Page 186 of 216 FirstFirst ... 86136176184185186187188196 ... LastLast
Results 1,851 to 1,860 of 2151

Thread: Bulldozer Leaks, Previews and Reviews

  1. #1851
    Quote Originally Posted by savantu View Post
    Feel free to believe the BS about "strategic reasons" and "priorities". As if selling $70 LLanos is better than $800 Opterons...
    Those Opterons are not delayed.

    How ?
    Crippled how? By allowing only up to 3-3.1GHz on the sample, which makes sense if you want to hide its true clock potential.

  2. #1852
    8-bit overflow
    Join Date
    Sep 2010
    Posts
    495
    Quote Originally Posted by James View Post
    Those Opterons are not delayed.
    Really ? I can go and order one from Dell/HP ?
    Having supplied Cray means nothing. They would have AMD people owrking with stencils and doing it by hand if need be, just to avoid the contract penalties.
    Secondly, Interlagos isn't affected by the inability of the 8 core products to exceed 3GHz since it is an MCM design and clocks are lower due to the thermals.

    To give you and example, Intel supplied Bull with Montecitos ( Montecito's faith can be re-traced in the mirror with BD ) in H2 2005 even if the chip launched in July 2006.

    http://www.cea.fr/english_portal/new..._powerful-3483

    Crippled how? By allowing only up to 3-3.1GHz on the sample, which makes sense if you want to hide it's true clock potential.
    Which reinforces my point. Performance per clock is set in stone, the variables are frequency and power.
    He's not dead, he's electroencephalographically challenged.

  3. #1853
    Quote Originally Posted by savantu View Post
    Really ? I can go and order one from Dell/HP ?
    Having supplied Cray means nothing. They would have AMD people owrking with stencils and doing it by hand if need be, just to avoid the contract penalties.
    Secondly, Interlagos isn't affected by the inability of the 8 core products to exceed 3GHz since it is an MCM design and clocks are lower due to the thermals.
    Pretty sure it was Q3 for server launch. As are you.

    Which reinforces my point. Performance per clock is set in stone, the variables are frequency and power.
    Yes but you have no idea what the "true" clock speed is on any of these BD ES's. That Llano sample on Coolaler is clearly clocked ~twice as low as it is showing. What if these BD samples are exactly the same, hmm?

  4. #1854
    Quote Originally Posted by savantu View Post
    Indeed. It just happens for AMD to have a great performing part, mobo makers all are ready to roll, boards are in the shops and at the last minute AMD says " 2-3 month delay..sorry".
    Limited amount of production space and huge demand for Llano (a sure sell in the market), which on short term could do well for AMD's marketshare and due to that AMD's mindshare, mind you.

    The simplest explanation is by far and away the most logical. It was said by all review sites, performance ( as in clocks ) isn't where it should be. Parts are clocking 1GHz less ( we see from the leaks it is true, most are under 3GHz while the launch target was around 4 ).
    The simplest explanation by far isn't always close to the correct one. In technology even more so. With the simple explanation, AMDs HD48xx and HD69xx should have been utter fails, since they were 'late' and no real performanceleaks were found. Must be hiding something there.

    Furthermore we only see engineering samples. There is no saying what AMD did to those chips. Who knows what hidden secrets the whole new Bulldozer architecture has. It certainly isn't as straight forward a design as what we have seen before.

    Feel free to believe the BS about "strategic reasons" and "priorities". As if selling $70 LLanos is better than $800 Opterons...
    Wait, that's why Cray is shipping new Bulldozer servers before they actually get launched?
    Gigabyte GA-A75M-D2H||AMD A8-3850||Corsair XMS3 PC3-16000(2000MHz)||Sapphire Radeon HD6670||Fractal Design Define Mini
    Life, it all of a sudden comes back to you and you have no clue who it is.

