AfterDawn: News

AMD confirms Hexa-core 'Thuban' processor

AMD has confirmed to Maximum PC that it will release a Hexa-core (6 cores) processor (codename Thuban) next year that will be backwards compatible with AM3 and AM2+ socket-equipped motherboards. The confirmation comes after numerous rumors suggested that the company had a new 6-core chip in the works. "We are all about platform longevity and long-lived upgrade paths," an AMD spokesman said.

The comment is likely a jab at Intel for offering three different incompatible socket infrastructures currently for Desktop computers. The "Thuban" chip is expected to be marketed as the Phenom II X6, and is derived from the Opteron chip that surfaced earlier in the summer. It will feature all six cores on single 45nm die.



Thuban is likely to pack 904 million transistors in a 346mm2 chip, compared to Intel's Core-i7-975's 262mm2/731 million transistors and the Core i5-750's 296mm2/774 million transistors. It is expected to feature 3MB of L2 cache and 6MB of L3 cache.

The company did not offer up clock-speed details to Maximum PC, but it is expected to be lower than the quad-core parts due to the additional thermal output of two extra cores.

Written by: James Delahunty @ 23 Sep 2009 18:57

• Previous article

• Calls for fees to use BBC iPlayer

• Next article

• Skype deal 'is going to close': eBay CEO

• 14 comments

• Lothros

  Holy sh*t.
  
  Dowant!
  
  And Compatible with AM2+/3? Epic. If it were intel we'd have to buy 4 kinds of new motherboard :D

  23.9.2009 20:29 #1

• 21Q

  With the right cooling, I can't imagine what the clock speed would be. It's even more amazing that they made it backwards compatible. Kudos to AMD.

  23.9.2009 22:02 #2

• ZippyDSM

  Unless the chipset on the mobo is a fast improvement, or the cache interface on the core's vastly improved or the core speed vastly improved its not going to help much. Game wise at least for apps that know what multi cores is you should get a good boost in speed but I wonder if the limits of the mobo chips are not going to hamper it some...

  23.9.2009 22:14 #3

• borhan9

  Quote:AMD has confirmed to Maximum PC that it will release a Hexa-core (6 cores) processor (codename Thuban) next year that will be backwards compatible with AM3 and AM2+ socket-equipped motherboards.Why do we need 6 processor.

  23.9.2009 23:07 #4

• ZippyDSM

  Quote:Quote:AMD has confirmed to Maximum PC that it will release a Hexa-core (6 cores) processor (codename Thuban) next year that will be backwards compatible with AM3 and AM2+ socket-equipped motherboards.Why do we need 6 processor.
  Because parallel processing is to difficult to fool as is speed.
  
  I wonder could they do a 6Ghz single die CPU with today's tech that would not require water cooling?

  23.9.2009 23:11 #5

• DXR88

  Originally posted by ZippyDSM: Unless the chipset on the mobo is a fast improvement, or the cache interface on the core's vastly improved or the core speed vastly improved its not going to help much. Game wise at least for apps that know what multi cores is you should get a good boost in speed but I wonder if the limits of the mobo chips are not going to hamper it some...the memory controller is no longer handled by the north bridge its actually in the processor so memory throughput is entirely dependent on your AMD CPU. the northbridge handles expansions, integrated SATA controllor,and integrated peripherals.
  
  so if you wanted quadfire you'd have to buy a new motherbored w/ a northbridge that could handle it.

  24.9.2009 03:59 #6

• ZippyDSM

  Quote:Originally posted by ZippyDSM: Unless the chipset on the mobo is a fast improvement, or the cache interface on the core's vastly improved or the core speed vastly improved its not going to help much. Game wise at least for apps that know what multi cores is you should get a good boost in speed but I wonder if the limits of the mobo chips are not going to hamper it some...the memory controller is no longer handled by the north bridge its actually in the processor so memory throughput is entirely dependent on your AMD CPU. the northbridge handles expansions, integrated SATA controllor,and integrated peripherals.
  
  so if you wanted quadfire you'd have to buy a new motherbored w/ a northbridge that could handle it.Sounds like when they get up to 8 or 10 cores it will be refined enough to drool over.

  24.9.2009 04:01 #7

• KillerBug

  If you look at the current offering of AMD 6-core processors that these will be based on, it hardly makes your mouth water. They are very slow, the fastest one is clocked at 2.6GHZ (15.6GHZ total), and costs $2,650! That is over $250 per GHZ per core, but it is still not as powerfull as the 3.33GHZ I7...a processor costing $75 per GHZ per core, and doing more with every GHZ.
  
  AMD makes big announcements proudly...but they seem to make them about a year too late. Heck, we are still waiting on the fusion chip...and it now looks like the 5870 will be available first...negating the high performance/low power goal that it is designed to meet!
  
  I'm as big an AMD fan and Intel hater as anyone who does not work for AMD/ATI, and I still have an AMD cpu in my main computer, but even I can see that AMD is way behind when it comes to high-end processors...and this 6-core will not fix that unless it comes out next week, costs less than $500, works with AM2+, and runs at 3.2GHZ or higher. Even then, they will only have a marginal lead, and only for a short time. The days of the K7 and AMD's domination are long gone.
  
  Then again, if they can make it cheap enough, and compatable with my AM2+ mainboard, I might be willing to buy one rather than spending over $500 the cheapest 1366 CPU/mainboard combo.

