Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Duh. What Mooly *thinks* AMD *will* have. For certain "standard" benchamrks. Reading comprehension troubles again?
Yeah, it's what Mooly thinks they'll have in 6 months, based on Intel's favorite benchmarks.
Keith, I guess Intel will be having a bad H1, if they feel the need to pump September/October performance now. Perhaps Intel sent him out in response to the analyst downgrades today.
His reference to "standard benchmarks" is telling, IMO, as well as his "what they have planned for H2". Well, plans change, and further, this line was particularly amusing:
"It will take at least a year and a half to two years to close such a gap."
Um... that's funny, because a few things like 65nm transition and a new core will be happening well before then.
Anyhow, that's some big talk from Mr. Eden. We'll see what happens come September.
gollem, my point is that without such data, one cannot look at changes in DC pricepoints an conclude anything about margins or "pricing pressure". It would be fine if he made "best guesses" but one needs to guess about changes in UNITS between SC and DC, as well as price changes, to draw the proper conclusions.
LOL! "Worse than wrong" about sums up wbmw's "power FUD".
It was complete nonsense, as mas said.
Sorry, but your post is completely wrong.
They're comparing against 4 SC Opterons. Weak.
That's a desktop part coming in early June. Taylor is what you want to look at.
14 week to 13 week comparisons, instead of only talking about the normalized numbers. Legal yet unethical and misleading.
Not for long(eom)
We still hear of shortages in chipsets, logic integrated circuit, etc. Some Intel chipsets are still on allocation, and according to some checks, Intel delivered only 70% of expected order volume.
What a weird post. So, as I thought, no real changes. Thanks.
The one piece of new info (other than a Q107 target for some new parts) is a 3800 X2 35W part.
You've provided no evidence of "pressure on CPU pricing due to competition" as opposed to the DC --> SC transition.
Ah, so you haven't done the analysis. That's what I thought.
There is no "shakeup" there (poor title), and what changes do you see?
There is more information about previously rumored things like the lower-power desktops, and the roadmap extends out further, but where are the changes?
125W FX TDP has been known for over 6 months now.
forcing AMD to give up many traditional price points that both companies held for years
A "price point" in isolation is meaningless. You have to know where the volumes are.
I still haven't seen you run any numbers that look at changes in both prices AND unit volumes at those prices, summing this up over all units. Without doing that, you can't draw any conclusions about "pricing pressure" or margins.
Unless your name is Hans Mosesmann, of course.
INTEL that had very different plans until very recently
So you know the mind of Intel? They couldn't have planned it all along?
And regardless, as DC ASPs are still well above SC ASPs as DC % grows strongly, there is certainly no pricing nor margin pressure to speak of.
vendors will have Opteron 185, 285 and 885 processors in stock on the March 7, 2006 launch date.
Good news.
Your post was still flawed, because you forgot about the extra week used in both comparisons you cited.
especially in combination with the sharper than expected price cuts
Sharper than expected only by those who don't understand that the initial premium for DC parts was... an INITIAL premium.
I'm sure those same people will be repeatedly "shocked" over the coming year at more "huge" price cuts in DC parts, as they become the mainstream.
units * ASPs = sc_units * sc_ASPs + dc_units * dc_ASPs
Sure dc_ASPs are falling, but they are higher than sc_ASPs, and dc_units are strongly rising as a %. And DC and SC costs are not that different.
You are talking about competing with Intel's high volume market
Uh, no. Clovertown will not be Intel's "high volume market" at any point in 2007.
Looks like Intel is dragging the market down. Again.
Don't forget the extra loads on the FSB. 1067 FSB probably isn't going to happen for Clovertown.
but at higher frequencies, the thermal curve gets pretty steep.
Once again, you're stuck in the past. There's these little things at 90nm coming, called "Rev F", and SiGe strain. The curve will shift to the right.
As it is, 2.2GHz may require an "SE" thermal envelope
Rev F parts are coming with a new higher TDP of 140W, in addition to the other thermal points, and this is supported by Socket F / 1207.
