Tom's Hardware

[mizerydearia]

I submitted an initial email to Tom's Hardware with only the following:

<mizery> tomshardware.com should start a section in their benchmarks to benchmark which hardware is best for bitcoin ^_^
<nanotube> mizery: hehe you should suggest it to them.

Chris Angelini responded:

Greetings Mizery,

I’ve just done some reading on bitcoin, and the question I’d ask would be: is there a way to even test which hardware is best for generating bitcoins? It seems like the generation of a bitcoin is not so much determined by the hardware you have, but by the constant average rate. And even then, depending on the amount of CPU power in play, this will change over time.

If there is a way that you think bitcoins can be tested, I’m up for trying to figure out a way to make it happen =)

All the best,

Chris Angelini

My follow up:

Hello Chris,

  The hardware that is best for generating Bitcoins is the hardware that has the highest khash/sec figure. An average may be necessary since the figure fluctuates.  Using the Bitcoin GUI there appears a khash/sec number (rounded to thousands, hence "k" for "thousands" in "khash") that represents how many hashes are calculated.  A more precise hash/second amount can be extracted using the getinfo API method documented at http://www.bitcoin.org/wiki/doku.php?id=api.  For example, in a linux environment you can execute `bitcoin getinfo` and monitor the "hashespersec" value.  You can extract this information using the Bitcoin GUI or the Bitcoin Daemon.

My computer is using an AMD Athlon II X4 620 cpu and with the latest version of official Bitcoin client v0.3.12 I can generate an average of approximately 9000-10,000khash/sec ("hashespersec" : 9439274).  On a computer (that is ~3 years old purchased from a local computer store, I am not sure what cpu is inside) at my previous job, it can generate approximately 300-700khash/sec.  There is a much obvious notable performance increase in my computer's cpu as compared to a general store-bought pc/cpu from ~3 years ago. It may be interesting to see how a variety of modern cpus fare for Bitcoin generation.

There has been recent interest from the Bitcoin community in determining which hardware is suitable or best for generating.  One individual blogged about it: http://bitcointalk.org/index.php?topic=955.  As of now, there isn't a kind of established way to determine which hardware is most suitable for generating Bitcoins.  At this time there are still many individual enthusiasts who are contributing towards generating Bitcoins, but in the next few years, it will be too costly for many individuals to generate Bitcoins and generating will be limited to businesses that established around generating Bitcoins.  Individuals could potentially establish such a business, and perhaps with information made available in a benchmark, it could be determined which hardware is best value for generating Bitcoins by reviewing Tom's Hardware benchmarks.

As of right now the official Bitcoin client relies on cpu processing for generating hashes.  However there are two other alternative clients that rely on gpu which are MUCH faster.  One is a proprietary 32-bit Windows CUDA-based implementation designed by puddinpop (intro: http://bitcointalk.org/index.php?topic=133.msg11940#msg11940, latest version: http://bitcointalk.org/index.php?topic=133.msg12255#msg12255) and there is another unreleased client by ArtForz (http://bitcointalk.org/index.php?action=profile;u=584), who is more available on irc @ freenode / #bitcoin-dev than on the forum, that is OpenCL/linux-based.  As of September 2nd, he has 12 gpus dedicated to generating about 1ghash/sec

<ArtForzZz> about 1.8kW for 1.05Ghash/s
<ArtForzZz> at 623.4 difficulty 1.05Gh/s = 70.83BTC/h

http://www.alloscomp.com/bitcoin/calculator.php
With the current difficulty at 712.88486454999997476988937705755234, generating 1gash/sec
ProbabilityPerSecond(0.00032660343238012378545490000000) Chances: Avg(0d 0:51:01) 25%(0d 0:14:40) 50%(0d 0:35:22) 75%(0d 1:10:44) 95%(0d 2:32:52) 99%(0d 3:55:00)

At this time, however, there is no open source or distributed version of Bitcoin client that uses OpenCL.  Such a client would be far superior than any cpu-based client, however, until one exists, cpu-based benchmarks will be useful for those who wish to buy hardware dedicated to generating Bitcoins.  It would be very nice, however, if there were an initiative to develop an open source OpenCL-based Bitcoin client.  As of yet, the Bitcoin community is still small with less than 100 regulars in freenode / #bitcoin-dev channel.  A few months ago as Bitcoin was posted on slashdot, it helped to establish a spike in awareness of Bitcoin, but the community is still small.

