Bitcoin x64 for Windows

I am... would it be even faster on SSE3?

I've got a spare server using a 8-core 64-bit Intel - 64 bit Windows 2003 server. I'd be curious if there is a difference from the stock, the machine currently does about 5000khash/s, I'd be curious if the 64bit native build would be better.

Yep.

I keep getting a "MSVCR100.dll" not found error when trying to start it. I copied the file over from another system into the system32 folder, but the program still can't find it for some reason. Do we need the full Microsoft dev environment installed to test it?

I'll guinea pig, I'm running it on a test system that has no balance yet, well trying to run, it won't start.

No missing DLL errors when you try to start it?

Get visual C 2010 redist from Microsoft -- pointers here:

http://www.mydigitallife.info/2010/04/17/visual-c-2010-runtime-redistributable-package-x86-x64-ia64-free-download/

The new code is full of win.. hash rate (4 core) went from 1350 to 1880 khash/s.  Sweet!

I recommend getting the redist from the official source:

x86 here: http://www.microsoft.com/downloads/details.aspx?FamilyID=a7b7a05e-6de6-4d3a-a423-37bf0912db84

x64 here: http://www.microsoft.com/downloads/details.aspx?FamilyID=bd512d9e-43c8-4655-81bf-9350143d5867

If there's demand for it, I can compile an Itanic Itanium build... for a reasonable consideration

that's very interesting, what processor/frequency are you on? x86 or x64?

Nice job, this is just my initial testing, but the 64 bit compile speeds up hashing by over 28% so far over the 32 bit counterpart.

So for example, my 8-core system does (600 khash/s per core =  4800 khash/s) for normally, but now averages 5700 khash/s

I'll certainly be sending some bitcoins your way 

]

do you have the Visual C++ 2010 runtime installed? if not, that's why.

Does that apply for the x86 one also, since that's what I was using to test the 32bit build he had?

[edit] I just checked, it was good for the 32bit 

Yes, straight from the download link on the previous page for x86.

I looked through the error log a little, didn't spot anything that stood out.

Sorry to be so skeptical, but there was a victim on this site when someone claimed to have compiled a CUDA client that used the graphics processor to hash.  One guy fell for it and lost some bitcoins.

But yes, if I get a chance I will look at the dlls I suppose before running it.

Or, maybe you can release the project files and we can compile it ourselves?

I think he should really present to this devs to get credit for it and some BC because I'm going to send some to him for raising my khash/s up and beyond the already insane levels that I have. 

Here's a weird question.. has anyone actually generated a block with this faster version?

I have a few machines that used to regularly generate, and since switching to this version -- zip.  I know these things are subject to random variation, and it could be a dry spell, and the difficulty is going up, but... could it be a bug?

Ichi: I got 2 blocks in a row back when I was running 1200khash.  I'm up to 2500 now, as per the update, so we'll see if I get some more blocks. =)

Hey, I just made this improvement that will vastly improve your khash/s!

In main.cpp, ~2751, change:
   string strStatus = strprintf("    %.0f khash/s", dHashesPerSec/1000.0);
to:
   string strStatus = strprintf("    %.0f khash/s", dHashesPerSec/300.0);

Then on ~2758, change:
  printf("hashmeter %3d CPUs %6.0f khash/s ", vnThreadsRunning[3], dHashesPerSec/1000.0);
to:
  printf("hashmeter %3d CPUs %6.0f khash/s ", vnThreadsRunning[3], dHashesPerSec/300.0);


(DISCLAIMER: THIS IS A JOKE.)

That's good news.  Thanks 

So --- Bitquux, that was Olipro's binary that found the coin?

It shouldn't be dirty, the SHA256 is untouched and I don't think the compiler has introduced any errors.

Yes, about 10 so far. The other "regular" clients acknowledge test Coin transfer between the two, so far it appears good. I also have a packet sniffer running on my test machine and it's basically doing what the other clients do, nothing out of the ordinary so far.

OK, I've made a new build of Bitcoin, this one is compiled using the Intel compiler which is considerably more advanced than the standard MS compiler.