So, wrong again.
Secondly, getting a low voltage on an engineering sample at 1.8GHz is not a large feat. While I think AMD should be able to deliver 1.8GHz @ 35W eventually, the volumes may not be adequate for more than a "PR" launch.
Whistling louder?
You do realize what that voltage implies about power don't you?
And you did notice that Taylor will be offered at both 35W and 25W?
The 140W socket is for beastly 65nm parts, which I can see reaching 2.8GHz, even with quad core. On the 90nm side, however, I think AMD is for more constrained than you think.
Now that's funny. The socket is only for 65nm parts? Hmmm, nope.
Because of Intel's FSB & chipset-memory chokepoints.
mas, also the DDR1 memory controller power is not multiplied by 4.
wbmw is whistling ever louder.
In 140W, with DDR2 instead of DDR1, 2.6GHz QC should be fine at 90nm.
At 90nm, AMD will be lucky to get much over 2.2GHz.
Because you want it to be that way? LOL, again.
but it[Clovertown] will also clock close to 3.0GHz
Yeah, sure it will.
Be that as it may, it's doubtful that anything less than 65nm will pose a threat to Clovertown.
LOL!!! You've got to be kidding us, right?
It's doubtful that Clovertown would pose a threat to an AMD 130nm QC part, let alone a 90nm one, or a 65nm one.
Hmmm, desktop shows no Conroe (non-EE) speed increase from 2.67GHz launch through Q107...
1.075V @ speed on 90nm? Nice. SiGe strain must be working well. As far as the July thing, I belive MSI's roadmap showed a DC Turion notebook launching in May, with more models in July.
See here:
http://www.epscontest.com/roadmaps/othermaps.htm
Page 23 features the May Taylor notebook, with 24 and 28 listing a number of late June / July Taylor (DC Turion) products.
Intel's chipset-chokepoint memory designs force them to adopt immature memory technologies in an attempt to supply enough memory capacity and bandwidth to multi-socket servers. AMD's IMC just keeps paying dividends.
Intel is releasing desktop FB-DIMM systems? Really? When?
AM2 is the desktop socket.
Currently folks use up to a whopping 2 DDR1-400+ DIMMs per channel, and many use 1 DIMM per channel, as timings can be faster.
Somehow, I don't think there'll be much trouble supporting both 1 & 2 DDR2-800 DIMMs per channel. As in the DDR1 case, "command rate" may need to relax in the 2/ch case.
So, if you think all existing AMD desktops are "benchmark specials", then I guess you think the new AM2 desktops will also be "benchmark specials".
Of course, many of us think of them more as "revenue & marketshare specials", instead.
As Mike pointed out, (1 waterfall step = nothing more than x85s coming within a month) this is 1 extra step beyond waterfall. Still a significant premium over SC prices, but this is one step along the way as DC prices approach SC prices.
Why don't you lay out some detailed numbers about DC unit % vs SC unit % expectations, as well as ASP expectations in each, in desktop segment, say, to support your 'pricing pressure' hypothesis.
Uh huh. Well, during Friday Happy Hour, I'll take a look.
Search for "conroe TDP 2.8" or something along those lines, and you'll find it.
You and they are misinterpreting the transition from DC to SC, and the resultant price-cuts in DC parts. DC parts appeared at a large premium, but that obviously cannot be maintained as they become the standard. Unless you expected everyone to suddenly pay a lot more for their CPUs.
DC price cuts are matched by increasing DC volumes (vs. SC).
Not really pricing pressure. You've got a transition from SC to DC underway in desktop as well as server. Surely you didn't expect DC parts to end up with a higher ASP than SC after the transition is done?
Then you've got Intel trying to sell LGA (Last Generation Architecture) parts to OEMs and consumers for the next few quarters (and then some) leading up to Conroe/Woodcrest/Merom.
With Conroe requiring different mobos, (VRM issue), what OEM wants to sign up for a 2Q run of Presler crap without some incentives?
And what knowledgeable (yet Intel-lovin') consumer doesn't want to wait for the NGA stuff?