  5. #1855
    8-bit overflow
    Join Date
    Sep 2010
    Posts
    495
    Quote Originally Posted by James View Post
    Pretty sure it was Q3 for server launch. As are you.
    That's also why Cray says "XK6" will be available in H2 2011. The first seed systems aren't indicative of BD's readiness.

    Yes but you have no idea what the "true" clock speed is on any of these BD ES's. That Llano sample on Coolaler is clearly clocked ~twice as low as it is showing. What if these BD samples are exactly the same, hmm?
    Just because CPU-Z cannot show LLano clock, it doesn't mean the BIOS and other programs ( like AMD overdrive and the mobo's own utilities which was validated with BD ) are wrong. Those mobo are in the market because they are ready to receive BD. You'd assume at least they are able to identify true clock speed, wouldn't you ?
    Don't generalize based on CPU-Z.
    He's not dead, he's electroencephalographically challenged.

  6. #1856
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,904
    Quote Originally Posted by 265586888 View Post
    The delay was NOT about performance issues.
    No more FUD about that, please.
    I hope I can buy BD until August then.
    Current rig: i5 3570 @3.8GHz, Asus 7750, Dell 24'' 2560X1440, Corsair Vengence 16GB Ram DDR3 @1600MHz, Chieftec 550W PSU, 480GB OCZ SSD + 500GB Seagate + 1TB WD hd, Windows 10 64-bit

  7. #1857
    640k who needs more?
    Join Date
    Mar 2010
    Posts
    995
    Quote Originally Posted by savantu View Post

    I've read that.


    Nice of you to keep track; :P I can write in every other sub forum there though. But I've lost any interest tbh.


    While CPU-z might show the wrong clock, AMD overdrive doesn't. So BD at least is running at whatever speed the program is saying.Secondly, even for Llano, the tests point very clearly it is in the same IPC range is Phenom II ( by your conclusion also ).




    How ?
    No. BD ES clock speed is internally set to 2x less of what any of those apps show.Only the latest AOD has preliminary support for Fam15h and this is of course only if appropriate BIOS hooks are in place .So AOD + BD ES= fail. Just as in case of CPUz and probably other apps.

    So to recap:
    BD ES IPC is probably set in stone and will be the same as what we will have in retail chips(unless microcode trickery is added into the mix,which we don't know). BD ES clock speeds are (internally) messed up and are in reality up to 2x lower than what is shown in BIOS or CPUz. Therefore we have 27s SPi score,4.6pts(!) C11.5 score for 8 core@ supposedly 3.2Ghz ,Fritz chess result which is in the line of ~2.5Ghz Thuban or ~3.6Ghz X4 etc. Just as in case of Llano ES which suffers from the same "illness",BD ES are having their max clock internally capped. One would need to OC "3.2Ghz" Zambezi ES to 6.4Ghz in order to see how it performs on default when it hits the retail.Or he can multiply most of the "3.2Ghz" results by factor of 2 and see it easy way.

    Fact n1: Zambezi will not be slower than Deneb
    Fact n2:Zambezi will not be slower than Thuban
    Possibility n1: Zambezi will be in line with 2600K
    Possibility n2: Zambezi will be faster than 2600K and in line with 980x/990x
    Possibility n3: Zambezi will be faster than 980x.

    Possibility 2 and 3 are most realistic in my opinion.

  8. #1858
    8-bit overflow
    Join Date
    Sep 2010
    Posts
    495
    Quote Originally Posted by Dahakon View Post
    Limited amount of production space and huge demand for Llano (a sure sell in the market), which on short term could do well for AMD's marketshare and due to that AMD's mindshare, mind you.
    BD market is 5-8x smaller than LLano and brings the same revenue. To have it ready and not ship it because of "strategic reasons" is beyond my comprehension.
    But I reserve the last laugh on the matter. Feel free to believe they have it ready, but they choose to supply 1mil more LLanos this Q than BDs.

