I think Barton will be a great product for the upgrader's market especially the millions of Socket A systems out there with Athlons from 800Mhz-1.4Ghz. Those boards may not support more than 266Mhz FAB though. It'd be nice if these new CPUS can work with older boards than don't support the higher FSBs. I think they can.

C

Subzero, I think Athlon 64 will boast a 400Mhz FSB like the Barton ( or maybe even higher ). It may also have more than 256K L2 Cache according to AMD sources.

C

subzero, I think AMD's rating holds no matter what apps they use to determine it. So, the benchmark suite they'll use for the Operon may be different than the one currently in use for the Athlon.

C

Deleted... Other people had the right response already....

wbmw, besides the fact that he's quite old, why not?

C

wbmw, Jealous (of Sanders) arent' we? :) Regardless, he would not have bought AMD stock unless he has a plan on how he's going to make more money out of it.

C

And he did it just for that....hmmm, you need to rethink that.

Sanders bought over 300,000 AMD shares on Friday!!!!!!!!!


Link:
http://biz.yahoo.com/djus/021125/0926000399_1.html

C

wbmw, AMD (or INTC) should have bought TMTA when it was under $1. AMD could have done as a stock deal and they could now have a good competitor to Banias and they could even fab Astro @ Dresden and probably get more Mhz out of it than the foundries do. I am very impressed even though we don't know if the mobile p4 was running on batteries which makes it run @ 50%-60% speed.

C

Semi, I agree completely that in absolute terms a 300mm will make more die for the cost of the gas and the wafer but the fixed cost of the equipment and the actuall running the company remains fixed it does not get cheaper. You still have to sell all of those chips to make more money. If you have to give them away at lower asps because you have a glut you may be doing yourself a disservice.

greg, Intel can afford to spend as much $$$ on fabs as they like. There are no guarantess as to when the semi market will turn. It's a gamble but it's one they can afford to take since they have tons of $$$$.

wbmw, I still don't see how 300mm can help a semi company in this market ( even Intel ). I thought the fixed costs of running a company are much greater than the variable costs. Unless you can sell all the additional capacity that 300mm offers, I don't see how it can be of a cost benefit. What it can definitely do is allow you to increase the # of transistors without dropping the number of cpus produced.

greg, I think they are delaying the venture until the market picks up. What's the sense in producing more chips when the market can't absorb the quantities already made?


p.s. All the baloney about lowering costs by using 300mm is meaningless. The way costs are lowered is by producing more chips per wafer hence more chips overall. In this market, more chips are not needed.

wbmw, and why can't they use HT? It's open and free as far as I know. Why reinvent the wheel?

C

OT: LOL!


C

Tim, even though I disagree with personal attacks on any board, I think that you can't be sued for libel against a fictional alias name on a board. It's not like your real persona was attacked :)

C

wbmw, thanks for the MPF coverage. AMD suggested that in Q4/Q1 they'll try to break even at $750 Million so they are tightening their belts. At those levels I think they can stay in business indefinitely and make tons of money if Hammer ever ships....

Regardless, since you were there and talked to the actual engineers, what's your gut feeling?

Thanks

C

Spokeshave, I agree with your assumptions but you have to account for the coming improvements in P4 like more L2 cache in Prescott. Dual DDR channels which should improve latency, newer, better chipsets, etc, etc...

wbmw, Dual 1.8Ghz 64-bit cores? Sounds cool and it may stabilize Apple's sales. I guess there go the dreams of many AMD fans to have Apple go with Hammer... On the other hand this shows that more companies are trying to find ways to improve cpu throughput without going for Max Mhz/Ghz.


C

Last I heard, high-bandwidth was good for all cpus. I think what may be more important for HT enabled cpus is lower latency. I think that's why Intel is going to the Dual-Channel DDR route. HT is interesting up to the point where you realize that non-HT conforming code may actually run significantly slower on an HT CPU probably due to cache thrashing. The majority of programmers have an aversion to threads because they complicate the development process. In certain apps, threads are absolutely necessary. I think HT will shine with Prescott when Intel increases the L2 to 1MB. Northwood with HT is more of a validation vehicle for HT than anything else.