  24.9.2009 06:20 #8

• xnonsuchx

  Originally posted by KillerBug: If you look at the current offering of AMD 6-core processors that these will be based on, it hardly makes your mouth water. They are very slow, the fastest one is clocked at 2.6GHZ (15.6GHZ total), and costs $2,650! That is over $250 per GHZ per core, but it is still not as powerfull as the 3.33GHZ I7...a processor costing $75 per GHZ per core, and doing more with every GHZ.
  Don't judge Opteron (server) CPUs w/ their 'desktop' CPU counterparts. Opterons are usually manufactured to significanly higher tolerances, which is part of the price discrepancy. Even when AMD's had similar performance and clock speeds between them, the desktop Athlon/Phenom CPUs have often been extremely cheap compared to similar Opterons. This is still several months off, so I wouldn't be surprised to see a 2.6-2.8GHz Phenom II 6X in the $300-400 range or so. Keep in mind, the highest end CPUs are usually FAR pricier than the ones a model or two behind them. Is a 3.33GHz i7 975 Extreme REALLY worth paying 75% more than a 3.06GHz i7 950? I'd definitely wait until these CPUs were out w/ definitive street prices before making any real judgment.

  24.9.2009 07:25 #9

• handsom

  It's very entertaining watching these two manufacturers go back and forth lately, ever since the first few credible sources even hinted at the idea of octal-cores in the next year or two. It's like Intel and AMD are suddenly in a race to get there now.
  
  Interestingly enough, it's kind of a catch 22 for gamers. If you buy CPU with more cores, at a lower speed, your system can do more different things at once, but it's still going to do a lot of them slower. Obviously newer versions of windows are being built with multi-threading in mind; but that's just your OS.
  
  Gaming on the other hand, is still trying to catch up on multithreading for the mostpart. A lot of games are reporting very minimal use on anything other than a dual core. Using 60% of one core and 15-30% of the rest isn't exactly a huge performance increase.
  
  Hopefully, more games will go multi-threaded, I believe that as soon as a few major engines are updated to use Multi-threading, we'll see this really blossom. Valve's source engine is beginning to look more appetizing for developers because it's already multi-thread optimized; as things like the new IDTech engine, and others hit the market, and become available for other companies to use for their own games; I think we'll see a lot better options becoming available.
  
  Look at how many things used the Quake II and III engines to do some really cool games without having to write all the code from scratch. I think the same thing will happen in the next year, allowing developers to make great games with this technology, building off the base code that other engineers have written.
  
  Hopefully this gibberish makes sense to someone out there... ^_^

  24.9.2009 15:30 #10

• ZippyDSM

  handsom
  When you did you crawl back I, not seen you around in a while!
  
  The architecture is what I look at if it can handle more throughput on each core then thats what I want to get.

  24.9.2009 15:34 #11

• handsom

  Heh. I've been doing some much more technical work for my employer for the last year or two; so my time has been ZAPPED. I'll probably start posting a lot more again in the next few months... Assuming everything goes as planned.
  
  I was a little sad to see that the PC forums seem so dead. I'm used to posting questions and seeing twenty people clamor over eachother to answer it, lol. Guess I'll have to see how much useful input I can start generating for folks. ^_^
  
  As for architecture... It's kind of a mixed bag. I get the feeling in the next five to seven years, CPU's may start sharing characteristics of current video cards, using a seemingly obscene number of slower cores that each handle smaller operations. Currently, it allows individual cores to process single effects and calculations like a particle that goes flying when you fire at a wall, while another core handles a large texture, etc.
  
  Theoretically, even an OS could see massive gains if properly coded for this. You could assign a set of cores to larger file operations For example you're moving a 2gb file, the OS could assign five or six 700Mhz cores to that task, while leaving the rest for other tasks; this would leave all the other cores to other tasks, allowing the os to better separate EXACTLY how much processing power each task can use. If general computing could be threaded this way at the most basic levels, we could see some really cool gains in performance, both gaming and otherwise. It would give much better control over resource management, and allow for much more flexible computing. I think we're several years off from this, but I'm confident that this is the direction we're going.

  24.9.2009 16:00 #12

• maitland

  we want sexa-core!

  1.10.2009 15:05 #13

• ZippyDSM

  Originally posted by handsom: Heh. I've been doing some much more technical work for my employer for the last year or two; so my time has been ZAPPED. I'll probably start posting a lot more again in the next few months... Assuming everything goes as planned.
  
  I was a little sad to see that the PC forums seem so dead. I'm used to posting questions and seeing twenty people clamor over eachother to answer it, lol. Guess I'll have to see how much useful input I can start generating for folks. ^_^
  
  As for architecture... It's kind of a mixed bag. I get the feeling in the next five to seven years, CPU's may start sharing characteristics of current video cards, using a seemingly obscene number of slower cores that each handle smaller operations. Currently, it allows individual cores to process single effects and calculations like a particle that goes flying when you fire at a wall, while another core handles a large texture, etc.
  
  Theoretically, even an OS could see massive gains if properly coded for this. You could assign a set of cores to larger file operations For example you're moving a 2gb file, the OS could assign five or six 700Mhz cores to that task, while leaving the rest for other tasks; this would leave all the other cores to other tasks, allowing the os to better separate EXACTLY how much processing power each task can use. If general computing could be threaded this way at the most basic levels, we could see some really cool gains in performance, both gaming and otherwise. It would give much better control over resource management, and allow for much more flexible computing. I think we're several years off from this, but I'm confident that this is the direction we're going.And that's the rub if its not coded for it its worthless than a CPU that's geared to higher performance per core.

  1.10.2009 15:15 #14