UltraSPARC III @ 1200Mhz

[Danzotron]

Ok guys,
Question with bitcoins as reward for good info.

I have a Sun Fire V880 with 8 UltraSPARC III Cpus @ 1200mhz at my disposal. This thing is a beast. Surely there is a way for me to get some decent hashing going on with it?

I've noticed there's not really much of a port, or anything that's been moved over doesn't get decent speeds.

Can someone give me the hot tip on how I can start really utilising this machine?

[1714021027]

1714021027

[1714021027]

1714021027

[1714021027]

1714021027

[niooron]

Didn't Sun hardware have a cryptography accelerator or something? It may be worthwhile.

[chromicant]

You want an UltraSparc T2 for SHA256 acceleration. I've had zero luck trying to get a test account on one....it's not common hardware someone would say "sure, have a shell account on this $10k server".

[error]

it's not common hardware someone would say "sure, have a shell account on this $10k server".

Why not? IBM does.

[frankiebits]

Ok guys,
Question with bitcoins as reward for good info.

I have a Sun Fire V880 with 8 UltraSPARC III Cpus @ 1200mhz at my disposal. This thing is a beast. Surely there is a way for me to get some decent hashing going on with it?

I've noticed there's not really much of a port, or anything that's been moved over doesn't get decent speeds.

Can someone give me the hot tip on how I can start really utilising this machine?

I bet my four core Intel CPU @ 3.4 Ghz can beat your 8 1200Mhz Sun CPUs.

Side bet, my AMD Cypress single core GPU at 1000Mhz will beat all 12 cores combined.

I bet you a whole bitcoin.

[urizane]

With a cryptography accelerator, does it really need to be massively parallel?  Running ufasoft's SSE2 miner on my Core i7 yields hashes at just under every 900 cycles on each core.  If a crytography accelerator can do that same work in, I don't know, say... 100 cycles, it could be worth it.  It would be more worth it than an AMD X2 or a Core Duo and I'm sure there are a few of those out there mining.

http://wikis.sun.com/display/CryptoPerf/UltraSPARC+cryptographic+accelerators
This article seems to outline only UltraSPARC T1 and UltraSPARC T2, although only T2 will do the job you want.  An UltraSPARC III does not have SHA-256 acceleration built into the core.  If you have something like a cryptography accelerator card, then you'll have to research the manufacturer and see if they made any separate libraries that you can use.  I wouldn't get my hopes up, though.  A cursory glance at separate crypto cards shows that a large number of them were designed to accelerate SSL, which wouldn't be of any help here.

If it turns out that you have a crypto card that doesn't have a co-processor on it that you can run arbitrary code on, you're probably not going to get much out of that system.  8x 1.2 GHz cores (in-order, superscalar) with highly optimized code could maybe net you roughly 10Mhash/s for the whole system.  It would be cheaper to buy bitcoins than burn through all of the wattage trying to mine for them.

[lfm]

Ok guys,
Question with bitcoins as reward for good info.

I have a Sun Fire V880 with 8 UltraSPARC III Cpus @ 1200mhz at my disposal. This thing is a beast. Surely there is a way for me to get some decent hashing going on with it?

I've noticed there's not really much of a port, or anything that's been moved over doesn't get decent speeds.

Can someone give me the hot tip on how I can start really utilising this machine?

It would be a large problem to port to this machine due to it being big endian. afaik bitocin has never been ported to big endian and I think Satoshi said some time ago that the bitcoin client was highly dependent on little endian architecture and unlikely to ever work on big endian.

Prehaps the java bitcoinj is better base to start with, not sure.

Not trying to start another flame war about what is better, just another design decision from early days of bitcoin that is a problem for you now.

[ArtForz]

According to sun, T2 can do "32 Gb/s/chip" for SHA-1
let's just assume it's as fast for SHA-256.
let's further assume it's magic and has 0 setup cost.
SHA256 block size is 512 bits, so 32Gb/s / 512 = 62.5M sha256 blocks/s
one bitcoinhash is 2 sha256 blocks, so 62.5M / 2 =
31.25Mhash/s.

Not bad for a crypto accel.
A $110 HD5770 does > 5x that.

[frankiebits]

I didnt post this for a biggest e-peen fest. I'm well aware of what the hardware is and what it is and isn't capable of. This thing was our production box for 7 years and did an excellent, albeit we wrote in house applications to take advantage of it for the task we were doing. I'm not too good at coding, but simply asked for help from someone with a better knowledge of architecture if this thing was worth while or not for mining. No need to come in here and have a wank about your GPUs.

I believe the machines we updated to have a cryptography accelerator in them, is there code out there to take advantage of such things? I could schedule it to mine at night.

I was just messing with you.. no e-peen fest. 

[urizane]

You could upgrade the storage on that system and turn it into a pretty bad ass database or web server system or some other pile of data on a domain.  You might as well play to its strong suits and put a separate low cost machine toward mining for bitcoins, unless the crypto accelerator can be reprogrammed to do SHA-256 hashes with ease.  Am I right in assuming it's installed in a PCI slot (or maybe multiple cards)?

http://en.wikipedia.org/wiki/File:Sun-crypto-accelerator-1000.jpg

EDIT: This question is aimed at the original post about an UltraSPARC III system.