Please note this build is 64-bit only since I see no reason to compile a build for 32-bit when most 32-bit processors lack the newer SSE instructions anyway.

Performance? my MSVC build averages about 2400K for me, this averages 2900K so you're looking at an improvement of about 125k per core or thereabouts although this does come at the cost of a larger EXE, improved performance is worth it in my opinion

My next goal is to see if this PolarSSL SHA256 algo really is faster. for now however, download Bitcoin x64 ICC optimised build here

Awesome, gives a 41% speed boost over the stock binary.

My experience so far.

Stock = Stock
Last x64 Build = 28% speed increase over stock
This x64 Build = 41% speed increase over stock

Nice stuff!

What would I expect for two Xeon 5570?

OK, I've made a new EXE, this one seems to get me an extra 100-200k (or about 25-50k per core).

Difference? I modified the ByteSwap function to operate on 64-bit integers; it does this by using the bswap intrinsic on a 64 bit register followed by rotate right through 32 bits to put the result in the correct order. it also initializes the SHA256 vectors using unsigned 64 bit values (however, the actual hashing still uses 32-bit so I doubt this is making much of a difference) and yes, I did convert the 32 bit numbers to 64 bit correctly (i.e. 0x12345678UL 0xabcdef0UL -> 0xabcdef012345678ULL) if that appears wrong to you, think about how little endian machines store 32 bit integers in memory.

anyway, grab it here

May I kindly ask you, mighty Olipro, to patch bitcoin 0.3.1 or may be 0.3.2? There were some issues in 0.3

my builds are from the latest SVN source code, of course, had you checked the about page, you would notice it clearly states itself as 0.3.2

then either it got re-broken or he also modified something in wxWidgets itself which I don't have.

on an unrelated note: I have CUDA working for SHA256 hashing... currently the host has to iterate the nonce so performance is terrible, working on moving the whole kit and kaboodle into the card and then performance should be pretty damn sexy.

Oli: One alternative is to start the CUDA part hashing with a nonce of MAXINT and subtract one, and the host client start at one and go upwards.  Then, either one could find a hash and you're not repeating work.

Ichi: Probably pretty fast, I saw a huge increase in performance to the tweaked version.  No tellin until you try it though.

the missing DLL is in my first release (libeay32.dll)

given that it's based on stock code and I doubt there's an issue with the compiler, it's more likely there's a bug in the SVN source, I'll make a build from the last stable production version just as soon as I've finished this CUDA code off.

My experience is similar, the first release was rock solid like the original client, the other releases, while still faster, randomly crash after minutes/hours/days  so it's kind of random.

So it's a trade off between stability and speed. The faster it gets, the less stable it seems. 

Of course, I have mine on a batch to restart if there is a crash and to count how many times the program crashed during the day for example. So far up to about 2 crashes a day.

Yeah, my Linux 64bit systems still have the leg up on my windows servers, even with this optimization that is made that speeds them up by 50%, seems Linux still rules the roost for coin generation speed.

Most of my coin generation comes from my Linux servers more than my windows servers, it's about a 1 to 4 ratio, for every 4 blocks made by my Linux servers, 1 will be made by one of the windows servers.

Yes, sorry, I left that part out, the Linux servers (same hardware) always generate higher khash/s than the windows machines, at least mine do. Your mileage may vary.

OK.  Could someone please explain the why of that, in terms that a non-programmer can understand?  Why is the stock Windows build slower than the stock Ubuntu build?

Thank you.  I'm glad that I'm moving to Ubuntu.

I really like Windows Server 2008, which is pretty close to Windows 7.  It'll be a while before I give that up.

OTOH, I'm thinking very seriously re going with Ubuntu for my next primary desktop.  As a test, I've recently installed Ubuntu 10.04 Server x64 + GNOME on an old Core 2 Quad machine.  Everything except boot lives on an encrypted RAID5 array managed by LVM.  Losing Excel 2007 might be a problem, though.

