Slashdot Mirror
i820 Chipset Under Recall
Dman33 writes "This Cnet story details how Intel hit another bump in the road with its i820 motherboards. This defect is in the memory translator hub which allows for the use of standard DIMMs as opposed to Rambus. Intel is planning on just replacing the standard memory with Rambus memory, but will replace the entire board at the user's request. " The estimated cost would put a big hurt on Intel's bottom line -- several hundred million dollars worth of it.
I agree. I think much of the slamming of Intel comes from people who think of them as a "Big Evil Corporation", just because they are large and have a large market share. They might be big but they are not evil. I think many of us use AMD chips simply because they are an alternative to Evil Intel. People like to side with the "little guy" even if it means some sort of sacrifice. I'm not saying that choosing AMD is a bad thing or that Intel is better.
Intel is doing a very good thing here and it's gonna cost them a fortune.
I'm not trying to flame anyone here...it's just how I see things.
Check Out Knexa.Com
Check Out Knexa.Com
KNowledge EXchange Auction
Read more carefully. This is a problem with SDRAM being used on a board originally intended for RDRAM, there is no problem at all with using RDRAM on an 820 board. Of course, there's nothing wrong with the SDRAM either, it's the memory translator hub slapped on after the fact.
As for Athlon's being a good upgrade path. Suddenly many of those who bought slotted Athlon MBs (which were fairly expensive for motherboards)are discovering that they must buy new boards again for the upgraded Athlons coming out over the next several months because of trace problems with the Slotted Thunderbirds.
It sounds like both companies are having trouble keeping things user-friendly at higher frequencies. Personally, I'd rather have 4 chips running in parallel at 500mhz than one struggling at 1.2ghz
THE YEAR WAS 2081, and everybody was finally equal...
June 1999: I start saving up every cent I can so I can replace my Pentium 166 (replaced after the FDIV bug discovery) workstation. I'm looking forward to a Camino (i820) based motherboard.
June 1999: i820 pushed back again.
Dec 1999: I place an order with CDW for an Intel CC820 mobo and an Intel Pentium III 533EB processor.
Jan 2000: RAM, Case, cooling, power supply, and everything else I ordered arrived weeks ago. My products, which were supposed to arrive in 3-5 days are pushed back "until further notice". Calls to Intel provide no further information.
Feb 2000: It's been 10 weeks. I place a call to CDW to be told that I'm still on backorder. I place calls to Intel to no avail.
Feb 2000: It's been 13 weeks. I call Intel to merely be blown off. I file a report to the BBB in regard to their conduct and how poorly this situation is being handled.
Feb 2000: 3 days later I receieve a phone call from Intel, gravely apologizing for the ordeal. My motherboard arrives the same day after receiving an apology call from CDW in which they explained they were shipping CPUs and motherboards to 'higher priority' customers during the processor shortage.
Mar 2000: 2 weeks later the BIOS fails. I suspected it when PCI steering and DMA went awry initially, but I had gotten it to work.
Mar 2000: My new motherboard arrives, and works fine. News stories surface on yer more problems with the i820 chipset.
May 2000: My motherboard is under recall.
Now seriously, what am I supposed to do here? Send in my motherboard for a replacement that'll prolly take several weeks to ensure I won't have any problems? And what about the memory? Do I have to send them my 128M stick of PC100 from Crucial to get an RDRAM replacement? Even then, is the memory I'm going to get in return quality memory? Is it PC800, or am I going to be given PC700 or PC600 (which won't work!)?
I've been quite patient through this and I've put most of it behind me and just enjoyed my spiffy new hand-made workstation. But now something else? This is ridiculous, and Intel BETTER be kicking themselves in the ass right now over wht is arguably their biggest goof ever: i820
I plan on getting a 1.5GHz Williamette when they hit in the planned Q1 2001, and that better well not have a single problem in it, as the motherboard it'll sit on.
Sorry if this has a bit of that "fresh rant feeling" to it, but I'm just a tad irritable over this. >_<
--
I haven't used Junkbuster. Can it be used to filter out this URL so that if I accidentally click on one of these stupid links, I don't go to it? An optional "blacklist" built into the browser would be a nice feature (and handy for work browsing).
Is there an option for hashed URL comparisons so that my blocked list can remain secret?