[chromicant]

According to sun, T2 can do "32 Gb/s/chip" for SHA-1
let's just assume it's as fast for SHA-256.
let's further assume it's magic and has 0 setup cost.
SHA256 block size is 512 bits, so 32Gb/s / 512 = 62.5M sha256 blocks/s
one bitcoinhash is 2 sha256 blocks, so 62.5M / 2 =
31.25Mhash/s.

Not bad for a crypto accel.
A $110 HD5770 does > 5x that.

I did find a student paper which measured the performance of the T2 here: http://cs.anu.edu.au/student/projects/10S2/Reports/Cody%20Christopher.pdf...and ArtForz really hit the nail on the head (like, freaky scarily on the head) compared to the experimental results. Seems like if you run 8 threads worth on a T2 (there's 8 hashing units on each T2 die, from what I can read), you would hit ~31.25Mhash/sec if you hash 8k of data.

Where the paper is a little unclear is if hashing 512 bits on the hardware is worthwhile, since it seems like the Sun supplied code will use the hardware if it's worth it. From section 4.2.1 in the paper, if we force the hardware to hash 512 bits, then we see that the speed of the hardware is only 1510.94kb/sec. This leads to a rate of 2.95M sha256 blocks/s, or 1.5Mhash/s.

Looks like the setup cost is prohibitive for small blocks, but for larger ones, it's negligible. Tradeoff for the T2 is around the 8k mark.

[lfm]

 This leads to a rate of 2.95M sha256 blocks/s, or 1.5Mhash/s.

Thats funny since it is exactly the same speed the VIA-C7 hashing engine achieves with bitcoin. ($67 including motherboard).

[chromicant]

Thats funny since it is exactly the same speed the VIA-C7 hashing engine achieves with bitcoin. ($67 including motherboard).

Even funnier. A T2 system is *cough* slightly *cough* more expensive:

http://cgi.ebay.com/SUN-Coolthreads-UltraSPARC-T2-Enterprise-T5120-Server-/180516461674

Only $10k USED!

Course, if you have a fab shop, you can make your own T2 (the design is Open Sourced)...in theory. Don't know if anyone has ever done that.

[dasein42]

The V880's co-processor board does have an API.  I used a V480 with the same board years ago, using the co-processor for HL7 encryption (medical software).  It's hard to imagine that old 700mhz encryption processor out mining a new dedicated machine, though, and writing a driver to use the board for bitcoin software is likely more work than it's worth just to get use out of an old machine.

Trying to use a more modern Sun machine, really any of the T series, is probably a waste of time for anything involving number crunching.  They aren't designed for that, they're built for data throughput.  They make phenomenal web servers, and phenomenal Java servers, so long as your Java code is CRUD + relatively simple business logic, which covers probably 95% of the Java server apps out there. 

If you wanted to look at used high end machines (often significantly cheaper on ebay than used PCs with equivalent power) look at Power6 or Power7 based machines.  Those are number crunchers, period, it's what they were designed to be good at and they are. 

The biggest difficulty with any older machines, though, for mining bitcoins is that they use a lot of electrical power.  The one reason for possibly considering a newer T series Sun is that the processors are ultra low power (an 8 core CPU uses about 20W), and you won't find decent performance in a low power intel CPU.  I've not tested the dedicated boxes as far as power consumption but from what I understand they use slightly less power than the value of the mined coins, so your return is negligible but at least not negative.

A potential solution to the underlying problem if you're planning to do enough mining, no matter what hardware you settle on, is using solar power to run the machine (wind power would work too, depending where you are).  A solar setup big enough to run a large Power7 based machine (or a few of the small dedicated units) would run about $5k.  You can work out, based on how much you can mine with your setup and adding the cost of the setup to the power gen unit's cost, how long it would take to pay for itself.  After that, as long as the solar power unit continues to work fine (most are warranted fully for at least 5 years) you're essentially continuing to mine bitcoins for nothing.  It's a gamble of course that mining bitcoins will be profitable 5 years from now, but you take that gamble with any setup that doesn't cost more to run than the value of the mined coins.

[MarketNeutral]

Nice machine you have, OP. Unfortunately, CPU mining bitcoins for profit has been dead for years—no exaggeration. A few people have managed to find ways to mine on big-endian machines, but none has managed to accumulate any meaningful quantity of bitcoins by doing so. If you want to make it work for the lulz or even mine some altcoins, I understand, but don't fool yourself into thinking you'll profit. Those electricity costs can sting.

[-ck]

Nice machine you have, OP. Unfortunately, CPU mining bitcoins for profit has been dead for years—no exaggeration. A few people have managed to find ways to mine on big-endian machines, but none has managed to accumulate any meaningful quantity of bitcoins by doing so. If you want to make it work for the lulz or even mine some altcoins, I understand, but don't fool yourself into thinking you'll profit. Those electricity costs can sting.

Did you look at the date on this thread you guys necro'd? In march 2011 he actually would have been mining meaningful amounts of bitcoin with CPU.

[MarketNeutral]

Nice machine you have, OP. Unfortunately, CPU mining bitcoins for profit has been dead for years—no exaggeration. A few people have managed to find ways to mine on big-endian machines, but none has managed to accumulate any meaningful quantity of bitcoins by doing so. If you want to make it work for the lulz or even mine some altcoins, I understand, but don't fool yourself into thinking you'll profit. Those electricity costs can sting.

Did you look at the date on this thread you guys necro'd? In march 2011 he actually would have been mining meaningful amounts of bitcoin with CPU.

No, I didn't notice the date. Thank you. My mistake.