Although I don't use Tor, I'm sure that there must be many setup guides for Ubuntu.

This is way off topic.  Sorry.

mind if i ask?
satoshi has released a couple of updates recently, mostly about security concerns as far as i can see.
how are the 64bit versions keeping up with these developments?

OK, I've made a new build now; this version uses the 64bit SHA256 Assembler code from Crypto++ which means the Byteswap function is now only used for re-ordering the resulting hash - and of course it's using 64-bit ASM to create the hash in the first place.

Performance? I've gone from an average of 2900k hashes on my previous builds up to a pretty stellar 3300k hashes (an improvement of about 100k per core) - I would be very surprised if the Linux builds outperform this.

Grab x64 Asm Optimised Bitcoin here

If you want the code, get Crypto++

after tinkering around a bit and reinstalling the vcredist, i even got the VS build to work now.

on the regular client i get around 1600khash/sec, your latest VS build currently runs at ~2750khash/sec,
+70% that's an outstanding performance!

OK, now for some absolutely incredible performance.

Credit to tcatm for the caching part of the SHA context - this offers absolutely brilliant performance. Additionally, the Intel compiler really comes into its own here as its parallelisation abilities give a massive performance boost over Visual Studio.

Performance: 4700khash/s on 4 cores, I think that speaks for itself.

I've included both the VS and Intel build, but there's really no comparison, the Intel build craps all over VS.

Grab SHA state caching Bitcoin here

Um.. Wow.  That last one was a bit of a leap.

Intel Core i7-870 (2.93 GHz) running 4950 khash/s here. (4 cores Turbo'd to 3.2 GHz)
(Intel version)

Anyone seeing a speed difference between VS and Intel?

It was indeed a large jump, not sure how stable it will be in the long run but, we will find out. The stock client ran fine 24 on this machine so I will post back some time tomorrow. I am getting less khash/s with the vc build, around 2600khash/s. The intel build for me is pushing the most khash/s.

I'm a total newb, I tried to use Olipro's file and it failed because I was missing
MSVCR100.dll I got it, now it says "The application was unable to start correctly." Should I just not mess with this? Is it supposed to be all good to go?

Just say Olipro's new thread. Got it working, so sweet.

OS is 64-bit. I loaded the runtime. It works now, but is the same khash as the x86.

I'm confused now, I don't see a VS or an Intel folder. When I unzipped vcredist_x64 I'm getting a bunch of numbered folders 1028, 1031, etc. I ran Setup and when I rerun it it just wants to repair Microsoft Visual C++ x64 Redistributable to it's original state each time.

I've also just noticed that it's getting incoming blocks like 3 or 4 at a time. It could be they're coming in close together, but I think I'm getting them like simultaneously.

EDIT: Okay, I see it now.

I hit 2700, up from 1250 stock.

I have 7 cores, how do I tell how many it's using? Can I control it?

Long time visitor, First time poster. How can you possibly have 7 cores? What Processor are you running, Cores come in even numbers . . . I would check your processor is working correctly. How do you even know that you have 7 cores and you don't know how to control them. I'm guessing your a Windows user.

I had to register to rebel against some of the crap I see been posted here!

Every time he posts I want to poke him in the eye, the one conveniently posted to the left of his friendly well-meaning nonsense 

This is going off topic completely but, there is a very slim theoretical chance that you could have 7 cores on an amd machine. You could disable a core on a dual quad-core system or it might be possible to get another core or two by core unlocking (using the extra cores that are some times on amd cpus but,"locked"). Based on how many khash/s they are getting they do not have 7 cores or a running in a vm. If they do have 7 cores they should post a cpuz screenshot because, it would seem that we have an outlier in the performance boosts and it would be good to figure out why.

Back on topic, My desktop have been crunching away since last night @  06:57 am forum time (9 hours) using the intel build, no crashes, blocks are coming in, and I have 29 connections.

I was thinking about Virtual Box when I wrote that 

the BitCoin threads are automatically assigned low priority; as soon as any other processes want to use the CPU it will automatically lose CPU time