Topic: AMD ASIC? (Read 3030 times)

You all are behind the game! I've been mining on AMD ASICs for over a year now!

I also agree here. I think the idea of a more experienced company taking the time to create Bitcoin-specific hardware is pretty intriguing, but at the same time I understand how very unlikely the scenario is. In order to garner support for such a project, Bitcoin would need to be taken seriously..or at least have such a substantial following that it warrants interest in larger organizations.

Or maybe if one of the execs at AMD, enamored by Bitcoin, gets bored one day and decides to set aside a few million for ASIC R&D. Hell, I'm sure they could just create a small shell company to do it all without necessarily spreading the word that AMD is fully supporting Bitcoin mining...

Yep, just an exercise the imagination thread. 

Probably closer to $100M in annual sales required for process time in current gen Fabs.  One can think of a Fab as a $10B investment that continually depreciates in face value.  Kinda like a mining bond but on the billion dollar scale. It has to throw off utterly insane cashflow in the early years just to keep up with depreciation.

Still AMD annual revenue is ~$7B.  Global demand for ASICs is probably <$20M.  Essentially a rounding error.  If we ever get to a point where annual mining rewards (subsidy + fee) is in the tens of millions then maybe a smaller established semiconductor company might leverage connections to get into the market but even that wouldn't be worth it for a company the size of AMD.

I'm not sure I understand this thread, is this hypothetical, wouldn't it be neat if it were true kinda stuff, or is this suggesting AMD might actually be considering this? Because they're not. Just like everyone shaking in their boots that "AMD makes AMD cards, so they could take over the bitcoin network!" were wrong, and the people who suggested that "AMD make a bitcoin based GPU" were silly, they're not going to make a bitcoin ASIC.

All the old usual reasons apply, the market is pathetic, volatile, uncertain as to longevity, and in general a horrible risk for any company of semi-reasonable size. While $5million probably seems like a lot to us chump-change miners, it's more like what AMD spends on toilet paper. I see that AMD put in a mention of mining, right alongside folding, in their GPU highlights. Meaning they equate the two hobbyist projects the same, though I see no evidence that they actually did anything mining or folding specific, just that bigger faster cards do both better.

If this is just a exercise the imagination thread though, I apologize, and it would be neat.

That's why I'm thinking it'd be a larger-process fab (less demanded), using 65nm or maybe 40nm.

assuming they haven't been designing them this whole time, they'd be coming into it awfully late.  You really don't want to be the last one in on the bitcoin market.  Plus, then their support dept would have to deal with BTC-related problems.  Stupid people would call them if their wallet file got stolen or for instruction on how to configure their mining client, etc.  It'd be rather expensive.  The alternative is to release a product and not support it at all basically, except hardware and software.  That's possible since they don't sell that many parts directly anyway but if they follow that course, we'll have Sapphire, PNY, MSI, Gigabyte mining cards, etc, cuz AMD would just supply the cores.  Then all of those companies would have to support bitcoin related questions to some extent.  Then their entire R&D budget would be wasted if anything bad happens to the bitcoin network.  Considering how badly they're doing right now, I don't think they'd take that big of a risk.

Isn't the volume required to access these foundries way above what makes sense for Bitcoin ASICs? I expect the minimum volumes to be in millions of $ or even 10s of millions but would like to get references on this subject.

Actually... AMD, Qualcomm, Broadcomm, STMicroelectrons, and produces ARM chips for Multiple MFG's.

Yes, but AMD and GF still work very closely together, do they not?  AMD would still be responsible for everything but actually creating the chips, I would think!

Contact Global Foundries... AMD is now a fabless company.

Wouldn't be too hard to get them on board. They're experienced with 40nm and 32nm products, and rolling out 28m dies for their next gen products.

BFL (Or any ASIC developer) would definitely research them for their second gen products.

Course... that's if they ship their first gen products.

Fair enough!   

They are a large, ponderous, confused corporation. Other than that missing detail, your analysis is reasonable.

What if AMD gets in the game and creates an ASIC at some point in the future?  They are obviously aware of Bitcoin mining, as they advertise mining as a selling point on their latest and greatest video cards.  So then, they are likely equally aware of the fact that GPU mining is on its way out the door.  And they should also be aware of Bitcoin continuing to grow, creating a larger and larger market for mining devices.

They have the fabs, the engineers, and the reputation to make it happen.

Here's what I picture if AMD did it...

- PCI-e based (it would live inside your computer just the same as a video card).
- Different levels of miners, up to 200w of consumption (using the normal 6-pin or 8-pin power connectors as necessary).
- Unless Bitcoin mining grows SIGNIFICANTLY (would require a price upwards of several hundred dollars), I don't see them putting much more than just a small amount of engineering and fab time behind this.  Thus, one could expect an older, larger process to be used - 65nm, or perhaps one step smaller. They have access to 28nm processes, but those modern processes are better off spent on GPU's and the like.
- Very competitive pricing (probably nothing that costs more than a top-end GPU).
- Targeted towards the more mainstream crowd of people, even towards (currently) non-Bitcoin users.
- Makes use of existing assets they have (perhaps reusing video card HSF assemblies, etc).

The 3rd point above is why I believe they haven't given it more than a token glance, but you never know what they might have going on behind the scenes.  Bitcoin miners would likely accept up to a 1 year present-value break-even, especially with AMD giving Bitcoin some extra credibility.  The remaining 3 ASIC vendors have likely pre-order-sold around $5M-$6M worth (I know this because of reasons), which means there's still $15M in potential sales to capture.  If AMD determines they can do a complete process, start to finish, for only a couple million, they might see that there's still plenty of potential profit in it for them.

Thoughts?