Thanks for reading.  I look forward to hearing back.

l3estest l2egardedsness,
Mizery

p.s. http://bitcointalk.org/index.php?topic=1011.0

note: I was too excited to write back to Tom's Hardware asap and didn't bother to revise or correct grammar/spelling mistakes.  However, in follow up above, I have corrected some grammatical/spelling mistakes.

Follow-up from Chris:

Thanks for the detail here Mizery, I’ll have to give this a closer look when I get back from IDF.

If the community is still small, then it’d probably be difficult to justify making this a part of the complete benchmark suite. However, perhaps a feature story on it might be doable. Is this something that you or any individual in the community might be interested in pursuing?

All the best!

Chris

My response:

Hello Chris,

You may review the Bitcoin wiki at http://bitcoin.org/wiki and additionally if you have any questions, I encourage you to come to our IRC channel at irc.freenode.net #bitcoin-dev and we can help to answer any questions you may be interested in asking.  If you do not have access to an IRC client, you may use http://webchat.freenode.net for web-based IRC client.  Simply enter a nickname and specify channel #bitcoin-dev.  We look forward to seeing you on IRC.

l3estest l2egardedsness,
Mizery

[1714171102]

1714171102

[1714171102]

1714171102

[1714171102]

1714171102

[1714171102]

1714171102

[mizerydearia]

Bitcoin on VPS - Perhaps there will be a type of VPS service established that offers Bitcoin hosting.  Perhaps potential Tom's Hardware Bitcoin benchmarks can help to provide which hardware is best value and to rely on that information for establishing a VPS hosting business specifically for Bitcoin.

There are many profit-oriented individuals that come to the IRC channel discussing excitement towards generating but sometimes are disappointed that their limited khash/sec will require weeks or even months to generate a block.  Some of these individuals may be turned on to VPS hosting to turn a profit.

A slightly more open approach to bitcoin on the GPU

[mizerydearia]

I updated the initial post with follow-up from Chris.

Perhaps I shouldn't have emphasized small at the end of my reply.  Boooo me.  Buuuuut, yay for honesty!

[LZ]

You had to say that we are 10 thousands.

[mizerydearia]

Response:

Hi Mizery,

Sorry for the delay! My wife and I had our first baby a couple of weeks ago, so I’m very behind on email now (in addition to sleep).

I actually hadn’t hopped on IRC yet—was it someone else named Chris perhaps? I can definitely try hopping on later, though it gets tough to sit for very long with the little guy waiting in the other room

As before, I’m still open to someone in the bitcoin community collaborating with us to come up with a standardized benchmark and perhaps do a story on the concept, software, and evolution. I did read up on the FAQ and am intrigued.

Hope all is well with you Miz! Take care,
Chris

[Anonymous]

Not only could they use bitcoin as a benchmark for CPU, but they could also do it for GPUI on OpenCL platform.

It certainly taxes your cpu to the max.  A hashmark for cpu's ? 

[mizerydearia]

I finally followed up to Chris's latest response from October 13, or so I thought.

Hello Chris,

I have endured a rather long hiatus from Bitcoin for a while, and I missed out on many latest developments, however, relaying your question to the active IRC community, the answer seems rather simple: "a way to test? yes. run a generator, see how fast it goes."

Additionally, ArtForz, a rather notable and popular individual within the Bitcoin community and widely known for his vast array of hardware resources obtained specifically for Bitcoin mining, suggested a 6-core phenom II can get up to 16m hash/sec using 4way.  This would be close to the highest possible generation rate per second (as of yet).  From my experiences using very low end systems that are able to run bitcoin, the lowest I've encountered is about 15k-20k hash/sec.  My computer using the official bitcoin client and using CPU for processing reached approximately 1m hash/sec.

ArtForz, has introduced GPU mining many months ago and has since established very successfully dominating in the realm of most hash/sec within the network.  I haven't been keeping up-to-date on GPU developments, so I cannot elaborate further on the subject, but from observing recent discussion from ArtForz, he has possible plans to invest $200,000 in establishing a particular mining project that will blow GPU mining away.

