All AMD R6xx chips are 65 nanometre chips, now

Yes, Teh Inq is back with more rag trash. This time it's the statement that AMD held up R600 to make ALL of the chips 65nm.

That means that they will be ahead of nVidia for mainstream 65nm. That's really good news. I wanted to wait and see if R600 would be as good as G80. Now it will more than likely be cooler running.
54 answers Last reply
More about r6xx chips nanometre chips
  1. Well, with that logic they should just delay it another year and release it on 45nm.
  2. Quote:
    Well, with that logic they should just delay it another year and release it on 45nm.


    There's a method to ATi's madness (I sill refuse to call them AMD until AMD do something). The R600 was to not only be large and warm but also take up considerable wattage using the 80nm process. Also the first iteration of the R600 was to use GDDR3.

    Therefore ATi opted to move it to 65nm with GDDR4. A good choice as it will nullify the effects of the 8900 series from nVIDIA.

    In other words ATi will have the top VPU for a year until nVIDIA's next gen GPU. (there's no reason for ATi to pull another R520/X1800XT).
  3. Great news!!!!!

    Now if only Nvidia would sign a contract with intel to manufacture their VPUs :D
  4. One of AMD's engineers stated at CeBit that 65nm process will be used for laptop versions, to reduce heat/power consumption.
    I think he's more credible that some ass from the Inq.
  5. Quote:
    One of AMD's engineers stated at CeBit that 65nm process will be used for laptop versions,


    Does that mean we wont be seeing it on the desktop versions? Dammmmmmmmmm :evil:
  6. Let's hope AMD can get it out soon.
  7. BTW When are they coming out with a mid range product ? (I personally buy midrange)
  8. Quote:
    BTW When are they coming out with a mid range product ? (I personally buy midrange)


    I think they're launching across several price points.
  9. Cool my bad I should have RTFA :(
  10. Quote:
    Don't worry, Baron, the 65 nm process does not buy much over the 80 nm process.... the oxide thickness wall has been reached... this is why AMD's 65 nm process does not out perform it's 90 nm process..


    Nonetheless, they will run slightly cooler, perhaps only consuming as much power as the nVidia offering.... ATI always consumed more power node for node than nVidia.... so at least it appears AMD will be on par with their competition this go round.



    I'm not worried. Physics says it will be better. I offered no numbers just that it will be cooler running. I guess as far as oxide thickness, I guess shrinking GPUs will never work and nVidia shouldn't even bother.
  11. Quote:
    BTW When are they coming out with a mid range product ? (I personally buy midrange)



    All levels are releasing at once in May. There are two low-midrange cards being released and those (RV610, RV630) will be configurable in terms of RAM and clock speed.

    But then since I don't feel like looking for links, I could be spreading FUD.
  12. Quote:
    Don't worry, Baron, the 65 nm process does not buy much over the 80 nm process.... the oxide thickness wall has been reached... this is why AMD's 65 nm process does not out perform it's 90 nm process..


    Nonetheless, they will run slightly cooler, perhaps only consuming as much power as the nVidia offering.... ATI always consumed more power node for node than nVidia.... so at least it appears AMD will be on par with their competition this go round.


    errr jack.. what does have AMD's process has to do with ATI's R600 ?
    isnt the UTMSC ( a fab name like that. dont remember it correctly atm) suposed to be the one producing it?(and thus the process? )
  13. Thanks for clearing that up
  14. Quote:
    Oh, it will be better for power than if it were fabbed at 80 nm.... most likely they needed to go to 65 nm because of this.... but it will not make it faster other than to bring it inline.

    The 90nm to 80 nm to 65 nm.... the performance curve is flattening out. If you have noticed, though nVidia shrunk to 80 nm, there was not a huge bang in clocks.... and power is still high. Same thing will happen with 65 nm R600.... if R600 out performs it will be architectural not by clock or fabbing.

    And don't speak of physics, you barely can spell physics.


