Scrypt mining only on GPU's? Any FPGA or ASIC companies out there?

[skeeterskeeter]

As of 0ct '13 I've looked around and can't find much. But is there any scrypt mining hardware out there or soon to be out that is not GPU based?

[1714771295]

1714771295

[1714771295]

1714771295

[1714771295]

1714771295

[btcrich]

No.  So far no ASIC has been developed for Scrypt.

[kramble]

As of 0ct '13 I've looked around and can't find much. But is there any scrypt mining hardware out there or soon to be out that is not GPU based?

Nothing you can buy right now, but there are several in development, see project development on the litecoin forum.

WARNING there are also several obvious scams out there, eg scryptasic.org https://bitcointalk.org/index.php?topic=259640.0

[tokeweed]

oh now you want asics.

[skeeterskeeter]

oh now you want asics.

lol

No I was just wondering, I had looked a few months ago and determined everything to be a scam. I just figured after peaking at LTCs market now that someone might have started or made something by now.

If anyone could, I've looked up scrypt on tarsnap and a few other places but I don't quite get it. It is an encryption algorithm? Could someone give a good overall pseudo-code for it?

I figure, from the BTC protocol, that you take the block header of the LTC block and run it through this algorithm to find a low probability output; the probability being based on the LTC difficulty setting of the network?

[kramble]

If anyone could, I've looked up scrypt on tarsnap and a few other places but I don't quite get it. It is an encryption algorithm? Could someone give a good overall pseudo-code for it?

Pseudo-code, no, but scrypt.c is a very readable implementation.

Not encryption, in mining usage its a hashing algorithm (but also usable for encryption, hence the confusion). And yes, its just a matter of comparing the resulting hash with the network target (1/difficulty). If its lower you've found a block (or a share if you're pool mining against a lower difficulty target)

[skeeterskeeter]

If anyone could, I've looked up scrypt on tarsnap and a few other places but I don't quite get it. It is an encryption algorithm? Could someone give a good overall pseudo-code for it?

Pseudo-code, no, but scrypt.c is a very readable implementation.

Not encryption, in mining usage its a hashing algorithm (but also usable for encryption, hence the confusion). And yes, its just a matter of comparing the resulting hash with the network target (1/difficulty). If its lower you've found a block (or a share if you're pool mining against a lower difficulty target)

The implementation is just above me. I get parts of it, but overall I can't follow it through. It looks like it uses SHA-256 function? But I also read that the computation is memory intensive. What is line of adjustments made to the input to produce the output?

Maybe does it do many SHA-256 hashes on the input, making it memory intensive as the amount of hashes needed to be performed and the data they perform on gets backed up in RAM?

[hulk]

Although many FPGA scrypt mining have been announce, but all of them are pre-order and none of them have shipped working FPGA. So no body knows..

[hope2907]

YES, in near future
Scrypt purpose is hardware intensive so that it limit the amount of fpga asic device can be use with same amount of money, unlike sha256 very cheap, but sell very high due to greedy of company an blindness of customer, scrypt asic or fpga will very expensive, and with current price/ dif it is not worth to buy -> no one buy, no one dev it

[kramble]

The implementation is just above me. I get parts of it, but overall I can't follow it through. It looks like it uses SHA-256 function? But I also read that the computation is memory intensive. What is line of adjustments made to the input to produce the output?

Maybe does it do many SHA-256 hashes on the input, making it memory intensive as the amount of hashes needed to be performed and the data they perform on gets backed up in RAM?

I did a very quick overview here

The SHA256 hash is just a wrapper around the meat of the salsa algorithm which happens between lines 373 and 400 of the code (two sets of 2048 calls each to the salsa20_8 half-round function).