Regarding time, time has no effect on generating.  For example, the difficulty has increased exponentially since the last half year when I began participating actively in the Bitcoin community.  This computer still generates at teh same 1m hash/sec (1,000,000 hashes) as it did.  The only thing that has changed since then is the difficulty.  e.g. the chances that I will generate a winning hash.  e.g. the odds have become worse.

Based on how Bitcoin is designed, if I generate a winning hash from my computer's attempt of generating one million hashes per second, then that will result in my generating a block and all nodes within the p2p network will eventually receive notice that I won that block, and additionally I will receive the x (at this time still 50) bitcoins.  The block count will increase by one, and all nodes in the p2p will continue trying to guess the next hash and create an additional block.

For every 2016 blocks that are generated, the difficulty is reconsidered.  As the Bitcoin network first started, the difficulty, which started at 1.0, remained at 1.0 for many of the 2016 blocks that were generated.  Here's a textual representation:

Block      0 was generated at 1231006505 (Sat, 03 Jan 2009 12:15:05 -0600)
Block   2016 was generated at 1233063531 (Tue, 27 Jan 2009 07:38:51 -0600)    1233063531 seconds interval 14271d 13:38:51   Difficulty: 1.00
Block   4032 was generated at 1234466190 (Thu, 12 Feb 2009 13:16:30 -0600)       1402659 seconds interval    16d 05:37:39   Difficulty: 1.00
Block   6048 was generated at 1235966513 (Sun, 01 Mar 2009 22:01:53 -0600)       1500323 seconds interval    17d 08:45:23   Difficulty: 1.00
Block   8064 was generated at 1237508786 (Thu, 19 Mar 2009 19:26:26 -0500)       1542273 seconds interval    17d 20:24:33   Difficulty: 1.00
Block  10080 was generated at 1239055463 (Mon, 06 Apr 2009 17:04:23 -0500)       1546677 seconds interval    17d 21:37:57   Difficulty: 1.00
Block  12096 was generated at 1240599098 (Fri, 24 Apr 2009 13:51:38 -0500)       1543635 seconds interval    17d 20:47:15   Difficulty: 1.00
Block  14112 was generated at 1242098425 (Mon, 11 May 2009 22:20:25 -0500)       1499327 seconds interval    17d 08:28:47   Difficulty: 1.00
Block  16128 was generated at 1243737085 (Sat, 30 May 2009 21:31:25 -0500)       1638660 seconds interval    18d 23:11:00   Difficulty: 1.00
Block  18144 was generated at 1246051973 (Fri, 26 Jun 2009 16:32:53 -0500)       2314888 seconds interval    26d 19:01:28   Difficulty: 1.00
Block  20160 was generated at 1248481816 (Fri, 24 Jul 2009 19:30:16 -0500)       2429843 seconds interval    28d 02:57:23   Difficulty: 1.00
Block  22176 was generated at 1252069298 (Fri, 04 Sep 2009 08:01:38 -0500)       3587482 seconds interval    41d 12:31:22   Difficulty: 1.00
Block  24192 was generated at 1254454028 (Thu, 01 Oct 2009 22:27:08 -0500)       2384730 seconds interval    27d 14:25:30   Difficulty: 1.00
Block  26208 was generated at 1257002900 (Sat, 31 Oct 2009 10:28:20 -0500)       2548872 seconds interval    29d 12:01:12   Difficulty: 1.00
Block  28224 was generated at 1259358667 (Fri, 27 Nov 2009 15:51:07 -0600)       2355767 seconds interval    27d 06:22:47   Difficulty: 1.00
Block  30240 was generated at 1261130161 (Fri, 18 Dec 2009 03:56:01 -0600)       1771494 seconds interval    20d 12:04:54   Difficulty: 1.00
Block  32256 was generated at 1262153464 (Wed, 30 Dec 2009 00:11:04 -0600)       1023303 seconds interval    11d 20:15:03   Difficulty: 1.18
Block  34272 was generated at 1263250117 (Mon, 11 Jan 2010 16:48:37 -0600)       1096653 seconds interval    12d 16:37:33   Difficulty: 1.30
Block  36288 was generated at 1264424879 (Mon, 25 Jan 2010 07:07:59 -0600)       1174762 seconds interval    13d 14:19:22   Difficulty: 1.34
Block  38304 was generated at 1265319794 (Thu, 04 Feb 2010 15:43:14 -0600)        894915 seconds interval    10d 08:35:15   Difficulty: 1.81
Block  40320 was generated at 1266191579 (Sun, 14 Feb 2010 17:52:59 -0600)        871785 seconds interval    10d 02:09:45   Difficulty: 2.52
