Slashdot Mirror
Athlon 64 Pushed Back to September
Orion writes "AMD confirmed today that their new Athlon 64 will indeed be pushed back to September. Originally planned to be released in April or May, AMD has decided to put all of its brainpower into the launch of the 64-bit Opteron, which is still scheduled to be released on April 22. This article explains that AMD is still going to try to get a few more Athlon XP processors out before the Athlon 64 hits stores. The 3000+ has a planned February 10 release date, and the 3200+ should be out by the middle of the year according to the article."
Also that AMD will not release until M$ is ready. The should release for Linux, but want to keep us hanging on as Intel's grip on the market tightens.
Did you even read the article?? Opteron is still scheduled for April 22. It is the release for Linux.
The XP lines "model number" indicates the equivalent Intel P4 processor speed (because, as well know, the Athlon, like the P3, achiees more per cycle than the P4 does. A 3Ghz P3 would absolutely stomp a 3Ghz P4), so if they effectively upped the speed by adding more cache, then it's entirely the right thing to do given the philosophy of the model number.
a tant-dupes is bad enough, but these technical issues are just astoundingly amateur for what is one of the larger websites out there.
One thing I really respect AMD for is how conservative they actually are with their "model numbers" : The XP 2800 actually trounces the P4 2.8 on most benchmarks, and slaps the 3.0Ghz around on several. If they took the Cyrix tact they would have called in the XP 8200+.
SLASHDOT NOTE: This is a repost because Slashdot lost track of my last comment. What the hell is going on with Slashdot? Not only has it slowed to an absolute crawl, but every couple of days some other poorly thought out alteration to the UI appears and then disappears, comments are being lost after they are posted, etc. The whole editors-not-reading-their-own-site-and-posting-bl
Not being very in tune with hardware, I am curious as to the advantages of a 64 bit processor, and why I as a consumer, would want them... Could anyone please post a bit of information about them or a link to further reading?
For AMD it's simple, product differentation and market prestige. AMD is in a position that they always look like they are feeding off of Intels table scraps. This is an opportunity for them to establish themselves as a tech. leader and not simply a me-too company.
That and the fact that the margins on the new processors will be significantly higher than existing chips, a much needed boost in revenues.
First of all it's much more than just a 64 bit processor. The design of the new processors, originally named sledgehammer and clawhammer, will release a whole new era for multiple cpu systems. In current Xeon systems you recieve little bang for the buck, of adding processors, after the second processor and you go to third, the processors start doing redundant processes because of memory cache issues. But with the new AMDs, each processor will have it's memory contorller built right into it. This means that as you add processors you actually add memory bandwidth, look into this and it will brighten the sky and put a smile on your face.
Actually, there are several applications, albeit specialized ones:
:)
:)
1) Databases - a lot of databases are too big for a pointer offset to fit into 32 bits. Ever notice that the 120Gig hard drive you just bought has more than 2^32 bytes on it? (yes - I know that the hard drive is split into 512-byte sectors, and that you won't overflow 32 bits until you get drives larger than 2 TB, but how long will that take
2) Video (editing, encoding, etc) - a single layer of a single side of a DVD is more than can be addressed by a 32-bit pointer. The amount of source data used to create the highly compressed DVD data is mind-boggling. (A high quality transfer from film is about 100M per frame. A 2-hour film has 172800 frames [assuming it's not IMax - that's higer resolution and more frames per second] - that's 17 terabytes of raw data!)
3) High dynamic range images (including photographs and extrme high color video games) - the data types being used by the GeForce FX (similar to the EXR format released by ILM) have 16 bits of data per channel - this totals 64 bits for each RGBA pixel.
I'm sure there are more - these few just jumped into mind quickly.
Of course, for those who use Windows, you'll need 64-bit CPU's to be able to load those Word XP-2004.Net documents
- The Sigless Wonder
Is that how many hertz it has? I still have a 333 and it works just fine. Can somebody tell me why I'd need a higher numbers of hertz?
... the chips are generally a little faster than the number would indicate. The actual chip probably runs at 2.5 GHz or so (my 2400+ MPs run at 2.0 GHz).
3000+ means its integer and floating point calculations are roughly on par with a 3000 MHz Intel P4. AMD has actually been quite good in being conservative with these numbers
As to why you would want a faster machine, well, that depends. If all you do is surf the net, read email, run gnucash to balance your checkbook, and do a little word processing then you're probably fine with what you have. If you have the misfortune of running Microsoft, you are quite possibly better off not upgrading, given those tasks.
If, on the other hand, you like to render animations using blender, povray, or what have you, or like to capture and convert video footage (cinelerra, kino, dvgrab, and transcode), or enjoy running an optimized, source based distribution such as Source Mage or Gentoo, then being able to compile your entire system, complete with open office, kde, mozilla, and so in in a few hours, rather than a few days, is kinda nice.