    I am afraid that R600 is to be the same fate of NV30... :?
    New process at new generation, high heat dissipation......
  15. Perhaps this would be more fitting in the Graphics section?
  16. Quote:
    Yeah, the GPUs are odd ... they are opposite of what CPUs are doing. While AMD touted power reduciton and Intel pushed thermal limits to get clock speed high, GPUs are just running up clockspeed pushing thermals as high as the process/architecture will allow....

    Make no mistake.... R600 will be a power hog, even with 65 nm... it will just be less of a power hog than if it were 80 nm.... it will probably equal or surpass the G80 in power when it is finally rolled out.
    Jack


    Maybe with a lower performance...... :(
  17. Maybe AMD and ATI build their GPU's similairly or have transitioned over to it...maybe its going to be cheaper to build everything in 65....this could be a new trend...just think of ddr5 on your gpu ddr3 for the cpu and everything at 45nm? It will be amazing 4 sure.
  18. Quote:
    Yeah, the GPUs are odd ... they are opposite of what CPUs are doing. While AMD touted power reduciton and Intel pushed thermal limits to get clock speed high, GPUs are just running up clockspeed pushing thermals as high as the process/architecture will allow....

    Make no mistake.... R600 will be a power hog, even with 65 nm... it will just be less of a power hog than if it were 80 nm.... it will probably equal or surpass the G80 in power when it is finally rolled out.
    Jack


    Maybe with a lower performance...... :(

    This is generally accepted as being unlikely. Even if some of the "leaked" benchmarks were fake, one would still expect R600 to outperform G80 by at least a small margin.

    Now, if only there were games to tax an R600 or G80 to its limits.
  19. Quote:
    Yeah, the GPUs are odd ... they are opposite of what CPUs are doing. While AMD touted power reduciton and Intel pushed thermal limits to get clock speed high, GPUs are just running up clockspeed pushing thermals as high as the process/architecture will allow....

    Make no mistake.... R600 will be a power hog, even with 65 nm... it will just be less of a power hog than if it were 80 nm.... it will probably equal or surpass the G80 in power when it is finally rolled out.
    Jack


    Maybe with a lower performance...... :(

    This is generally accepted as being unlikely. Even if some of the "leaked" benchmarks were fake, one would still expect R600 to outperform G80 by at least a small margin.

    Now, if only there were games to tax an R600 or G80 to its limits.

    Oh... as competitive as the GPU market is, and AMD's focus on performance.... the R600 will out do the G80 when it launches I am pretty certain. This is my hunch anyway....

    Which will be fun to see how they demonstrate it .... it is pretty clear that in some cases (not all) but some, the top AMD processor bottlnecks the G80 at 16x12 with all the eye candy on. If they do a side by side compare using AMD CPUs, it will make them look even in some cases .... :)

    I wonder if they will pop in a OCed X6800 or go with a QuadFX setup?

    Well...now that you mention that...it would be quite the publicity stunt to demo a K10 rig at the R600 launch, wouldn't it...?
  20. Quote:
    This is generally accepted as being unlikely. Even if some of the "leaked" benchmarks were fake, one would still expect R600 to outperform G80 by at least a small margin.

    Now, if only there were games to tax an R600 or G80 to its limits.
    Oblivion and Rainbow Six Vegas.
  21. I've heard in a different report that this will not only allow ATi to push the R600 to over 1ghz 8O :twisted: , but also cut the power requirements in half. :P :P :P

    Well worth the wait IMO.
  22. Quote:
    BTW When are they coming out with a mid range product ? (I personally buy midrange)



    All levels are releasing at once in May. There are two low-midrange cards being released and those (RV610, RV630) will be configurable in terms of RAM and clock speed.

    But then since I don't feel like looking for links, I could be spreading FUD.

    And I'd wager $20 that you won't be able to purchase a single card until July - August...
  23. Quote:
    Don't worry, Baron, the 65 nm process does not buy much over the 80 nm process.... the oxide thickness wall has been reached... this is why AMD's 65 nm process does not out perform it's 90 nm process..


    Nonetheless, they will run slightly cooler, perhaps only consuming as much power as the nVidia offering.... ATI always consumed more power node for node than nVidia.... so at least it appears AMD will be on par with their competition this go round.



