Slashdot Mirror
AMD's DDR-Capable 760 Chipset Reviewed X3
An unnamed correspondent writes: "The Tech Report has posted a review of AMD's 760 chipset. This is the one that includes a 133 MHz DDR bus, with support for 133 MHz DDR (a.k.a. PC2100) SDRAM. Benchmarks were done using a 1.2 GHz Athlon, and include everything from memory bandwidth tests to a variety of Quake III scores; they even attempted Linux tests, but Linux and the 760 wouldn't play nice." For another point of view, Fr0child writes "Today is the day that AMD officially announces their DDR SDRAM supporting chipset, the AMD 760. They promise "Increasing Memory Data Rate by Up to 100 Percent," which is quite promising to say the least. Of course, who would sit back and believe what a manufacturer says without verification? Anandtech has taken an in depth look at all the performance and features of the AMD 760. Looks like the combination of DDR + Athlon easily topples the RDRAM + Intel platforms out there."
And on the other, other hand, romeomustdie writes: "According to this [Sharky Extreme] piece, AMD is finally debuting the 760 DDR capable chipset, which is, for the most part, an evolutionary step up from the 750 chipset which has been out for the past year. Boasting a faster system bus, support for DDR memory, and a brand new South Bridge, AMD has set themselves up to not only surpass their first-generation offering, but also the current performance Athlon chipset, VIA's KT133. DDR is finally here to stay."
This is a popular misconception. The granularity of RDRAM makes it an attractive solution for Sony, and at the price that Rambus must sell it to them for, it must be almost as cheap if not cheaper than if they had to go out and buy SDRAM.
l ink</a>
Also, for the kind of work that the memory does, RDRAM trumps SDRAM. Read about how it works into the PS2 model on Ars. <a href="http://arstechnica.com/paedia/3dtech.html">
OMG, Sony is using RDRAM for what it is supposed to be used for!!!
/me faints.
-inq
The 760 is NOT at fault. The Linux error he reported is a known bug -- see Red Hat bugzilla ID 19535 (I added the workaround this morning, after realizing that we hadn't put anything up about this on our website). It is basically the kernel trying to disable the P3 serial number -- on the Athlon processor. (oops) It only affects Thunderbird-core Athlons and Durons. It has been fixed in all 2.2.16 and later kernels. I emailed damage to let him know about -- hopefully we'll get to see some benchmarks.
After installation, at the LILO boot: prompt use this command:
linux x86_serial_nr=1That'll get you booted, and you can upgrade/recompile/append to lilo.conf from there. There was supposed to be a Gotcha added for this, but obviously it isn't up yet. I'll track things down today and get it added to the Gotchas page for 6.2
Aetiushell it might be the only hope of having netscape render tables!!!!
HHell let make things more interesting and start up staroffice!!!
Non-Deterministic Finite Automata
>DOESN'T WORK ON [ALL] ATHLON CHIPS[...]
:-)
...ETS
A quote from the webpage:
"AMD Athlon processors with certain chipset combinations will cause problems for the Dazzle Digital Video Creator II"
What matters: certain chipset combinations. Not the processor.
I think what they are trying to say is that one of the current Athlon Chipsets (either the AMD one or the VIA one) is not compatible. But the Athlon CPU itself is perfectly compatible.
>And no, this is not a case of bribed or lazy programmers favoring Intel or "not writing in AMD compatibility".
It's likely a case of poor hardware. Either the chipset is at fault (which is what I'll believe, I know how poor the VIA KT133 chipset is firsthand... I owned a Vortex 2 soundcard and an ATI Radeon, ouch!) or the card is.
>AMD itself doesn't know what's going on
Only if the AMD chipset is the broken item. Not the processor.
>But I simply cannot afford to "take a chance" that all kinds of stuff I need to do won't run on "mostly compatible" AMD chips.
I have only ever heard of a single "major" error in AMD CPUs (there may be more...) that affected the K6 CPU when a jump larger than 32 MB was executed.
I have also heard of major errors on Intel's part, making them incompatible with Intel specs (haha) or just broken. Like the FDIV bug and the F00F bug for example.