"It's overkill, of course. But you can never have too much overkill." - Anonymous Slashdot Coward
But RDRAM is the problem, insofar as Intel is ramming it down their engineers' and customers' throats with no regard for reality. Intel management has committed so strongly to RDRAM that they'll practically have to commit seppuku before changing their "strategy". They ought to have looked at RDRAM's risks before committing.
I agree that RDRAM can be made to work. But the reality is that RDRAM has numerous challenges:
I'm not knocking the RDRAM concept itself, just Intel's Dilbertian implementation of it. In fact, I think narrow superfast busses are the Wave Of The Future(tm). Imagine buying an extra 16 gigs of RAM, taking it home, and plugging in the optical fiber without turning the machine off. Ditto for hard drives, or monitors. Would be much easier, especially for the average person.
-- ;-)
Kuro5hin.org: where the good times never end.
Apparently the rumor is that the problem is in the motherboard and how it has designed around the i820. Perhaps other motherboards don't have this problem (or perhaps they have their own problems).
now we need to go OSS in diesel cars
It isn't.
Dell has never passed up an opportunity to trash-talk AMD chips. Their loyalty means they get almost as many CPUs as they want, while their competitors go begging. Athlon? Hey, I just bought one and am a shareholder, but any idiot can tell you AMD is at the 1% mark in the business arena where Dell is king.
Dell gets very preferential treatment for their loyalty, and they have chosen to keep it that way. If Michael Dell called up AMD tomorrow and asked to place an order, Jerry Sanders would load up a wheelbarrow full of whatever Dell wanted and wheel it over personally. So don't tell me Dell doesn't have a choice.
-cwk.
I run a small systems company that builds high-end gaming systems for educated consumers. Three months ago, we removed Intel mainboards and processors from our system lineup. The reasoning had to do mainly with price/performance issues, along with Intel's inclination to force you into a new mainboard everytime you want to upgrade your processor. I'm pleased to report that the 90+ customers who've purchased Athlon systems from us have reported no problems at all with their ASUS K7M and ABIT KA7 mainboards. And they all survived the ILOVEYOU virus due to our policy of deleting IE and Outlook from their systems and getting them started with Netscape and Eudora Light, but that's another issue.
Fascism should more properly be called corporatism, since it is the merger of state and corporate power.
But RAMBUS is moving towards more analog sensitivity. There may be fewer data lines but there are more discrete voltage levels and timing intervals between clock pulses to get screwed up by spurious signals. At least with pure 1's and 0's, ths signal is sufficiently quantized as to be readily corrected until the noise pushes it past the boundary. RAMBUS seems to be squeezing those boundaries ever closer. Lots of bandwidth improving technology has gone the way of more data lines, such as wide SCSI and gigabit ethernet. Generating noise vs. noise sensitivity seems to be the issue here.
One of the problems I do see in motherboard design is that the paths that data and addresses have to take between the various components in the layout is, in a large number of cases, not the shortest.
One of my motherboards, an Intel SR440BX, has a sinusoidal ringing pattern in sync with the horizontal sync pulses. The video has to traverse across the motherboard, unshielded, to reach the video connector. It's not actually getting any digital hash, but it is encountering the impedance variations along the way. Maybe this is just saying that motherboard designers don't get to (or can't) take all things into consideration as others (for example video card designers) could. Which end of a video card would you put the D/A conversion? Which end would you put the VRAM?
Sometimes I wonder if these digital designers were sleeping in Microwave RF Engineering class
now we need to go OSS in diesel cars
It's good to see that Intel is finally handling this issue. It's too bad they didn't announce this a few weeks ago, before I snapped my cc820 in half and mounted it on my wall, to serve as a constant reminder to avoid Intel products at all costs.
(And you think I'm kidding...)
MSK
Yep - Intel is contractually obligated to hype (pretty much literally!) Rambus thru the end of 2002, and to push it as their sole memory option. I think the reason Intel was willing to sign such an odd contract is that they wanted to move more motherboard value onto their CPUs (i.e. get bigger % of the PC pie), and perhaps (seems a bit naieve, though...) genuinely thought that SDRAM couldn't be extended further.
Not only is their the contract, but Intel have for a long time been propping their earnings up with capital gains from various holdings that they sell each quarter. They'd have missed estimates last quarter without this (i.e. that fat P/E you're paying for is buying you one-time capital gains, not recurring growing earnings!!!). Intel own a large number of RMBS stock warrants as part of their deal with Rambus, and need to push RDRAM to be able to turn a profit on these to help earnings.