C

Nitt, I think AMD will ship to OEMs in Q1 and will book revenues from Bartons and Hammers in Q1. In Q2 we'll see boxes bartons and hammers.By Q3, AMD could be at 50% hammers. All this is hypothetical of course and assumes no more slips.

C

EP, I was talking about Q1 of '03 since that's when Clawhammer is supposed to be introduced. Also, it seems that Intel is pricing their top of the line chip at $650+.

C

Nitt, that's a valid arguments. Let's take Intel's cpus for example. Given the distribution from 1.8Ghz to 2.8Ghz how would you allocate the percentage sold for each speed grade ( they have to total to 100% ). The other consideration is that AMD's ASPS are suppressed because they don't own the high-end anymore. Having the performance leadership allows all the manufacturer to pull up the prices of all their chips. For example, the lower-volume Athlon MPs still fetch well over $100 but there are not enough of those to make a dent. For your information the $199 price listed on Pwatch for the 2400+ is the lowest one. There's one for $238 also. I think that AMD will price the clawhammer 3400+ at 25% off the 3Ghz P4 which should price the 3400+ at around $350, then subtract some $50 for the OEMS and they'll get around $300. By then they'll have the 2800+ and 3000+ bartons ( hopefully ) which can fetch anywhere from $150 to $250 and they should be able to pull up their ASPS. They just have to execute like you say. If they can sell 5 million TBreds @ $50-$75 and 1 million Bartons and Hammers at ~$150, they will do ok.

C

Too expensive....

Nitt, ok, let's see. We know AMD has a rather large inventory of parts that they can sell while they ramp Hammer so they can afford to move more wafer starts to Hammer than they would otherwise have to. I don't know the exact numbers. Also, clawhammer is tiny, slightly larger than TBred-B so even if AMD startes with 15% of the wafer starts for hammer they should be able to make 1 million Clawhammers for Q1. As far as ASPS I would scale their current Athlon linearly up to 3400+ ( even though higher end part prices are usually higher than a linear scaling would indicate ). So, the 2400+ goes for $199 I'd say that a 3400+ would go for around $299 which is ok by me, that should bring AMD back into the black. Of course this is a best case scenario. AMD can alway make less than 1 million claws for Q1 and they can sell them at $150 which will keep them in the red. I think by Q2 though they'll have their act together and be back in the black.

C

Re:That would be the apples to oranges comparison. Thoroughbred is a shrink of Palomino, not Thunderbird, so it should be compared to Palomino.

According to EP and you guys, Palomino was using .13 features already :)

Re:I have not seen any indication of this. Some overclocking experiments have reached 2.5GHz in reviews, but these were not at the default voltage. And if 2.5GHz was so easy, that doesn't explain how AMD was able to ship a few 2.0GHz parts, but no 2.13GHz parts, yet.

Almost all the reviewed samples overclocked to 2.4Ghz air-cooled reliably. Maybe they were hand picked, maybe not.We'll see on that one.

Re:I don't see how it makes a difference. Intel is ahead at .13u and they will probably be ahead at 90nm, too. It's apples to apples, though your point is that AMD has room for improvement, which may be a factor, or may not.

OK

Re:What difference does that make? The end result is that Pentium 4 has higher performance at the frequencies it can reach. Unless the Athlon scales higher, it will not match Intel in performance.

I am not arguing about total performance ( even though P4's REAL performance is still questionable ). This thread was talking about AMD's .13 process failure. I was trying to make a point that we don't know for sure that the process failed.

Re:I have no knowledge of any capacitance changes, but given the reduced power levels in TBred-B, it seems that capacitance had to have been reduced. It probably took both process changes and transistor backend design changes to pull that off.

