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.
Dan3, ? - how is this relevant to the discussion?
Paul
wbmw, my point is that there are problems other than "process" defects that can affect the number of die that are usable when you have particular requirements to be marketable, such as Banias' need for low power consumption. You could have a defect-free process, but if there's a design flaw or weakness, or maybe a speed path issue, like early Opterons appear to have had, then you're going to have much lower production than with a different design on the same process that doesn't have those sorts of issues. Again, theoretical discussion.
Paul
That is good news, Keith. eom
wbmw, I wasn't assuming it is production limited, just that the production of P4 doesn't necessarily mean the production of Banias is as problem free. It may well be, but you couldn't prove that by the fact that P4 is doing well.
Paul
EP, since they're totally different designs, couldn't issues come up such as high numbers of die with excessive current draw(in the case of Banias) that would render it useless? Or a different speed distribution curve that wouldn't match up adequately with what's needed? Again, this is all theoretical, I'm not saying those issues exist. Yes, I understand the equivalency in defect density, but that's not the only thing that would prevent a Banias from being marketable. BTW, I keep forgetting - what is the official name of Banias now - Pentium M? Without the wireless doodads, that is.
Paul
chipguy, not that it's necessarily the case, but I don't see how Intel's pumping out P4s necessarily has anything to do with their production of Banias. Different cores, different problems. Again, not that I think there's a problem, just pointing out the logic error.
Paul
wbmw, thanks, will do. eom
Paul
EP, chipguy, wbmw - Any thoughts on Intel's delays in the new roadmaps? No, I'm not trolling or tweaking, I'd really like to know what you think.
Thanks,
Paul
http://www.theinquirer.net/?article=12014
@$13.30, quite an opening jump.
Interesting that newegg has dropped the price of the A64 3200 by almost 10% from 448 to 408. Volume increasing?
Paul
http://www.newegg.com/app/viewproduct.asp?featureItem=1&description=19-103-413&DEPA=0
blauboad, very good question, especially since they didn't appear to say anything substantive about it in the article. If this is the case, there are going to be some happy high-end gamers. Since it really is essentially just a re-badged Opteron, perhaps they figured there's no harm done in allowing FX dual rigs - it's not like they're losing money selling those! And Socket940 is not long for this world anyhow. Good eye!
Paul
Besides, EP, SPEC has nothing to do with the target audience of these two processors - gamers.
Paul
Keith, Staples is one of the 3 big nationwide office supply stores, along with Office Depot and Office Max.
Paul
sgolds, I don't know if I can buy that. They(MB manuf.) said "projected sales". Are you saying the sales are to the MB manuf. to bundles with boards? 1.6 million seems a little high to me. Once they get to that level of production, I'd think the MB manuf. would be less concerned about availability. Unless, of course, they think it's going to be so hot of a product that they want to make sure it's their MB that gets the chip in it! ;)
Yeesh! 3.3GHz, 27s - Katsumi just keeps pushing it!
Paul
http://www.vr-zone.com/?i=161&s=1
http://www16.big.or.jp/~bunnywk/cgi-bin/superpi.cgi
This could conceivably become a useful source of unintended info.
Paul
http://forums.amd.com/
Fascinating. Perhaps Prescott's salvation? ;)
Paul
http://www.theregister.co.uk/content/3/33252.html
wbmw, forgive me, but between the things I mentioned and the insatiable growth needs for "bandwidth", I don't see it being underutilized for long. Sounds a lot like the "No one would ever want more than 640K of memory" argument to me. Just look at the increase in Intel's bus bandwidth over the past 3 or 4 years. If you think that storage is going to remain limited to the ruts it's been in - well, I doubt it; one way or another system bandwidth is where the most stands to be gained in overall system performance. I'll just say I'm glad to see they're thinking ahead, even if it should prove to be overkill for now.
Paul
gollem - I especially like the AMD "Northwood" chipset mentioned. I had no idea AMD and Intel had gotten so close. ;)
Paul
wbmw - the point? Throughput, of course. What are you getting at? With the upcoming PCI-Xpress graphics cards and various other peripherals (like Gb Ethernet), it will get used before too long.
Paul
chipguy, since that quote was from the middle of a section about SiS support of K8, I suspect they will be, though that's not explicitly stated. Besides, I haven't heard anyone say nVidia is having trouble with 800MHz, I think it was just a stupid, failed market segmentation attempt.
Paul
HT bus to be bumped up by 25% soon?
"Roadmap for AMD processors is not so broad, as for Intel’s chips, but is still quite interesting. First of all, we should note that AMD will pump up the frequency of HyperTransport bus to 1000MHz from current 800MHz next year. This will result in 2000Mb/s transfer speed in each direction. SiS’ 756 and 761 PCI Express x16 tunnels will support AMD Athlon 64 processors for Socket 754/939 and Socket 754 respectively. The SiS761 will also sport that mysterious DirectX 9.0-compliant graphics core from SiS, it transpired."
http://www.xbitlabs.com/news/chipsets/display/20031005150819.html
neye_eve - one would think so...
neye_eve, it may not be nVidia's fault; may be more the fault of the MB technical writers. I haven't seen one yet that clearly and specifically covers this. In any case, we figured it out.
Paul
Jules2, re dual-channel. What you posted is correct, assuming that the DIMMs you labeled as 1 & 2 are the DIMMS that share a channel. What we were getting at is a situation where all 3 DIMMs are populated. Dual-channel is possible in this configuration as well, as long as the conditions I detailed are met.
Paul
Edit: posted before I saw your second post on this subject, where you realized the DIMM 1 & 2 slots were on the same channel.
gollem, check here:
http://www.investorshub.com/boards/read_msg.asp?message_id=1537970
Paul
InternetPlay, Mike was mistaken. I read the Hexus article first thing this morning, and it didn't say anything about the P4 EE, it said that Intel would release a 3.2GHz Xeon w/533MHz FSB on Monday. As a matter of fact, the story is still there(scroll down a little):
http://www.hexus.net/
Paul
sgolds,CJ,gollem - re: dual channel. I've been scouring trying to find definitive word on this, and it's a little hard to get. But here goes with what I've found. Yes, you can use 3 DIMMs and still get dual-channel, as I remembered - with these limitations:
1)The 2 channels must be balanced as far as their total memory sizes(e.g. 2 256MB in the one channel with 2 DIMMS, and 1 512MB in the single DIMM channel), and they should be of similar, preferably identical design(other than the size).
2) The total number of banks must not exceed six.
I have read a number of posts from guys at nforcershq that verify this. Info from a number of other sources could be interpreted either way (Anand, digit-life, etc.). I'm satisfied with the guys who are looking at their boot screens and know what they have in their systems. It does tell me that if I want to upgrade, I'll be getting a 512MB Corsair 3200LL module and putting it in the lone DIMM channel.
Paul
gollem, w/ 3 DIMMs you'll still have dual channel, but you may have to reduce the frequency or increase the timings somewhat because of the additional capacitance. The 1st nForce chipset automatically went to very conservative timings with a 3rd DIMM because of stability issues.
Paul
Socket 754 mobile Athlon XP "Dublin" info.
http://www.xbitlabs.com/news/mobile/display/20031005034354.html
Keith, ULi must have a very low cost structure, since they are offering the K8 chipsets at a lower price than their competitors and still are excited about the margins. If they can do it right and offer the necessary features, they have a great opportunity to grab a good chunk of this market.
Paul
Keith, welcome back! Nice vacation? I was beginning to wonder if you'd dropped off the face of the earth! ;)
Paul
neye, just curious, why didn't you go for the low-end Barton? $84 shipped from newegg.
Paul
spokeshave, yes I found the protestations amusing as well. From more than one person.
Paul
Thanks, Klaus! eom
gollem, although it could be stretched this way and that way, I'd suspect they were referring to 90nm in 1H04, not the dual core.
Paul
sgolds, perhaps some of the native German speakers can verify, but I believe it means that the beta behaves more like a pre-beta. Which is, I suppose, par for the course for MS, since many consider their releases to be more like mass betas.
Paul
Some initial Win64/A64 benchmark results.
Paul
http://www.tweakpc.de/hardware/tests/cpu/athlon64_fx_64/athlon64_fx_64.htm
sgolds, I am curious how well they'll be able to deal with various hardware variations. Particularly video cards and network cards. Of course, if they cover nVidia Detonator drivers and ATI Catalysts, that would cover the huge majority of the graphics of those interested. Hmmm.. does ATI have Catalysts for Linux yet? In any case, it seems like a pretty cool idea, if they can do it right!
Paul
gollem, I can't give you an exact figure, but seems like I remember ca. 3% difference in Sandra bandwidth results without involving integrated graphics. Not much to worry about, the FSB is more the limiting factor. Now, if you involve the int. graphics, it makes a quite substantial difference. If I remember correctly, Anand may have had something about it. I imagine Johan did, too.
Paul