Why not just interleave 2 or 4 banks of SDRAM DIMMs into 4-way interleaved cache on the motherboard? Cost: 216 more data lines and a few more control lines, plus the larger cache (because there's really 4 caches), plus the fact that you have to have 4 identical DIMMs in place (2 for 2-way mode). How much would this design, with 512MB of RAM organized as 4x 128MB, cost, compared to 512MB of RIMMs and the associated technology?
I declined to choose the 810 or 820 chipsets anyway because they don't support ECC (which I suspect might even be more important with RAMBUS). So I'll be using BX technology unless and until someone makes a K7 based motherboard with correct server technology (The Abit KA7 board doesn't reboot on interrupted power, so I'm avoiding that for my servers).
now we need to go OSS in diesel cars
When you send yours in for replacement, borrow some extra DIMMs and populate the mother to the max! They might replace the board in that case, but obviously it would have to be one that at least handles that much. Beware they might replace 4 128MB DIMMs with 2 256MB DIMMs on a board with only 2 DIMM sockets. Whatever number of sockets you do have, max them out with the largest DIMMs you can get. At worst case, you get your DIMMs back.
now we need to go OSS in diesel cars
Does anyone know why it is that Intel has been having all these problems lately? I mean, first RDRAM and the Camino chipset thing (where they put a piece of metal over one of the DIMM slots to keep you from using it), low yields on the faster P-IIIs, now this. Is it just a string of bad luck, or is Intel doing rushing production on these things in order to compete with AMD, or what?
... must be pretty big for Intel to put up with this crap. Sometimes I wonder if they went with RAMBus just to prove that they owned the market. Some peopl are so gung-ho Intel (and usually MS) that they'd buy TurdRAM if Intel sold it, with a Feces Translator Hub.
Gamingmuseum.com: Give your 3D accelerator a rest.
a while until picking up on new chipsets until the bugs are worked out... Like a few years later.
-- On the side of the software box, in the system requirements section it said "Requires Windows 95 or better." So I i
This defect is in the memory translator hub which allows for the use of standard DIMMs as opposed to Rambus.
Why can't they fix with the good ol' fashioned method of "reversing the polarity of the positron emission trail" or "modifying the deflector shield harmonics to emit a resonance"?
I mean, what is the problem! Are these people stupid or something?
It strikes me that intel's in a dilly of a pickle if they plan to swap regular pc100 dimms for rdram. What's to stop a person from claiming (maybe even legitimately) that they needed 2G RAM on their new mobo? Go out and spend whatever it'll cost for 2G of pc100, and then let intel swap it up to rdram? You'd make a killing!
Another application would be to buy several of these, as needed. Buy a boatload of ram for one, and when they replace it with rdram, use that to feed the others.
They weren't kidding when they talked about damage into the hundreds of millions. That rambus investment is starting to look more and more painful for intel shareholders.
This has gotta hurt. It takes a lot of guts for a company to eat a bunch of revenue like that, especially when they aren't doing so hot in other areas.
[Off-the-wall and highly improbable conspiracy theory]
Or maybe they are doing it purely to entrench RDRAM in the market.
Step 1) Oh no, that SDRAM makes things buggy. Let's replace it for free with RDRAM for ya! (Who wouldn't go for it?)
Step 2) Now that EVERY i820 board has RDRAM, more RAM companies make RDRAM, the price of RDRAM goes down, and the price of SDRAM goes up
Step 3) Revisionist history starts pointing the finger at SDRAM for being an unreliable technology, not the i820.
Step 4) More and more consumers scoff at anything that doesn't use RDRAM ("Sheesh, your RAM runs at a puny little 133Mhz! Mine runs at 800!")
Step 5) AMD is hurt by the fact that none of their CPUs run on a chipset that uses RDRAM.
Step 6) While Intel flushes cash fighting AMD, HP takes over the majority of the CPU biz. (OK, maybe not.)
Intel has a name, created by a gigantic campaign, and continued by a gigantic name. It is the name that makes Intel the industry leader, and they will swallow just amount any amount of money to keep that name.
Hopefully I didn't put any [] around my words.
Intel's engineers are humans too, they make mistakes.
The problem is, this isn't solely a symptom of Intel's engineers goofing... Intel has been trying to get Camino out the door for quite a long time now (I believe the first publicly announced roll out date was more than a year ago). If they can deliver millions of CPU's (which are arguably much more complex than a memory controller) why then, can't they deliver an 820 that works. Remember, this is a company with billions of dollars to spend and thousands of the brightest minds in the industry... They ought to be able to deliver a memory controller that works in a reasonable amount of time.