All that having been said, my firewall remains a Gentoo box on a k6, so older, slower hardware is by no means worthless with GNU/Linux around, even if the newer, sexier, faster hardware really shines under FreeBSD and Linux.
The Future of Human Evolution: Autonomy
1) Cool stuff at the kernel level. No more high-mem (to address memory beyond the 1-2GB that can be mapped into the kernel. Single address space OSs. Persistant object model OSs (which go well with the new database FSs coming out). Finally making mmap() useful again for 2GB+ files.
2) 2GB (pretty much the max for a 32-bit machine) of PC2700 is $300. How long before desktop machines are coming equiped with that much, feeling the hurt for more memory?
A deep unwavering belief is a sure sign you're missing something...
The Opteron's debut is set for April 22nd .
They say it'll be out in a month... my guess is you'll be able to by the first one off the production line when you get your Opteron and Duke Nukem Forever all on the same day.
Help Brendan pay off his student loans
http://www.digit-life.com/articles2/amd-hammer-fam ily/
about opteron and athlon64
So, in other words, right now a 64 bit CPU is not needed for this. Why did you list it?
Uh... sure it is. Right now you can't easily address a file >2 GB on a 32-bit CPU. Doing so requires a file pointer larger than 32 bits in size (most vendors go to a 64-bit unsigned int, but implementations do vary) and that causes a pretty dramatic slow down on a 32-bit CPU. A 2 GB DB may have been large once upon a time, but it's trivial nowadays. Medium sized databases are in the hundreds of gigs, large in the terabyte range, and some of the biggest are pushing a petabyte.
Moving a database (or any other large file I/O heavy operation) to a 64-bit CPU can dramatically improve performance for this reason alone.
I believe the original poster said common applications
I believe watching a DVD on a PC is becoming increasingly common. HDTV on PCs isn't too uncommon, and HDTV dumps make DVDs look puny - even when compressed. Video editing is becoming more common as well, which utilizes both large files and can take advantage of the larger operations on a 64-bit CPU.
probably done on a unix system with a 64 bit cpu already
Yes, as are all of the applications... and it only costs 100x as much for a slower CPU. The point is that x86-64 will bring 64-bit computing to an entirely new price point - you'll be able to build a fast 64-bit PC for less than the price of a single 64-bit chip from Sun, IBM, Intel, or HP. That's pretty significant.
Honestly, there isn't much need for a 64-bit desktop CPU. But there isn't much need for a 2 GHz desktop CPU either. For those that can take advantage of the higher bit width, or speed, or both, the improvements are indeed massive.
You have to use a signed int because you can do seeks with a negative offset. So you're limited to a 2 GB seek space, and virtually everyone uses the same type for the seek offset as they do for storing the actual pointer into the file.
A few vendors do it differently, but you get a performance hit for it. And, frankly, it's just not worth it - if a 2 GB file isn't big enough, a 4 GB one isn't likely to be either. Go for 48-bit or 64-bit and you're much better off. Most vendors go 64-bit for simplicity and future expandability.
In this case, the "64-bit" really refers to the size of a virtual address, and not a lot more. A 64-bit processor (and an OS that can use it) allows problem sizes to grow past the 32-bit address limitation. By itself, a 64-bit processor doesn't buy you any speed improvement - in fact, it can be a bit slower than an otherwise identical 32-bit processor because doubling the address size comes at a cost in chip circuitry and OS overhead. There's also a memory usage tradeoff - if you're storing lots of addresses, it takes twice as much memory as it does for 32-bits.
I've been through the 16 to 32 bit transition as well as 32 to 64 (ten years ago!). While the marketeers love to tout the "bit-ness", what has really happened each time is that the processor architecture was redesigned to accomodate new technology, and this is what delivers overall performance improvement.
Where a larger address space really shines is in VERY large applications, such as relational databases, weather modelling, etc., which have to use disk segmentation to work around the limits of a smaller address space. This is why 64-bit processors are much more important for servers than for desktops.
As for 128-bit, etc. - I suspect you are looking at some specialized processors which operate on data that size, not virtual addresses. Will we move to 128-bit someday? Perhaps someday - after all, software expands to fill the available address space (Windows certainly demonstrates that!) Once 64-bit becomes mainstream, I expect it to not be supplanted for at least a dozen years.
[snip: about the limitations of the built in memory controller holdning the Athlon64 back]
AFAIK the Athlon64 memory controller can easely be bypassed by chipset manufactures. So Athlon64 and Opteron motherboards with RDRAM or 400MHz DDR are possible (though unlikely). Besides that 333MHz DDR memory is standard, the SIS 755 chipset support 800MHz FSB speeds. The "Athens" version of the Opteron will include onboard 400MHz DDR II support.
The biggest problem with the Athlon64 cpu is, that I can't buy one until september.
In the meantime I will drool over the system from www.newisys.com : dual Opteron, onboard PPC cpu running Linux; http, ssh, ssl for management, dual channel u320 scsi w/mirroring (LSI logic with ARM cpu?) hotswap drives, all packed in a 2U casing.