AMD prepares dual-core 45nm surprise.

Rick, you can't go at engineering but go after physics (owned by Mother Nature).
When these things are processed it's not possible to control all the densities of the channel wells and traces. Due to molecular differences and flow the etching process can sometimes erode too much causing caverns under the layers between the resist and exposed semiconductor shrinking the trace or transistor stack. Any of these can cause the component to be too thin to handle current of the heavy creating lag and in some cases shorted. Most of the disc edge stuff I'm sure is chunked. This is the first production run. Soon a second run will adjust the troubled spots attempting to improve the mask which might prevent some losses. This happens all the time, it just takes getting out enough stock before the change is implemented. In the mean time take advantage of moving bunk stock that would other be waste. Again, todays manufacturing is focused of reducing scrap which is a big loss in this case.

I too would love to see a true dual again at 45nm. It would probably kick arse.

Another thing, AMD does have it's hands on narrow transistor tech which will increase the density of RAM.

For now, enjoy large cashes because we had them taken away in AM2 before Phenom!

hooflung said:

Intel does chip sampling too. The e4300 is a prime example of it. It has all the virtualization bits on die but its turned off. It also has 2mb of it's cache disabled.

Usually when the initial product comes to market its because of defects. If market demand goes through the roof they will start crippling it's big brothers to meet demand. A chip sold is a chip sold. It doesn't matter if they can meet market demand for every man woman and child on the planet, if people won't buy certain parts at certain pricepoints they will always make a cheaper product. Economic's 101.

AMD just also tend to sample their cores now like ATI did pre-buyout. If you can take 2 cores that don't work and sqeeze a product line out of it then why not? Its either that or the garbage bin.

Intel is also going to be doing that with their Neh. chips soon as Arstechnica stated last week. They will be harvesting cores for different products. IE, now intel has a chip that CAN be treated like this and not have the Celeron name to it.

Click to expand...

Have a to go with the half eaten Neha!

How much faster would a C2D with 12mb of L2 cache compared to 6mb? Is it the cache that makes the C2D more peppy then the X2's?

Particle said:

Your response assumes that I'm somehow surprised or shocked by the information. I'm not. What I listed was a counter to the claim that 99.9% should be achieved for high volume lithography and that anything below should be considered faulty production (eg, time for an examination as to why things are going so badly). I'm not sure what you're laughing about.

Click to expand...

D**n my shoe taste bad I used that number without thinking. Patooey anybody got some salt water to cut this bad taste, oh crap there is crap on my shoe.

I have PCB and electronics fab exp (My statement about 99.XX% was evidently not a qualified statement) and in that scene that number is what I shoot for or start looking for a problem.

Rick James said:

How much faster would a C2D with 12mb of L2 cache compared to 6mb? Is it the cache that makes the C2D more peppy then the X2's?

Click to expand...

Depends on the app, more interesting is I think how much PH2 triples and duals benefit from more L3. Note they wrote about only 6MB L3 so they disable some of the cache no matter what, so we might never really tell. T_T Don't want more skus or there aren't enough 8MB duals, or maybe that would be too good ?

You can see the C2D cache scaling from 1-4MB here. Link

There might be newer ones but most compared 65 to 45 nm and that would be misleading although in a different way than this one (low clockspeeds mean slower cache with higher lat so larger cache has an edge here) ^^;;

Wolfdale VS Conroe.

C2Ds real strength is its monster integer execution, that's why the changes to neha were made to feed the beast, and the large cache was there to hide the effect of the FSB the best they could.

Might sound stupid... but given it's on the large Deneb chip, it should run pretty cool giving more room for overclocking, yes?

Based on this, I believe AMD is having a yield issue. The dual core processor still dominate in low end boxes from HP and Dell. Those are the big volume pusher. There is a reason why the price is so cheap on those boxes. They need volume to stay profitable. If AMD start doing this strategy, they would have a hard time to squeeze more profit out of these die. In addition, competing against Pentum Dual Core? With that die size, Intel can slash the price to the point that AMD makes no profit and still win. We are talking about 3x die area in advantage!

Well the fact that they are selling them is helpful to them. As long as they can sell them and gain a return beyond what it would cost them to just throw them away then it is helpful.

