Slashdot Mirror
AMD's 'Newcastle' Budget Athlon64 Chips Analyzed
Edward Scissorhands writes "CNET News.com reported on Thursday that AMD had released a new "budget" Athlon64 CPU. Appearing on the AMD roadmap under the codename of "Newcastle", these chips are identical to the 754-pin Athlon64 3200+ in every way except for the size of their L2 cache (512KB vs. 1MB). CNET suggests that some of these chips may be 3200's that don't pass QA as having full 1MB caches. Newcastle chips are about half the cost of their 1MB cached counterparts, though preliminary benchmarks from Anand indicate favourable performance/price."
This is what many companies do. If certain chips cannot pass Q&A then remark them down and you do not lose your inventory.
cheers
Rick
http://www.hardocp.com/article.html?art=NTYw
Kyle of HardOCP makes a bunch of speculations as to AMD's purpose for releasing these chips, and comes to basically the same conclusions that CNet does.
He sugguests that these chips are also just the ones that only had partial working cache (a portion of the cache was working, the other portion was not) and to save money they are selling these as a "budget" chip. Seems like a good idea to me!
"You had this look that of an angel, it was such a bad disguise" --Dishwalla
There haven't been any official statements that their new processors are going 939-pin.. that is a speculation by a few highly-voiced individuals and off-the-record statements
"You had this look that of an angel, it was such a bad disguise" --Dishwalla
$213USD seems to be the lowest on pricewatch, for those who are wondering
http://www.pricewatch.com/1/3/5867-1.htm
well since they run 32 bit code natively and FASTER than the top of the line P4, I would say the shelf life is good.
I am the Alpha and the Omega-3
Overclockers.com did a mini-feature on the Newcastles last week, including why you shouldn't buy one too soon.
I hadn't known there were so many idiots in the world until I started using the Internet -Stanislaw Lem
I'm not sure, but I would think that as long as they could find 512K of contiguous cache that passes, they could use just that without any major modifications.
An excellent beer to namesake your chip with, though. Smooth, reliable, and makes you dizzy.
As it is the 9xx pin FX jobs don't really have that much of a reason to exist yet because the performance increase is marginal, maybe 5%. Sometimes dual channel actually slows things down by a percent or two. I hope that this changes before the 7xx pin version goes away. I imagine that at a higher clock the difference becomes more noticable.
The only reason to get a 9xx pin chip is to get multiprocessing in the form of Opteron.
The cache is allocated in blocks. There are 2 512k blocks, and if there are bad cells in just one of them, it is disabled and the chip is sold with only 1 512k block enabled. If there are bad cells in both, they throw it away.
I hadn't known there were so many idiots in the world until I started using the Internet -Stanislaw Lem
Obviously, to keep other sites and everyone else from "stealing" their charts. I know it's a pain in the ass, but Anand is a pretty large and profitable review site. If their charts were just JPGs, what's to stop some unscrupulous site from snatching their pics, changing the colors in a batch job, and reposting the results as their own? Granted, they could probably just screen-cap their pics now, but that involves a little more work.
Not quite. They're on the roadmap for Q1, which would just miss christmas at the earliest.
...has 5 in stock, and that's here in Norway even. So it might not arrive before Christmas, but if you hurry you'll probably still go 64 bit in 2003 :)
I'm perfectly happy with the PC I have though. Usually, whenever Christmas is nearing I get questions about whether I'd want something for my computer. This year, for the first time in as long as I can remember, the answer is "not really". No big itch to scratch... I have CPU, GPU, RAM, HDD enough, broadband, most everything really.
I must say, I'd still like to improve noise and style though. Performance, well it's not that critical anymore. But the noise is pretty bad, even after I replaced my WD disks with Seagate. And I admit a Shuttle XPC + LCD looks ten times better than my beige box. Maybe next PC, but that one is not now. Not for a while yet...
Kjella
Live today, because you never know what tomorrow brings
They can still copy the Flash files as well. Makes no bloody difference, expect they are using non W3C standards. Bad, very bad.
Have you considered the Antec Sonata case?
(And of course, the other options like quieter CPU fans, quieter exhaust fans, quieter power supplies. For my home office, 2004 is the year of "quiet", my goal is to make a serious dent in the amount of PC noise going on in here.)
Wolde you bothe eate your cake, and have your cake?
Here's example systems you could build, with the best possible motherboards. Each assumes you need to buy some DDR400 RAM so that is not included, since it's all the same:
My Athlon XP system:
Athlon XP 2800+: $150
Abit NF7: $100
Total: $250
I'm quite happy with it. Best price/performance choice (last week, anyway.)
Top-End Athlon XP system:
Athlon XP 3200+: $289
Abit NF7: $100
Total: $389
A complete waste of money, especially after today.
P4 3.2 system for comparison:
P4 3.2 CPU: $366
Asus P4C800-E: $164
Total: $530
Better than both of the above, but only by a few percent for most things.
