     ▀▀██████████████████████▄▄
▄▄        ▀██████████████████████▄
████▄▄        ▀▀██████████████████
█████████▄▄        ▀▀█████████████
█████████████▄▄        ▀▀█████████
██████████████████▄▄        ▀▀████
▀██████████████████████▄▄
  ▀▀████████████████████████
      ▀▀█████████████████▀▀
           ▀▀███████▀▀



.SEMUX
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
  Semux uses .100% original codebase.
  Superfast with .30 seconds instant finality.
  Tested .5000 tx per block. on open network
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
█ █
Etlase2
Hero Member
*****
Offline Offline

Activity: 798
Merit: 1000


View Profile
November 07, 2013, 07:04:58 AM
 #347

The bounty isn't for whether or not GPUs can mine it, the bounty is for whether or not there is a non-linear speed increase available without the memory tradeoff. If GPUs end up being faster, big deal, it isn't going to be an order of magnitude faster unless you have an order of magnitude increase in the memory bandwidth or in the memory available. At least, that is the presumption by my understanding.

Luckybit
Hero Member
*****
Offline Offline

Activity: 714
Merit: 510



View Profile
November 07, 2013, 07:07:24 AM
 #348


I can't work with bytemaster because he and I don't agree. He wants to make a socialist coin where everyone receives their fair share just by sitting on their coins, and I would want to make a capitalist coin which doesn't steal (ahem redistribute) value to give to the collective of coin owners. The coin owners are already profiting from the rise in value of the coin.

I do talk with others behind the scenes.
Are you saying he doesn't believe in the survival of the fittest?
He doesn't believe in the market?

That doesn't seem to be what his documents say on his websites. What gives you the idea that its a socialist coin? I never heard of that before.
AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 07:14:05 AM
Last edit: November 07, 2013, 07:36:03 AM by AnonyMint
 #349

The bounty isn't for whether or not GPUs can mine it, the bounty is for whether or not there is a non-linear speed increase available without the memory tradeoff.

Correct the terms of the bounty disincentivized me.

If GPUs end up being faster, big deal, it isn't going to be an order of magnitude faster unless you have an order of magnitude increase in the memory bandwidth or in the memory available. At least, that is the presumption by my understanding.

You (and bytemaster) are apparently not factoring that the CPU does not hit its memory bandwidth (except in special situations such as using wide SIMD instructions in serial memory space which doesn't apply here because this is random lookup of saved values). Rather the CPU is limited by the order-of-magnitude slower random access memory latency, due to not having enough hardware threads.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 07:16:06 AM
Last edit: November 07, 2013, 07:46:33 AM by AnonyMint
 #350


I can't work with bytemaster because he and I don't agree. He wants to make a socialist coin where everyone receives their fair share just by sitting on their coins, and I would want to make a capitalist coin which doesn't steal (ahem redistribute) value to give to the collective of coin owners. The coin owners are already profiting from the rise in value of the coin.

I do talk with others behind the scenes.
Are you saying he doesn't believe in the survival of the fittest?
He doesn't believe in the market?

That doesn't seem to be what his documents say on his websites. What gives you the idea that its a socialist coin? I never heard of that before.

The public discussions he and I had (in his bitcointalk threads) about how everyone gets what is akin to an interest payment, and how everyone who owns the domain name coin gets a payment. He claims to be Austrian economics based, yet apparently he and I disagree about what that entails. I'd rather not rehash that here. The prior discussions are available. My summary is inaccurate I am sure. Better to go read the prior discussions.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
FreeTrade (OP)
Legendary
*
Offline Offline

Activity: 1428
Merit: 1030



View Profile
November 07, 2013, 07:21:28 AM
 #351


The limiting factor on the CPU is not memory bandwidth, rather memory latency which is less than 1 GB per second because the memory latency of main memory (if outside of L3) is several hundred clock cycles. With only 8 hardware hyperthreads, that latency isn't entirely masked away.

Whereas, the GPU can run 1000s of hardware threads (not software threads!) which masks away the latency and hits the memory bandwidth as the limit.

This is why the GPU blows away the CPU.

A hardware thread has its own copy of registers so there is nearly no cost to blocking the thread on memory access, so another thread can run which was blocked and is ready to proceed.

I think an optimized algorithm, either for CPU or GPU is going to need really detailed knowledge of memory latency issues - more knowledge than I have. You might be onto something with this - I don't have enough knowledge in this area to know if or what level of improvement managing memory latency could offer. You'll have to wait for BM's response.

Membercoin - Layer 1 Coin used for the member.cash decentralized social network.
10% Interest On All Balances. Browser and Solo Mining. 100% Distributed to Users and Developers.
AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 07:23:40 AM
Last edit: November 07, 2013, 08:17:30 AM by AnonyMint
 #352


The limiting factor on the CPU is not memory bandwidth, rather memory latency which is less than 1 GB per second because the memory latency of main memory (if outside of L3) is several hundred clock cycles. With only 8 hardware hyperthreads, that latency isn't entirely masked away.

Whereas, the GPU can run 1000s of hardware threads (not software threads!) which masks away the latency and hits the memory bandwidth as the limit.

This is why the GPU blows away the CPU.

A hardware thread has its own copy of registers so there is nearly no cost to blocking the thread on memory access, so another thread can run which was blocked and is ready to proceed.

I think an optimized algorithm, either for CPU or GPU is going to need really detailed knowledge of memory latency issues - more knowledge than I have. You might be onto something with this - I don't have enough knowledge in this area to know if or what level of improvement managing memory latency could offer. You'll have to wait for BM's response.

I have that knowledge and have done all the research.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
Sy
Legendary
*
Offline Offline

Activity: 1484
Merit: 1003


Bounty Detective


View Profile
November 07, 2013, 07:25:40 AM
 #353

Interesting thing, a 32 core xeon server gets only ~45 hpm (-10 if i increase threads from 32 to 64) and found 1 block so far - where a decent quadcore (sandy and newer) will run at 20-25 hpm - just a little "fun fact" for all your bot and datacenter resistance comments.

Oh and Windows generally performs worse since its 32 bit, 3 threads net me 11 hpm in Windows and around 16-18 in Linux, 4 got me to the mentioned 20-22 hpm with an i5-3570k non oc.

BOUNTY DETECTIVE


















Powered by,
AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 07:43:09 AM
 #354

I believe someone is already GPU mining it and planning to dump thousands of coins.
And your evidence for this is what?

People kept believing that Primecoin was being GPU mined, without any evidence. And still no-one has admitted to being able to mine it on GPUs, the two well publicised efforts both ending in failure.

Any thing that is computationally intensive can be mined faster on the GPU employing 1000s of hardware threads, so I find it impossible to believe that Primecoin is a CPU-only coin.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
Etlase2
Hero Member
*****
Offline Offline

Activity: 798
Merit: 1000


View Profile
November 07, 2013, 07:45:14 AM
 #355

Note the 768MB is within L3 cache, which has much lower memory latency, so the specific case might not be as bad as a larger memory case.

When did L3 caches become 768MB?

AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 07:48:46 AM
 #356

Note the 768MB is within L3 cache, which has much lower memory latency, so the specific case might not be as bad as a larger memory case.

When did L3 caches become 768MB?

Excuse me I was thinking 768 KB. You are correct, it is outside of L3. I've been rather sleepless lately.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
FreeTrade (OP)
Legendary
*
Offline Offline

Activity: 1428
Merit: 1030



View Profile
November 07, 2013, 07:59:22 AM
 #357

Inflammatory, offtopic post about primecoin removed.

Membercoin - Layer 1 Coin used for the member.cash decentralized social network.
10% Interest On All Balances. Browser and Solo Mining. 100% Distributed to Users and Developers.
Etlase2
Hero Member
*****
Offline Offline

Activity: 798
Merit: 1000


View Profile
November 07, 2013, 08:01:32 AM
 #358

Any thing that is computationally intensive can be mined faster on the GPU employing 1000s of hardware threads

A GPU is a relatively general purpose device, and with GPGPU on the horizon, is it really a problem? What is the fascination with CPU only? As long as ASICs remain specific and an algorithm is easy to change, ASICs can't have an advantage over GPU/GPGPUs because they can easily become obsolete. With this algo, it is trivial to change the hashing algorithm while keeping the verification simple and general purpose devices relevant.

AnonyMint
Hero Member
*****
Offline Offline

Activity: 518
Merit: 521


View Profile
November 07, 2013, 08:12:17 AM
 #359

Any thing that is computationally intensive can be mined faster on the GPU employing 1000s of hardware threads

A GPU is a relatively general purpose device, and with GPGPU on the horizon, is it really a problem? What is the fascination with CPU only? As long as ASICs remain specific and an algorithm is easy to change, ASICs can't have an advantage over GPU/GPGPUs because they can easily become obsolete. With this algo, it is trivial to change the hashing algorithm while keeping the verification simple and general purpose devices relevant.

The point of CPU-only is the millions mass your coin can gain in the market since anyone can download and get some coin. Mass leads to economic network effects, which perhaps leads to competing against Bitcoin effectively, given Bitcoin only has 350,000 users.

I have programmed software all by myself which had a million downloads when the internet 10 times smaller.

Also the mining is in theory not cartelized. This is very important in my opinion. If the mining is controlled by a few entities, you no longer have a decentralized money and the government can then take over.

Changing algorithms is not practical for an established coin, because the existing peers may not accept the hard fork. So ASIC immune design needs to also be considered.

unheresy.com - Prodigiously Elucidating the Profoundly ObtuseTHIS FORUM ACCOUNT IS NO LONGER ACTIVE
digitalindustry
Hero Member
*****
Offline Offline

Activity: 798
Merit: 1000


‘Try to be nice’


View Profile WWW
November 07, 2013, 08:20:27 AM
 #360

The bounty isn't for whether or not GPUs can mine it, the bounty is for whether or not there is a non-linear speed increase available without the memory tradeoff. If GPUs end up being faster, big deal, it isn't going to be an order of magnitude faster unless you have an order of magnitude increase in the memory bandwidth or in the memory available. At least, that is the presumption by my understanding.

^
That

- Twitter @Kolin_Quark
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 [18] 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 ... 71 »
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!