It was discussed on SI. Anyway, a backend design change is different than a "regular" design change so does it have its own name?

Thanks

C

Nitt, the MB manufacturers have had Hammer samples for many months now, I don't think that's an issue. Most Hammer boards will come with USB 2.0, AGP 8x, Serial ATA and Firewire on board. There are no Intel boards with all of those features and none are slated for anytime soon. Some 3rd party P4 boards are starting to appear with those features though.

C

wbmw, we're comparing apples to oranges. Intel went from a .18 aluminum process to a .13 copper process and will derive a 50% speed improvement ( yes, I am including the 2 Ghz Wilammete since you are including the 1.73 Ghz Palomino even though you should be comparing it to the .18 Thunderbird which topped out at 1.4Ghz ). Anyway, the early Tbred-B samples indicate that 2.5Ghz is not out of the question which will be about 50% of the Palomino and this is without getting the benefit of moving to a copper process. Also, Intel is a whole year ahead of AMD in developing/refining their .13 process so we're now comparing apples to cucumbers. And, if it was all about the process, then we can see that Intel's P3 @ .13 topped out at 1.4Ghz while AMD's is now crossing 2Ghz and moving up. It sounds like it's not the process and more likely the 20-stage vs 10 stage pipeline. Basically, AMD is trying to make a 10-stage cpu scale as well as a 20-stage and they may pull-it off. I have a question for all of you. AMD fine-tuned some components in TBred-B to change the capacitance ( excuse my lack of knowledge ) in order to achieve higher-clock speeds. Is such a change considered a process change, a design change or some other change?

C

EP, I like your sarcasm but it's not true that the .13 won't scale. I think that % wise it's scaling fine. It seems that the P4 will have scaled 50% going from .18 to .13. 50% for the Athlon is about 2.25 Ghz and it seems that we'll see those before year end. I think all you guys are concentrating on the process thingy too much even though we witnessed delays with everything past the 1.4Ghz Thunderbird including the .18 Palominos. To me it's more scary that they have delays due to organizational issues than process issues.

C

Nitt, I don't know. There are tons of Hammer boards waiting for CPUS. These boards have the latest feature sets on them unlike the Intel MBS for the P4. Of course these boards are not available for purchase yet but they were due in Q4/02, Q1/03 last time I looked. So, the support is there for Hammer, it's AMD that's holding everything up.

C

wbmw, yes, I am "Constantine". Anyway, all of AMD's products that I mentioned are not "powerpoint" products but real products that have been demonstrated at numerous venues. The only valid question that you have is whether AMD can manufacture these things. My guess is yes. As to the reason why they have been floundering around the last 18 months or so, I don't know. Every single product introduction in the last 18 months has been delayed with no explanation given. It seems to me that must have some internal bs going on that we don't know about. It can't be something as simple as "their .13 is broken" because the delays were widespread and started way before the .13 process.

C

Greg, thanks. I will be taking it easy on this board like I did on the other board.

C

Semi, thanks, I wanted to check up on you guys since the SI intc thread has been overrun by all the intc shorts. To get back on topic, I think that AMD & Intc will do well in '03 unless we have a double-dip recession. All the computers bought in the late 90's will be breaking down soon and fixing them will be more expensive than purchasing new ones.

C

John, you and wbmw must be having funky (I don't want to use the other word ) dreams about AMD going out of business. Sure, it's a possibility but I don't think that all the work that went into the Hammer series will vanish like this no matter what. There is x86-64, HT, up to 32-way multiprocessing, on board memory controller ( maybe dual core ), a bunch of partners (motherboard and OS) having the infrastructure ready. All that's missing is the actual CPU. If and when AMD manages to open up the spigot it wouldn't be long before they're making money with Hammer. I am expecting AMD to either cust costs immediately or their board to fire Ruiz for incompetence. In the case of Ruiz getting fired it would mean that chaos has taken over and AMD will go belly up.

C