That was the situation last week. Including an Athlon 64, Athlon 64 FX, or P4Extreme in the examples would have been useless since they're insanely expensive.
However:
Athlon 64 system you can build now:
Athlon 64 3000+ 512k cache: $230
Gigabyte GAK8VT800M: $106
Total: $336
Yowza.
So, to jump from the top-end Athlon XP to an entry level Athlon 64 actually saves you $53. I could have spent an extra $86 and got all this. The Athlon 64 system will now save you $194 over the best P4 Intel has to offer, and it will beat it (for virtually all applications.) Of course, if $336 is too much, you can still build a good Athlon XP system and cut costs dramatically, but $336 is very reasonable for building a brand new system. It'll be interesting when Intel gets it's P4Extreme down to a reasonable price, and AMD starts ratcheting up the Athlon 64 speeds.
Proves it's always better to wait just one more week. I should have known that there would be major cuts in the 64-bit world soon after the processor debut.
Hope all this is useful to anyone considering building a system. Keep in mind that 1gb of dual-channel DDR400 RAM is gonna run at least $150.
All prices are PriceWatch.com and the Athlon 64 CPU price is from a link on AnandTech. I know PriceWatch prices are hard to get and you have to deal with shipping and all that.
# Erik
Bandwith - the major cost of runing any hardware review site. By making a simple flash movie that displays a graph based upon passed in text data, they save bandwith. You download the flash movie once (or once every time they update it), and then everytime they display a graph your browser uses the cached movie, but with the new param set for what data labels and values to display. Sending even 15 lines of data is a lot smaller than sending a 300x120 pixel image for every single graph - especialy when a multi-page review may have 15+ graphs.
Yes, it sucks for the user who has to download whatever version of flash they use - and it sucks even more if the version they require is so new it isn't avail on all platforms, but they Do have a good reason.
man is machine
Ah yes, good ol' fud. CURRENT top of the line Intel processors run hotter and consume more power than top of the line AMD processors. You're a good 4 major revisions behind on your info. The Thunderbird was the last really hot Athlon (and it really wasn't THAT bad) They cooled off considerably with the release of the Palamino, and again with the Thoroughbred. Barton is basically the same as the Tbred, and the A64 chips don't seem to be running too hot, either.
Who cares how good the retail "hintsinks" are. They're pretty decent, these days, but anyone who cares will replace it with a better performing model. No different than with Intel CPUs.
You're entitled your opinion, but your post is pretty much just 3 year old FUD.
Not just Linux. Solaris, FreeBSD, and NetBSD have working ports already.
On FreeBSD, the AMD64 architecture is a "Tier 1" platform, which means fully supported. See here.
On NetBSD, the port seems to be basically functional and mostly done and is expected to be included as part of the next release. See here
Sun has announced amd64 support along with amd64 hardware products as well (Opteron entry-level servers). See here and here. Solaris 10 (which isn't out yet, but which does have "early access" releases available to the public) supports it too. Sun is rumored to be working hard right now on tuning its own compiler to generate good amd64 code. And, significantly, Sun has been selling 64-bit machines for nearly 10 years, so they have pretty much solved the issues of integration, making it easy to manage the combination of 32-bit and 64-bit applications. (Older Sun machines can run either a 32-bit or 64-bit kernel and will automatically select 32-bit apps on the 32-bit kernel but 64-bit (unless unavailable) on the 64-bit kernel with the same OS install image.)
All of which is good, because the x86 market tends to procastinate, and the 32-bit to 64-bit transition is IMHO a bit overdue. (And because it is possible that it will make Intel pee its pants if this really takes off.)
AMD's Athlon MP is their 32 bit dual processor capable offering, Opteron is their 64 bit dual capable. Athlon64/64-FX are NOT dual capable.
I don't know about this chip specifically, but for chips in general there would be more than just two blocks. In fact, it would probably be along the lines of 36 cells of 32K a piece. If any of these cells fail validation, you can disable them.
So long as the number of busted cells is 4 or less (just using my example numbers here), you can sell the processor with 1024KB of cache enabled. Obviously, if there are no cells that are bad you still disable 4 of them to keep the parts consistant (OEMs don't want one chip with 1024KB of cache, the next with 1056KB of cache and the next with only 992KB). If more than 4 cells are bad, you disable all but 16 of them and you've got a chip with 512KB of cache.
Intel and AMD both do this (as do others), and at various times both have had chips sold at different performance and price points depending on how much cache (or some other feature, ie hypertransport links, hyperthreading, floating point units, etc.) is enabled.
How the fuck? BTW, a NetBurst Celery has half the L2 cache, which doesn't sound as bad at first, but when you realize that the cache is what's keeping NetBurst CPUs from going into the gutter performance wise, the Celeries suck - just look at Anand's benchmarks of sub-$100 Intel and AMD CPUs (with a 1.8A P4 as a baseline) - except for the Duron 1.6, AMD CPUs and the P4 MURDERED Celeries. As for the Duron, it usually held it's own against the Celeries and 1.8 P4.