You take a chance on whatever you buy. If you want a totally error free processor, you need to buy a 486. That way any errors that exist are very likely known and accounted for (and workarounds enabled). Why do you think NASA didn't put a PIII in the Hubble Telescope? Because it isn't tested well enough.
But don't take my word for it. You really have to decide for yourself.
If you could be told what you can see or read, then it follows that you could be told what to say or think - BoC
There's another review here as well.
Let's hear it for the onward march of technology! Marketer-speak aside, aren't we all just glad that our wonderful machines are going faster, going farther, and doing more? I for one remember my first computer, a TI-99 4/A, with a tape cassette as it's read/write drive. It was a wonderful thing, and lasted until my mother threw it out when I went to college. Now I sit in front of a Penium-2 on top of a nice fat DSL line, with winamp going, a linux partition just a dual-boot away, and my drives are full of luscious games and applications that would have exploded my good old TI-99/4A into so much hamburger if it had even thought about running them. Where will we be 10 years from now? I don't know, but I was at a Zellers [Canadian sort of Wal-Mart, I'm an American-in-Canada, BTW] this past weekend and there was a glorified 3D snazzy-background Pong-clone for sale, with big plastic paddles. A few decades development from it's birth and it's now -3D- [with goofy controllers] Pong :)
In July O7, I got a mac pro. There's no punchline. Just endless joy and wonder.
SMP support. I know with the 760 chipset (a revision of the chipset at least), is supposed to have support for 2 processors. DDR is great and all, but when you can scale to 2 1.2 GHz Thunderbirds, imagine the power you can have. You could have 2 Netscape windwos open in Gnome, each running on a seperate processor. It will double the chances of getting that f1r5t p05t on /.
Its not what it is, its something else.
Its not what it is, its something else.
Red Hat didn't add that feature, Linux 2.2.x post CPU ID fiasco includes a feature to disable the unique ID during boot, before userspace apps have a chance to read it.
Unfortunately it triggers for non-Intel hardware too, so later revisions (x >= 16) check first. Kernel boot line overrides it (see previous posts)
This bit me on a hand-built kernel with same old disks but a new MB, Duron 700, etc.
Booting a rescue kernel buys you an opportunity to check the solutions database, find the magic bootline and install a fixed kernel. Bingo.
Temporary copy of NT4 survived this switch (but doesn't do ATA66, USB, etc.)
On a friend's upgrade Win95 (98?) went haywire and now refuses to see USB or work properly with his Dual Head G400.
Actually Rambus isn't hard to make except in the higher grades. Since the entire stick has to be fabricated including the attached heatsink before it can be tested for speed, you get a lot of lower grade stuff and not much higher grade stuff. Where Rambust fails is in the noise immunity when it comes to making chipsets, thus Intel's problems with their various efforts.
Misinformation is running rampant here lately...
REDHAT shipped with the "Disable CPU serial number at boot" kernel option enabled. disable_x86_serial_nr=1 passed to LILO gets over this.
On with my rant.... just because REDHAT jacks something up does NOT mean tht Linux is fucked up.
Sorry abou tgetting all heated but really people use your heads and a search engine every once in a while.
Jc is reporting that Dell pulled their P4 system out of a "shootout" vs the DDR Athlon systems after looking at their performance. He also has numerous other links to 760-land. It is really looking like Intel had better dump Rambust and get with the program if they are ever going to sell anything with the P4. Likewise, Dell had better re-consider its Intelicide policy and start making AMD machines, especially when the multi-processor version of the 760 goes commercial. They have held the server market because Intel was the only multi-processor game in town. This is going to change soon bigtime.
If you really want the full DDR SDRAM promise then wait for the Mustang. It should be out shortly and addresses some of the Thunderbird's shortcomings. The combination of DDR, 760 and the Mustang has been the goal. DDR for the Thunderbird is merely introducing the new technology, the Mustang will take full advantage of it.
--
A feeling of having made the same mistake before: Deja Foobar
This is about DDR being the abbreviation for East Germany, and BRD West Germany, isn't it?
I'll remember to set my chip to Eastern-bloc mode.