Slashdot Mirror
AMD Launches Athlon 64 FX-57
Kez writes "Today AMD release what could be the fastest x86 processor to date. The FX-57 is the first 90nm Athlon 64 FX from AMD, clocked to 2.8GHz, with 1MiB of L2 cache and support for SSE3. The memory controller has also been tweaked to support mismatched memory module sizes - something some enthusiasts have been crying out for. Hexus.net reviews the new processor, which, in gaming benchmarks, walks all over any of Intel's offerings." There's going to be plenty of reviews I'm sure - if you've found other links, please post them below.
Probably the best place to look for info:w s&file=article&sid=3165&mode=thread&order=0&thold= 0
http://amdzone.com/modules.php?op=modload&name=Ne
No gods, no demons, and no masters. Secular Humanism!
Toms Hardware Review
Coral Cache: http://www.hexus.net.nyud.net:8090/content/reviews /review.php?dXJsX3Jldmlld19JRD0xMzE3
Because the market for fuel cells for your mouse isn't driven by the gaming industry, while the processor industry (at the high end, as this CPU is) is. :)
No gods, no demons, and no masters. Secular Humanism!
What is 1 MiB of L2 cache? 1 Million Bytes?
It's the binary count of bytes, 2^n, instead of the decimal *10^n
Let the commencement BEGINULATE!
SI unit for 1 megabyte (1024) is MiB, 1 megabyte (1000) is MB.
SI units were changed back in the 90s to reflect base-10, the base-2 units were renamed to use the letter 'i'.
For a definition of MiB see this wikipedia article -- http://en.wikipedia.org/wiki/Mebibyte
n 64/fx57/001.htm
And here is a random review that includes the actual cache numbers -- http://www.gdhardware.com/hardware/cpus/amd/athlo
If you are confused, This guide is pretty good at explaining the differences between the current AMD offerings.
The reason it costs so much is you just outfitted your rig with some seriously serious hardware. You went close to top of the line of your own volition. The top-shelf stuff will always cost you a premium - one that doesn't usually scale to the increase in performance - because you're going for best of the best.
Well AMD's production capacity isn't as big as Intel's - that's one of the big issues. AMD has an agreement with Intel to use x86, but this means that AMD may only outsource a limited amount of its production to foundries. But the new fab in Dresden, Germany will soon be opened to boost AMD's production capacity.
Intel also has big deals with huge system builders such as Dell and HP.
Venice.
The venice core is the new one. It's a few unimportant percent faster than the others. It's got a few not-terribly-important new features. It's rather more than a few percent less power-hungry. AMD doggedly sticks to its 89W max figure, but Venice maxes out at more like 50W, according to measurements / approximations.
So, to recap:
Faster
Better
Lower power
Nicely overclockable
Buy from a respectable retailer and they'll tell you not only which core you're buying, but the clockspeed, cache size, etc. rather than just the model number.
High-speed Road Trip (18.000KPH)
It has nothing to do with the smarts (or lack of) of the market. It has to do with their capacity and with marketing.
99% of the users out there don't even need a 2ghz P4, let alone a 2.8ghz FX-57. AMD has long been catering to the enthusiast market which is a very, very small fraction of the overall market.
Programs in the system tray are mainly memory hogs, not processor time consuming. After all those just sit there and wait for something to happen. And interactive programs like chats just don't eat to much processor time either, the average amount of data to process is dwarfed by the amount the game engine is crunching.
Pull up task manager and look at how much CPU usage you have when your machine is just sitting 'idle', not running a game or other CPU intensive app. If you have a higher end CPU, it's probably maybe averaging 1-2%. That's not a lot of stuff to toss on the other CPU. The FX chip will run at much more than 2% faster than the dual-core chips available now, so it's the better option for games (99.9% of which won't take advantage of 2 CPUs). In a couple years when more games are written for SMP systems, then the dual-core will be the way to go.
All the FX chips are Socket 939. The only chips that are socket 940 are Opterons. The Athlon-FX chips are basically Socket 939 Opterons.
Socket 939 is for the single CPU systems (including multiple cores).
Socket 940 is for multiple CPU systems (ie. Opteron).
The ratio of people to cake is too big
MB: Megabyte (1000*1000=10^6 bytes)
MiB: Mebibyte (1024*1024=2^20 bytes)
theefer
Also, the 940 boards have to be 6-layer (costs more), and they have to have ECC (IIRC), whereas 939 can be 4-layer, and use cheaper, faster non-ECC memory.
Here's all of the current sockets that AMD uses:
462: Newer Classic Athlons, Athlon MPs (server chip), Athlon XPs, Duron, Old Semprons
754: Old Athlon 64s, Semprons, Turion 64s (mobile chip)
939: Newer Athlon 64s, most Athlon 64 FX's (53, 55, 57), and in the near future, newer Semprons and 1xx (single CPU only) Opterons
940: Opterons, Athlon 64 FX-51 and some FX-53s
Then Intel will need two bolt of lightning.
The original FXs were 940s, However, the newer ones have been 939 (they might also have a 940, but I don't believe so.)
Also, it's not just one pin removed, the actual layout is different between 939 and 940. (940 also supports smp, that being the main difference.)
The big difference between 754 and 939/940 is that 939/940 support dual-channel memory. This is important, because if they didn't, while the athlon 64 is relatively non-memory bandwidth hungry (as compared to the p4), performance would suffer with two cores on single channel memory.
Well, here's another review:
:-))
tweakers.net
It's in Dutch but it has some nice benchmarkresults which should be clear to anybody, benchmarked next to an AMD dual-core 4200+ and a AMD 3800+ (and I know for a fact those results are correct cause it's all been benchmarked on my own desk
Check it out. Anandtech has a review as well.
AMD has limited production facilities. These "fabs" produce all of their semiconductor products. The FX57 is currently the fastest chip they produce, and typically yields on faster processors are lower than slower chips. AMD already has commitments to clients to supply certain quantities of slower chips as well. Additionally AMD doesn't want to devalue their slower processors by flooding the market with the fastest chips. These factors all boil down to the single fact that AMD can only produce a limited number of FX57 processors.
The basic laws of supply and demand dictate that there is an inverse relationship between price and demand. If we assume that AMD has a fixed supply then their goal is to set a price that will match demand to supply as closely as possible. If the price is too high then some of their chips will go unsold, and AMD will either have to stockpile or cut the price. If the price is too low then AMD will lose profit and have a backlog of buyers.