    I'm not worried. Physics says it will be better. I offered no numbers just that it will be cooler running. I guess as far as oxide thickness, I guess shrinking GPUs will never work and nVidia shouldn't even bother.Physics said that Prescott would be better than Northwood too. Don't hang your ego on specs. :oops:
  24. Quote:
    ...You can see that Td (gate delay, or transistor speed) goes up as Tox goes up:...
    Gate delay increases as gate oxide thickness increases, so I think transistor speed goes down not up.
    BTW, Edgar Wallace is wrong.
  25. Quote:
    BTW When are they coming out with a mid range product ? (I personally buy midrange)



    All levels are releasing at once in May. There are two low-midrange cards being released and those (RV610, RV630) will be configurable in terms of RAM and clock speed.

    But then since I don't feel like looking for links, I could be spreading FUD.

    And I'd wager $20 that you won't be able to purchase a single card until July - August...

    I wouldn't.
  26. Quote:
    Not quite, in the application of gate oxide the dielectric constant, if you will, is sorta a measure of how effective the dielectric can allow electric fields to pass through to the other side. Low dielectric constants diminish the electric field through the material (hence low K for back end to avoid 'cross talk' in layman terms between wires). However, for a good gate you want 'cross talk' to the channel so that the electric field can modulate the inversion layer and cause the transistor to conduct. It is not about slowing or speeding up the negative charge, it is about how fast can it allow the charge to make the connection between source and drain so that the transistor can be 'ON'.
    Jack
    Time for my dumb question. Transistor switching speed (some inverse function of Td) must be faster than clock speed. In rough terms, is when these speeds approach the same value when we see systems crash?
  27. So unless AMD can use/adapt/steal/improve on Intels new process, they are screwed?
  28. Thank you Jack. Your posts are awesome.

  29. Quote:
    Don't worry, Baron, the 65 nm process does not buy much over the 80 nm process.... the oxide thickness wall has been reached... this is why AMD's 65 nm process does not out perform it's 90 nm process..


    Nonetheless, they will run slightly cooler, perhaps only consuming as much power as the nVidia offering.... ATI always consumed more power node for node than nVidia.... so at least it appears AMD will be on par with their competition this go round.



    I'm not worried. Physics says it will be better. I offered no numbers just that it will be cooler running. I guess as far as oxide thickness, I guess shrinking GPUs will never work and nVidia shouldn't even bother.

    Oh, it will be better for power than if it were fabbed at 80 nm.... most likely they needed to go to 65 nm because of this.... but it will not make it faster other than to bring it inline.

    The 90nm to 80 nm to 65 nm.... the performance curve is flattening out. If you have noticed, though nVidia shrunk to 80 nm, there was not a huge bang in clocks.... and power is still high. Same thing will happen with 65 nm R600.... if R600 out performs it will be architectural not by clock or fabbing.

    And don't speak of physics, you barely can spell physics.


    Isn't physics where you stand under something for too long and it falls and bashes your thick skull in?
    I can speak of whatever I want. You're not the boss of me. Why don't you spread your negativity around the house? We don't need it.
  30. Quote:

    This is why I constantly beat on Baron about the 40% number.... 40% at today's clock speed is not the same as comparing against Penryn when it launches --- we don't know the clock speed to be honest and I may be over estimating, but 3.5 GHz is not unlikely based on the information Intel has published and is in alignment with the leaked speeds published by VR-ZONE:
    http://www.vr-zone.com/?i=4184



    Why can't you leave my name out of your mouth? If you mean the projected 40% speed increase of switching for AMDs mature 65nm process with DSL and the other thingamajig, I guess we'd have to have a 65nm chip without it and test it at the switching level.

    Maybe you could write a white paper on it for AMD?
  31. Quote:
    8O friggen hell :!: what a pain in the azz. Is modelling useful, i.e. computer aided simulations and such?

    You really have to know your materials and your process and run the basic equations on what is possible.

    I guess if you know your material properties and what the process does to change them its just a matter of practical trial and error to get to a desired result.
    That's what the paper Jack quoted was all about. If you can predict relationships between Td, Tox and Toxe at various voltages, then a lot of the trial and error is not necessary. This gives a good indication of where the sweet spots are. You always have to have some trials and errors though (job security 8) ).
    http://www.eecs.berkeley.edu/~hu/PUBLICATIONS/Hu_papers/Hu_JNL/HuC_JNL_031.pdf
  32. So what youre saying Jack, is that using high K in their processes, both Intel and AMD will only lead them back to their highest clock? As in the P4's and the K8? P4 being around 3.6 and K8 being 3.0? That is still using the same Silicon?
  33. Quote:
    ...is this equation correct?...
    Not really. I have noticed that Baron can be very helpful in solving forum members computer problems. Your model needs more empirical data.
  34. My question then is, who will benefit more from high K? Will it take a complete overhaul from AMD to use the "better" silicon ? And then of course get the same effects?
  35. Quote:
    My question then is, who will benefit more from high K? Will it take a complete overhaul from AMD to use the "better" silicon ? And then of course get the same effects?
    Looks like Intel at least for this year...
  36. Thanks for the laugh Baron. :lol:
  37. Quote:
    Yeah, the GPUs are odd ... they are opposite of what CPUs are doing. While AMD touted power reduciton and Intel pushed thermal limits to get clock speed high, GPUs are just running up clockspeed pushing thermals as high as the process/architecture will allow....

    Make no mistake.... R600 will be a power hog, even with 65 nm... it will just be less of a power hog than if it were 80 nm.... it will probably equal or surpass the G80 in power when it is finally rolled out.
    Jack


    Maybe with a lower performance...... :(

    This is generally accepted as being unlikely. Even if some of the "leaked" benchmarks were fake, one would still expect R600 to outperform G80 by at least a small margin.

    Now, if only there were games to tax an R600 or G80 to its limits.

    I'd expect something around a 20% performance boost if not twice that over G80... and nothing less.
  38. Quote:
    It depends.... Barcelona will come out.... it will be a great chip, and I am certain it will compete to out perform C2D at today's clock.... the question is when will 45 nm release and will the clock speed be as high as what the intrinsic data and the leaked info suggest it will be?

    If so, then I expect AMD to hold a small advantage in server, and lose on desktop and lose majorly in mobile.

    That is my guess anyway.... but there is no guarantee Intel's 45 nm will arrive this year, Intel has been saying 2H 2007, but it may go to 2008...


    I love the horse race as much as the technology. If K10 matches or surpasses Core even for 6-8 months it will give AMD a nice boost. It will also prove the Horde wrong because most of us will be cheering along.

    I'd be sold on a K10 quad but not until power comes down. Intel has managed to do well at 65nm without high K so I expect similar performance from AMD once their process matures and the chip itself evolves. I'm a sucker for low power though and Intel already has 50 watt quad cores.
  39. Quote:



    Isn't physics where you stand under something for too long and it falls and bashes your thick skull in?
    I can speak of whatever I want. You're not the boss of me. Why don't you spread your negativity around the house? We don't need it.


    Yes, phyhsics is where I stand together with the data.... unfortunately, you know nothing of niether so it is best you don't speak at all...

    Finally, the bolded item is the 'rudeness' that I spoke of in our PMs, this is your exact phrase you used when I provided friendly advice and you provided this.... Baron, some leapords never change their spots and the color of your character is what will hold you down in life and in any attempt at success. People can read you like a book, and the book is pretty pornographic.

    Jack


    If you take all of the parts of Windows programming the little capacitance crap you talk about is minuscule. That's my point. I don't have time to read white papers until I'm psychotic just to prove I read white papers until I was blue in the face. Suffice it to say that the purpose is not to keep getting smaller and faster but to widen execution.

    Unfortunately everything has to be a war with these lined up against those when if everyone was working together we'd already have 10GHz chips running at 65W with multiple processors running at2 slower speeds.

    What if I were to spend a few hundred on a couple of IEEE journals and post all of this crap would that mean I'm now smart or worthy or would it mean I obviously have been letting my development work slip?
  40. Quote:
    Thanks for the laugh Baron. :lol:



    No one should take this too seriously. I need to remember that too.
  41. It's a bug. ;)
  42. Did I miss something?
  43. Wow, haven't bought a video card in too long. I didn't realize Nvidia had hit the 80nm mark. I made an oops, that I looked around for a little, but Wiki was wrong.... yes.... I said Wiki. Sorry guys, I looked around before I posted, and have not been keeping up with the cards as much because it's generally a pretty easy decision when it comes time to buy one. Anyway.... sorry, I thought the information would be ATLEAST that accurate.... wow, never realized how bad it was, probably from damn fanboys editing that crap. I have normally been an Nvidia fan, and the two companies have been neck and neck for a while now. Thought ATI ended up pulling the process edge over on them a bit. I WAS WRONG. Funny thing is, no one caught it....? It was in another thread....

    wes
  44. Quote:


    If you take all of the parts of Windows programming the little capacitance crap you talk about is minuscule. That's my point. I don't have time to read white papers until I'm psychotic just to prove I read white papers until I was blue in the face. Suffice it to say that the purpose is not to keep getting smaller and faster but to widen execution.

    Unfortunately everything has to be a war with these lined up against those when if everyone was working together we'd already have 10GHz chips running at 65W with multiple processors running at2 slower speeds.

    What if I were to spend a few hundred on a couple of IEEE journals and post all of this crap would that mean I'm now smart or worthy or would it mean I obviously have been letting my development work slip?

    hey.. if what you're looking for is consensus on your statement, you better list some sources and data, or this "believe me because i said so" crap is not working at all.
  45. Quote:


    If you take all of the parts of Windows programming the little capacitance crap you talk about is minuscule. That's my point. I don't have time to read white papers until I'm psychotic just to prove I read white papers until I was blue in the face. Suffice it to say that the purpose is not to keep getting smaller and faster but to widen execution.

    Unfortunately everything has to be a war with these lined up against those when if everyone was working together we'd already have 10GHz chips running at 65W with multiple processors running at2 slower speeds.

    What if I were to spend a few hundred on a couple of IEEE journals and post all of this crap would that mean I'm now smart or worthy or would it mean I obviously have been letting my development work slip?

    hey.. if what you're looking for is consensus on your statement, you better list some sources and data, or this "believe me because i said so" crap is not working at all.

    So you mean I got fired?

    Oh wait you're CPU is everything mode. It's not. I just like to take people to the abyss and see if they come back. Wait, that's a part of your username.
    I guess you didn't come back.
  46. Quote:

    So you mean I got fired?

    Oh wait you're CPU is everything mode. It's not. I just like to take people to the abyss and see if they come back. Wait, that's a part of your username.
    I guess you didn't come back.

    oh ok. so we are name calling now?
  47. Quote:
    I just like to take people to the abyss and see if they come back.


    lol troll :lol:
  48. Quote:

    So you mean I got fired?

    Oh wait you're CPU is everything mode. It's not. I just like to take people to the abyss and see if they come back. Wait, that's a part of your username.
    I guess you didn't come back.

    oh ok. so we are name calling now?

    This is the 9th position on the list of methodologies Baron uses to make an argument.

    1. If you don't know try to fool them by acting like an authority on the topic.
    2. Twist the detail around to fit your definition.
    3. If #2 fails, twist the definition around to assert your argument.
    4. Post "I said before" to make the supposed fact appear more real.
    5. Take said reference, pick out the one statement that sound best for AMD, and reference it out of context.
    6. Take #5, which is an opinion and present it as a conclusion.
    7. Backtrack the argument to reconclude 'that is what I meant in the first place'.
    8. Apply Baron nLogic® that always works, and makes AMD appear unstoppable, ignore all actual data, and conclude the opposite of what the data really says.
    9. If all else fails, call them names like Jumping Jagoff or A-hole.

    JackYou missed one Jack. When speaking of AMD vs Intel....

    #10. X2 is fast enough.
  49. Quote:
    Thanks, fixed....

    i think he also employs another tactic.

    if a company has inferior, or immoral marketing tactics (in this case, intel), its products are inferior, thus he refuses to buy them.
Ask a new question

Read More

CPUs AMD Nvidia