Block  42336 was generated at 1267000864 (Wed, 24 Feb 2010 02:41:04 -0600)        809285 seconds interval    09d 08:48:05   Difficulty: 3.78
Block  44352 was generated at 1268010873 (Sun, 07 Mar 2010 19:14:33 -0600)       1010009 seconds interval    11d 16:33:29   Difficulty: 4.53
Block  46368 was generated at 1269212064 (Sun, 21 Mar 2010 17:54:24 -0500)       1201191 seconds interval    13d 21:39:51   Difficulty: 4.56
Block  48384 was generated at 1270120042 (Thu, 01 Apr 2010 06:07:22 -0500)        907978 seconds interval    10d 12:12:58   Difficulty: 6.08
Block  50400 was generated at 1271061586 (Mon, 12 Apr 2010 03:39:46 -0500)        941544 seconds interval    10d 21:32:24   Difficulty: 7.81
Block  52416 was generated at 1271886772 (Wed, 21 Apr 2010 16:52:52 -0500)        825186 seconds interval    09d 13:13:06   Difficulty: 11.46
Block  54432 was generated at 1272966376 (Tue, 04 May 2010 04:46:16 -0500)       1079604 seconds interval    12d 11:53:24   Difficulty: 12.84
Block  56448 was generated at 1274278435 (Wed, 19 May 2010 09:13:55 -0500)       1312059 seconds interval    15d 04:27:39   Difficulty: 11.84
Block  58464 was generated at 1275141448 (Sat, 29 May 2010 08:57:28 -0500)        863013 seconds interval    09d 23:43:33   Difficulty: 16.61
Block  60480 was generated at 1276298786 (Fri, 11 Jun 2010 18:26:26 -0500)       1157338 seconds interval    13d 09:28:58   Difficulty: 17.38
Block  62496 was generated at 1277382446 (Thu, 24 Jun 2010 07:27:26 -0500)       1083660 seconds interval    12d 13:01:00   Difficulty: 19.40
Block  64512 was generated at 1278381464 (Mon, 05 Jul 2010 20:57:44 -0500)        999018 seconds interval    11d 13:30:18   Difficulty: 23.50
Block  66528 was generated at 1279008237 (Tue, 13 Jul 2010 03:03:57 -0500)        626773 seconds interval    07d 06:06:13   Difficulty: 45.38
Block  68544 was generated at 1279297779 (Fri, 16 Jul 2010 11:29:39 -0500)        289542 seconds interval    03d 08:25:42   Difficulty: 181.54
Block  70560 was generated at 1280198558 (Mon, 26 Jul 2010 21:42:38 -0500)        900779 seconds interval    10d 10:12:59   Difficulty: 244.21
Block  72576 was generated at 1281037595 (Thu, 05 Aug 2010 14:46:35 -0500)        839037 seconds interval    09d 17:03:57   Difficulty: 352.16
Block  74592 was generated at 1281870671 (Sun, 15 Aug 2010 06:11:11 -0500)        833076 seconds interval    09d 15:24:36   Difficulty: 511.77
Block  76608 was generated at 1282864403 (Thu, 26 Aug 2010 18:13:23 -0500)        993732 seconds interval    11d 12:02:12   Difficulty: 623.38
Block  78624 was generated at 1283922289 (Wed, 08 Sep 2010 00:04:49 -0500)       1057886 seconds interval    12d 05:51:26   Difficulty: 712.88
Block  80640 was generated at 1284861847 (Sat, 18 Sep 2010 21:04:07 -0500)        939558 seconds interval    10d 20:59:18   Difficulty: 917.83
Block  82656 was generated at 1285703908 (Tue, 28 Sep 2010 14:58:28 -0500)        842061 seconds interval    09d 17:54:21   Difficulty: 1318.67
Block  84672 was generated at 1286861705 (Tue, 12 Oct 2010 00:35:05 -0500)       1157797 seconds interval    13d 09:36:37   Difficulty: 1378.02
Block  86688 was generated at 1287637995 (Thu, 21 Oct 2010 00:13:15 -0500)        776290 seconds interval    08d 23:38:10   Difficulty: 2149.02
Block  88704 was generated at 1288479527 (Sat, 30 Oct 2010 17:58:47 -0500)        841532 seconds interval    09d 17:45:32   Difficulty: 3091.73
Block  90720 was generated at 1289305768 (Tue, 09 Nov 2010 06:29:28 -0600)        826241 seconds interval    09d 13:30:41   Difficulty: 4536.35
Block  92736 was generated at 1290105874 (Thu, 18 Nov 2010 12:44:34 -0600)        800106 seconds interval    09d 06:15:06   Difficulty: 6866.89
Block  94752 was generated at 1291135075 (Tue, 30 Nov 2010 10:37:55 -0600)       1029201 seconds interval    11d 21:53:21   Difficulty: 8078.19
Block  96768 was generated at 1291933202 (Thu, 09 Dec 2010 16:20:02 -0600)        798127 seconds interval    09d 05:42:07   Difficulty: 12251.99

The reason the difficulty stayed at 1.0 for almost a year was because Bitcoin is designed to generate 2016 blocks in exactly two weeks (in seconds).  If it takes longer than two weeks to generate 2016 blocks, then the difficulty decreases.  You can notice that the difficulty decreased at block 56448 back down to ~11.84.  However, the difficulty can never be less than 1.0, therefore for the first year it stayed at 1.0.  If 2016 blocks are generated before two weeks, then the difficulty increases.  The calculation for determining the new difficulty is based on how long it took to generate 2016 blocks and Bitcoin is designed to establish a new difficulty that will help to bring generating the next 2016 blocks closer to a two week time frame.

Since the Bitcoin community is drastically increasing and most notable recent hardware and software (OpenCL/CUDA clients) developments especially from ArtForz, have greatly impacted how quickly blocks are generated as well as the drastically increasing difficulty.

When I first established myself into the Bitcoin community, the difficulty was ~23.50, and I managed to generate a few blocks, and in only a day, sometimes even a couple times a day.  I don't bother to generate anymore because according to calculations at my computer's hash rate will take almost nine weeks to have ~60% chance to generate a block.  Because of this I no longer generate anymore.  I only have one computer and am nowhere near as well invested as ArtForz into Bitcoin mining.

When I first contacted you regarding the idea of using Bitcoin generation as a kind of benchmark, I was excited about determining which hardwares were best for generating Bitcoins.  Additionally there were many others who asked similar questions.  For me, and I believe many others, it isn't so important, at least not unless individuals plan on investing lots of money into building large powerful clusters of computers to mine Bitcoins similarly as ArtForz has established for himself.

In regards to the original idea of using Bitcoin as a kind of benchmark, it may still be interesting or worthy of doing.  Perhaps it may be beneficial in terms of raw processing power for constant computation of data, especially in terms of recent establishing of p2p-related services  e.g. Bitcoin is p2p currency, bitdns or dns-p2p (http://dot-p2p.org) is p2p dns, and there is potential for many other future p2p established services in the future.  In terms of determining what the best hardware is for such p2p uses. perhaps a kind of benchmark for this purpose may seem useful.

As I have done previously, I will post this to our forum at http://bitcointalk.org/index.php?topic=1011.0  If you have any interest you are welcome to join our forum community and/or irc community and establish further information as necessary.

Also, I noticed pcworld.com have written an article about Bitcoin earlier today: http://www.pcworld.com/businesscenter/article/213230/could_the_wikileaks_scandal_lead_to_new_virtual_currency.html

l3estest l2egardedsness,
Mizery

I had written the above response after focusing on his first email to me.  During entire time of writing my response I was under impression that the first email was his latest response.  I later realized my mistake and wrote:

Hi Chris,

I just realized that I responded to your FIRST email and not your latest.  >_<  Gah.... sorriez about that!  hehe...  blah!

Congrats on the baby!

In regards to further collaborating with someone from the community, I recommend creating a new thread on our forum at http://bitcointalk.org

Previously I had suggested our IRC community, which is still convenient, active and occasionally more noticed by many of the core developers, but it may be more useful to use the forum for better documented responses and especially for linking if necessary.

Well, take care.

l3estest l2egardedsness,
Mizery