This latest 820 debacle is more of symptom of political infighting and a company collapsing under its own wieght. There's an awful lot of 'us vs. them' infighting between the various processor and chipset marketing/design teams. Trust me.
Just like they replaced all those Pentiums with the FDIV bug
...but only after a tremendous backlash from customers after Intel told them that Intel's engineers would decide on a case by case basis who would, wouldn't get replacments.
The problem has nothing to do with RDRAM
...excpet that it has EVERYTHING to do with RDRAM. Camino is an RDRAM memory controller, right? So... I don't know how you could say that this has nothing to do with RDRAM.
So please, guys, cut Intel some slack here.
No... they don't deserve it. They've screwed countless customers (think Dell and many other OEM's) with promises of Camino based motherboard delivery, and every time, they've managed to screw it up. Dell has taken it in the shorts for several quarters now due to Intel's inability to deliver Camino and Camino based motherboards. I'm no great lover of Dell, but its fortunes shouldn't be based on (or ruined by) the ineptness of a monopolistic supplier. Expect the governement to start looking at Intel again before too long... I do.
I remember Andy Grove once saying that Intel's biggest enemy was itself.... That's looking more true everyday.
Don't believe Intel's hype, and don't feel sorry for them. They work their engineers and production workers VERY HARD and VERY LONG for relatively low pay (like I said, trust me) and make up the difference with stock options. Well, if the company, as a collective, performs poorly (as Intel has over the past year and half or so), alot of the compensation that comes in the form of stocks options will devaluate and Intel's talent will look elsewhere for a company that can manage its own complexity and avoid political infighting an turf battles.
-t
RAMBUS is an expensive joke! The i820 can't even keep up with the "old" 440BX based solutions. Have a look at a March 8th article that goes into all the gory details. To quote Mr Pabst: "Rambus seems to be a rather nasty disappointment."
Not everyone will want to overclock a 440BX board to 133 MHz, but since so little hardware is actually overclocked, and even then only by 30%, you stand little chance of ruining anything. Just make sure you use a CPU and SDRAM rated for 133 MHz FSB.
If anything, wait for i815 to come out. This is supposed to be the "real" successor to 440BX.
Use Evolution instead of Outlook? Bewa
Intel's engineers are humans too, they make mistakes. But, when the mistakes happen, Intel takes care of its customers. Just like they replaced all those Pentiums with the FDIV bug.
Oh, and for all you RDRAM hatemongers: The problem has nothing to do with RDRAM; the memory controller used in specific motherboards was not shielded properly, causing intermittent lockups.
So please, guys, cut Intel some slack here. It's not like AMD is bulletproof, either (Athlon/GeForce, anyone?).
NathanSo, they make this chipset, and it's supposed to save you all this money, and lower TCO of workstations, by performing multiple operations.
It actually sucks, it's dog slow, etc - but it IS cheap, and it sells like hotcakes, and now a million or so peecees are around there using it.
Hehehe, sorry, I can't stop giggling at the irony of it, this chip that's made to save money is costing everyone out the butt.
In a perfect world, maybe some IT managers would sit up and say 'hrm, well, maybe we should not keep buying the cheapest piece of junk workstation that ibm/compaq/hp makes..'
Or maybe not.
--
blue
i browse at -1 because they're funnier than you are.
...which is why MS stamps theirs out. Intel has competition and has for some time, so they have to innovate the old fashioned way.
Of course, that means mistakes, rush jobs, recalls, and all the problems that go with them. I also tend to think that users are not seeing the bang for the buck out of these new RAM technologies. PC100 was a good thing, but how much farther can we climb in sophistication before there really is no visible difference in performance/the technology itself tops out?
In space, no one can hear you moo.
Doesn't Intel have to support Rambus contractually? I mean, isn't there some kind of deal between the two, where Intel has to use their memory and be seen supporting them in good faith?
If so, when this deal expires, watch out RMBS...
Your Working Boy,
No Linux here
No Linux, or just no X?
Will I retire or break 10K?
I would suggest that in that case, they'll probably replace the motherboard. They are offering to upgrade the RAM, because if the loss is the same, they want to start moving people to RDRAM... However, I doubt that they'll let you get away with rediculousness... they'll just ship out a new board.
Alex