If intel slashes prices to the point to where amd is not making a profit on the chip it will still be better for them to sell them because it is recouping some of the expense of manufacturing the batch to begin with. (the whole batch of chips, quads on down).

I would also agree that they may be having yield issues though. The only alternative to this is that they are taking good quads and neutering them just to make the numbers. That may be what they are doing though if they can still make a small profit on each dual/tri sold to the box companies. No point in having a ton of quads sitting around if you aren't selling them so just sell the best and cut the others down and sell off as tri/duals just to keep product going out the door and money coming in.

I think this is a good approach for them though. I also think it will turn out better for them in the long term after intel reaches EOL on all of the 45nm C2 chips and the only thing left for the box guys to buy is neha parts. If this happens then AMD could tie up the budged market fairly nicely. Of course I doubt intel will let it happen that easily but it is a possibility. They could continue production until either they have a replacement chip that is cheap to sell or the cost to produce low end nehas comes way down. If this doesn't happen though then AMD will easily be able to fill the void, if they can make the numbers.

Anyway, I think it has been and will continue to be a good business move for AMD to sell off the defective quads as tri/dual core chips. They need all the help they can get and with moves like this (and others seen in more recent times) the future might be looking a bit brighter for them.

Firestrider said:

I think the original Phenoms should have been the same principle as Core 2 Quads. Two native dual-core MCM connected via HyperTransport.
<snip>

Click to expand...

There was a reason for AMD taking the route they did. When it came time for the die shrink Phenom II had few design issues to iron out. What issues there are mostly come from the shrink, not the architecture. The architectural issues were ironed out with Phenom I.

Results speak for themselves ...

Yahira said:

If intel slashes prices to the point to where amd is not making a profit on the chip it will still be better for them to sell them because it is recouping some of the expense of manufacturing the batch to begin with. (the whole batch of chips, quads on down).

Click to expand...

Intel does not really cut prices unless the move to a new node, they kill the lowest clocked model an the lineup falls one step. Q6600 really fell when there were plenty of 45nm cpus from Intel and they barely made 65nm cpus anymore.
So if AMD can release a new stepping with higher top clocked cpus they can do the same and keep their prices.

I would expect another step drop from Intel for their back to school price cut to make room for I5 so AMD has 2 quarters to deliver they did it with barcs so it could happen.

solstice said:

I also think it will turn out better for them in the long term after intel reaches EOL on all of the 45nm C2 chips and the only thing left for the box guys to buy is neha parts.

Click to expand...

First of all C2D stays till 2011 after that I don't see K10 as a major seller especially against 32 and 25nm low end cpus.
I5 won't be that expensive especially on 32nm, mobos will be cheaper than I7, and DDR3 ram prices are coming down to the acceptable levels. Ok there is not much to gain from cheap DDR3 but by the time I5 comes out it will be a rather small price difference for the same performance as DDR2.

QuietIce said:

There was a reason for AMD taking the route they did. When it came time for the die shrink Phenom II had few design issues to iron out. What issues there are mostly come from the shrink, not the architecture. The architectural issues were ironed out with Phenom I.

Results speak for themselves ...

Click to expand...

Well as I recall the new CEO said they regret going that way in light of 4 quarterly losses thanks to the big, slow, native quadcore, I would not call it a major win that they could deliver a good product for the second try.

If you think about it, Intel has been manufacturing at 45nm for over a year now. They might just have their process so fine tuned that maybe 2 or 3 are bad out of a 100,000 maybe more. But the fact that they shrank the c2d's and c2q's before making nate shows that they might have had yield issues they wanted to avoid with their new baby.

hooflung said:

Intel does chip sampling too. The e4300 is a prime example of it. It has all the virtualization bits on die but its turned off. It also has 2mb of it's cache disabled.

Click to expand...

Core 2 Solo anyone
truth is at the beginning of a chips life, there are always realatively low yealds for it. Example is the core 2 solo which was a conroe/allendale with one core disabled because of bad manufacturing (At least I think it is)

I'm real behind on my computer knowledge right now, but whatever AMD needs to do right now is what they need to do. They haven't been making a profit, and just like everyone else they gotta do things a bit differently with this economic climate.

^^^ very true. Unfortunately, this is why I don't have 1066 ram, a new mobo and a PII 940 right now...