    The simplest explanation by far isn't always close to the correct one. In technology even more so. With the simple explanation, AMDs HD48xx and HD69xx should have been utter fails, since they were 'late' and no real performanceleaks were found. Must be hiding something there.
    This is a straw man argument. Nobody assumed AMD was delaying HD48xx just to ship more HD3300 for "strategic reasons" as you imply with BD.
    Furthermore we only see engineering samples. There is no saying what AMD did to those chips. Who knows what hidden secrets the whole new Bulldozer architecture has. It certainly isn't as straight forward a design as what we have seen before.
    Like what? Quantum computing ? Why are people expecting miracles where there is none ?
    Wait, that's why Cray is shipping new Bulldozer servers before they actually get launched?
    Because those deals was done in 2009-2010 and back then AMD didn't know it will flop with early steppings. Just as Cray took TLB bugged 2GHz hot running Barcelonas, it will take anything that can run whatever code the lab has. Otherwise, somebody has to open the purse.
    He's not dead, he's electroencephalographically challenged.

  9. #1859
    8-bit overflow
    Join Date
    Sep 2010
    Posts
    495
    Quote Originally Posted by inf64 View Post
    No. BD ES clock speed is internally set to 2x less of what any of those apps show.Only the latest AOD has preliminary support for Fam15h and this is of course only if appropriate BIOS hooks are in place .So AOD + BD ES= fail. Just as in case of CPUz and probably other apps.
    CPU-Z is flawed ( as it has been proven countless times ).
    AOD and the mobo's utilities are not. They were designed to run the damn things.
    So to recap:
    BD ES IPC is probably set in stone and will be the same as what we will have in retail chips(unless microcode trickery is added into the mix,which we don't know). BD ES clock speeds are (internally) messed up and are in reality up to 2x lower than what is shown in BIOS or CPUz.
    No it's not. You're simply extrapolating from LLano CPU-Z photo where you could clearly see the lack of multiplier and base clock details ( and you only had CPU-Z as reference ) to BD where we have CPU-Z details, BIOS, AOD and Cinebench. Doesn't add up , no matter how much you recap.
    BD speed is correct and it is 2.8GHz.
    Therefore we have 27s SPi
    It'a 23s actually inline with a 2.8GHz Phenom.
    score,4.6pts(!) C11.5 score for 8 core@ supposedly 3.2Ghz
    It's running stock as you can see from the Cinebench photo, 2.8GHz. Nothing abnormal about this. Each module can offer 4 DP FLOPs per cycle, same as Nehalem and Phenom 2. Obviously, Cinebench doesn't use the FMACs.
    ,Fritz chess result which is in the line of ~2.5Ghz Thuban or ~3.6Ghz X4 etc.
    Again, what's abnormal ? A 2.5GHz Thuban has more FPU power than a 2.8GHz BD without FMAs.
    ...

    Fact n1: Zambezi will not be slower than Deneb
    Fact n2:Zambezi will not be slower than Thuban
    Possibility n1: Zambezi will be in line with 2600K
    Possibility n2: Zambezi will be faster than 2600K and in line with 980x/990x
    Possibility n3: Zambezi will be faster than 980x.

    Possibility 2 and 3 are most realistic in my opinion.
    I give it 1 and that is if they reach 3.8-4GHz stock. They weren't idiots when they positioned it there. If it could only smell 980x, AMD would have launched $500 and $800 parts and claim performance/$ crown.
    He's not dead, he's electroencephalographically challenged.

  10. #1860
    Banned
    Join Date
    May 2011
    Posts
    150
    savantu you are rather amusing but why don't you chill already.Intel stock won't go up just because you love it so much.
    BD server is not delayed as far as we know,it was announced for late summer and there is nothing to suggest it won't come at that time.We don't know why desktop BD is late but for AMD it doesn't matter much.BD is a server chip,Llano is targeting the notebook market.AMD has low share in notebook and server while they are quite fine with their share in desktop and Llano notebook might just have higher margins than BD desktop since notebook CPU's ASP is quite high and the Llano die is smaller (smaller means more chips/wafer too so they can have more units if capacity is limited).

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux