Thuban, the worlds fastest desktop cpu coming soon from AMD.

$295, nice....., would have though a little more than that

might have to pick one up to do some crypto programming on

EDIT: though i don't know about fastest (it is for my board at least)

aaaaaagh! i want to buy one, but i dont know if i should wait till Bulldozer to upgrade

Whoa look at the 1055T: 6 cores at 2.8ghz for $199. Not good for the i5.

Wow and I was about to get a Phenom ii x4 955... maybe I should wait lol.

Hmm I didn't really think about that, but yeah you're right the i5 can't compete with this.

It also looks like SOI did some good for AMD, as even though Thuban has 6 cores and still is at 3.2Ghz, it has a tdp of 125 watts.

Thanks. =B

Nope - April !!

They would get massively overshadowed by the release of Fermi if they were released on March 26th.

I'll wait for AMD 8 or 12 core.

I think the NDA ends on March 26th, while products will launch on April 6th. Paper Launch FTW!

Edit:

Is jennyh the new Thunderman? The post title says, "Thuban, the worlds fasted desktop cpu coming soon from AMD." Then she says in the post the 980X is the only one that's faster. So is it the fastest or not? Which one is it, jennyh?

I agree 100%.

I've read somewhere that it might be released earlier. 19th of April has been mentioned.

Gulftown has 12MB cache, it won't beat it

LOL - actually if you look at the Xbitlabs article, where they compare 6-core Istanbuls (2.6GHz & 2.4GHz) to both an Intel i7-920 and an AMD P2-965 quad core, the Istanbul loses rather badly in most of the benchmarks .

So, mods please change the title to "Thuban - world's slowest desktop coming" in the interest of accuracy .

Thuban is worth it if you can utilize those cores. But for most of us it'll be faster quad ftw. I don't see i5 having any trouble. Anyone who is buying a cpu for those tasks is going to have i7 already over i5.

Very interesting price point though. For the cost, I would imagine they will sell a bunch to folks who really don't need them to begin with.

AMD clearly need to sack their entire marketing department if they have 'the fastest desktop CPU' but are planning to sell it for a quarter as much as their competition would.

My thoughts for the entire life span of the Phenom 2 line. They could easily command more money from their chips but would rather flirt with the thin margins to gain marketshare.

There are a large number of games that can use more than 2 threads and show performance gains doing so, this isn't 2007 anymore. It is true that no game requires more than a fast dual core yet, but there is clearly performance gains going to 4 cores, however you are right that 6 are usesless and that can be seen with the gulftown benchmarks. Speak for yourself having no use for the cores, some of us do.

Does it matter how many cores you have if the OS or software apps don't take advantage of them?

I'm thinking of a new build and I'm wondering that over the course of the computer's 2 or 3 year life, is there any point in building anything other than an i7 right now from a cost benefit basis. Over the course of 2 or 3 years, I don't know if a couple of hundred dollars matters. It works out to pennies a day.

I mostly do online stock and currency trading with multiple monitors. Speed is important in currency trading, the market moves very quickly.

What AMD CPU's will be viable years down the line? How about Intel C2 quads, i5's, etc.

Thanks all for your opinions and knowledge and recommendations.

the limitation is the os...

i personally have upgraded from a 9550 to a i920 and the performance increase was minimal.

so i got a virutal 8 core processor with no virtual 8 core apps to run on it..

yeah photo editing and video conversion etc is quicker but then again a 9550 was no slouch either... but all the games were playable on a 9550, not one struggled on it..


Its far to soon to jump on the 6 core processor in my eyes, but some will feel happy that they got on in their ali box with a heatsink the size of notre dam and a fan the size of a windmill keeping it cool.

Lets use the current tech to proper use. A faster processor means more sloppy programming as the processor can do it anyway.

The assessment that the OS is the limitation is not correct. The OS handles the cores just fine. i.e. It can see and utilize whichever cores of my processor - whether on the fly or assigned through pinning an app. And though it is true that Vista and Win 7 are both more efficient than XP as far as multicore/thread usage, even XP is perfectly capable of simple load balancing and proper scheduling among all of the available (real or virtual) processor cores. And it makes that hardware resource available to any and all applications which may be running. And in a manner that is perfectly transparent to those applications. (Meaning *nobody* has to code what parts of what app run on which core and when. That is all handled by the OS's scheduler. All the app/programmer knows/cares is that it's passing a call to be processed. The details are completely irrelevant) So it's not an issue with the OS.

The limitation is that, with a few exceptions, application programmers don't/haven't written applications that take proper advantage of that kind of hardware enviroment. Why not??? Because writing highly threaded code is Really F*cking Hard To Do, and takes orders of magnitude more effort and better tools than have been necessary to date. Not to mention there is also cause to argue that writing truely multithreaded applications isn't necessarily good design practice, either. And while there are clearly heavier duty apps that are a good fit, there are also many (most!?!?) cases where tearing the whole thing apart to change it doesn't deliver any added value to the user.

The Hardware platform is there. The operating system is capable of the necessary load balancing and scheduling. But the apps don't exist to utilize it. So it's not the fault of the OS. Unless, of course, someone writes an OS that can deconstruct single/poorly threaded programming and somehow rewrite it on the fly to be efficient in a multithreaded environment...

Of course, if were to opine an OS exists which can do that, I'd suppose they also have a nice bridge in Sonora to sell...

The Xbit article was pretty awful. There are a multitude of problems with it:

- The title of the first page is wrong. The Istanbul is not a "Server Processor with Desktop Genealogy" by a long shot. AMD K8s and later were initially designed as server chips, initially released as server chips, and then only later released as desktop CPUs. The Istanbul's die is currently only available in server-only Socket F packages for crying out loud!
- The KN9U Speedster is very old. Xbitlabs did at least note that in the article, to their credit.
- They used slower-than-supported RAM with all of the AMD systems, which hinders their performance relative to the Core i7. The Istanbul can take DDR2-800; the Phenom IIs DDR2-1066/DDR3-1333. Registered DDR2-800 server RAM is quite a bit more expensive than DDR2-667, so I can see why they used it. Using the same-spec unbuffered DDR2-667 in the Phenom IIs makes sense to compare those two chip generations. It doesn't make sense when comparing versus the Core i7 line, since it does not allow the AMD CPUs to perform to full potential. Humorously, the i7 920 actually has to downclock the RAM they gave it- the i7 920 only runs its IMC at DDR3-1066 at stock.
- They don't say if they used the DDR2-667 or the DDR2-800 with the Phenom IIs in the benchmarks except for the X4 910 vs. Opteron 2435 page.
- Using their Istanbul setup to try to predict how well Thuban will perform is silly as Thuban has far higher memory bandwidth, lower memory latency (the Istanbul uses registered ECC DIMMs for crying out loud!), and is rumored to have a Turbo Boost-like feature that Istanbuls do not have. All of those affect performance in desktop benchmarks.

If anybody is going to change anything, it should be Xbit reworking their article to make it worth half a damn.