Kuroimaho said:

Intel does not really cut prices unless the move to a new node, they kill the lowest clocked model an the lineup falls one step. Q6600 really fell when there were plenty of 45nm cpus from Intel and they barely made 65nm cpus anymore.
So if AMD can release a new stepping with higher top clocked cpus they can do the same and keep their prices.

I would expect another step drop from Intel for their back to school price cut to make room for I5 so AMD has 2 quarters to deliver they did it with barcs so it could happen.

Click to expand...

I know, I said this more to go along with my point that even if Intel did start to aggressively cut prices, forcing AMD to go extremely low with their own prices, it is still better that AMD sell off the chips in this fashion than just trashing them.

It seems as though some in this thread were suggesting that it is a bad move for AMD to sell off "neutered" quads as tri/dual core chips. I am just stating that from a business standpoint it is a good idea. The vast majority of the consumers will never know the difference. It lowers costs by only needing to start the production of a single type of chip (all start on the quad "line") and it recoups some of the money spent making the chips as opposed to just throwing them away and having the expense of making native dual/tri core chips (or not and still losing the money from not selling the chip at all).

Even in good times when AMD is making native dual core (or what ever we are at for the lowest core chip at the time) I think this would be a good idea. Sell the native duals/tris/quads/whatever chips as the medium and high end chips and sell the neutered chips as low end both to consumers building budget rigs and to OEMs building low dollar boxes.

If for example if we are at 8 core chips and they took and sold off the "neutered" octas as 6 and 4 core chips for a price of $50 and $60 each it would still be a good idea (if for some reason they couldn't make more on them). At that price it would probably be a loss per chip from manufacturing but it is still better than just throwing the chip away and making nothing on it. In business it isn't always about making money, you also have to work to lose as little money as possible.

Actually, I believe one thing we make a serious assumption about and I am not 100% positive is that Intel and AMD have similar wafer costs.

Check this article out:
http://www.electronicsweekly.com/Ar...n-insulator-cuts-manufacturing-costs-says.htm

So if I read it right by cutting out marketing BS, it add 10 to 15% per wafer costs of a wafer compare to regular wafers. However, HKME may add extra cost by increasing the amount of time it takes to make the stuff (hour = $) or by using more expensive materials. Although, I kinda doubt that is the case. At least as an apple to apple comparisons, SOI does not seem to be cheap.

The marketing BS is that you can use less logic and less memory which is clearly not the case with AMD. The transistor count and die size is about the same with Intel i7.

Ah! I had mourned about this a few weeks back. Finally a cheap AMD chip I can swap in place of my X2 4850e/780G, same price, with definitive more performance, tweaking and very low power (which will certainly get tweaked to even lower, heh).

Will these really be AM3 chips only? Hmmm..

Now its all about price. I'll take the slowest X2 they can throw out.

AMD, hurry up... such priced products are where the largest volume market lies.

If only you could get a low priced 790GX mATX too.

KTE said:

Will these really be AM3 chips only? Hmmm..

Click to expand...

Since AM3 is backwards compatible with AM2+... makes no sense for them not to be...

Makes sense to make another small step this way. It's a good start to getting the dual IMC working, which I've heard has been problematic.

Kuroimaho said:

Well as I recall the new CEO said they regret going that way in light of 4 quarterly losses thanks to the big, slow, native quadcore, I would not call it a major win that they could deliver a good product for the second try.

Click to expand...

It may have been a business error but, IMO, it wasn't an R&D error. Not a major win but AMD has a history of taking one step at a time and the Phenom I release fits that history. I'm sure they would have had losses even without a native quad and the shrink would have cost them more ...

solstice said:

Actually, I believe one thing we make a serious assumption about and I am not 100% positive is that Intel and AMD have similar wafer costs.

Click to expand...

More than just one I believe, from assuming fab equipment price is the same that intel does not get lower prices by upgrading 4 fabs a time etc, but have no idea about those details.

QuietIce said:

It may have been a business error but, IMO, it wasn't an R&D error.

Click to expand...

Well it could be compared to the I7, which did add IMC, HT and became a native quad and was a hole in one, although it could be said Intel has more experience with huge dice thanks to Itanium.

Nonetheless it was ahead of it's time, when quads made up less than a third of the cpu sales making triples and duals from quads was far from being the ideal solution.
They need native dual cpu and keep the one with L3 as a higher end product.