(AMD) Problems with SOI

[GoThr3k]

http://overclockers.com/tips01182/

interesting read

[safan80]

look at the ball bounce... AMD needs to figure out what they are doing.

[turtle]

ID Tag Process IDDC1@HighV IDDC1@LowV
ADO3600IAA5DD 65nm 32.2A @ 1.30V 11.2A @ 1.1V

HOLY .

[Shintai]

Could explain the low speedbin barcelonas too. Aswell as no flagship 65nm or Opterons.

[Lightman]

AMD already is doing better @65nm.
First Brisbanes was working @1.35V which is a lot for low clocked 65nm parts, but last week I purchased X2 3600+ rated 1.2V and X2 4800+ rated 1.25V.
Both produced after 0714.
I think we will see huge improvements towards end of this year, because AMD needs State of The Art process to ramp Phenom CPUs.

[nn_step]

HOLY .

That is right, you drop the Voltage on AMD's 65nm chips. Their power consumption drops like a brick. My backup rig doesn't use or need a heatsink and it is running WCG 24/7

[largon]

My backup rig doesn't use or need a heatsink and it is running WCG 24/7

You wanna refrase that?

[nn_step]

You wanna refrase that?

Nope, but if you don't believe me. Drop the voltage on your Proc and see what happens

[Shintai]

Maybe he should lower the frequency aswell to about 400mhz.

And if you lower the voltage too much the transistors wont switch aswell.

So another piece of bulldong?

[largon]

You're actually running a Brisbane without a HEATSINK? That's just plain awesome!


I think I'll pass trying it on my E6600 tho...

[nn_step]

Maybe he should lower the frequency aswell to about 400mhz.

And if you lower the voltage too much the transistors wont switch aswell.

So another piece of bulldong?

I haven't gotten any validation errors, so any errors are happening they don't involve the operating system or any processing that is noticeable.
But it is actually running above stock speed

[Shintai]

I haven't gotten any validation errors, so any errors are happening they don't involve the operating system or any processing that is noticeable.
But it is actually running above stock speed

So lets see, over 1.9Ghz dualcore with no heatsink?

Sure, and the moon is a big green cheese.

[ted3]

IBM crap got AMD into this, i doubt IBM can get them out of it again. Isnt AMD supposed to work with IBM on 45nm also? Make the table ready for Intel to increase their production advantage

Hard to believe that some AMD supporters still think AMD 65nm is good, its been obvious from the first day brisbanes came out that the process is not at all up to what IBM/AMD has claimed and unfortunately it will continue to be the limiting factor also for K10.

Maybe its time to look at how excellent this process was before they actually started using it? Dig up some old links anyone? This was supposed to be the greatest invention in chipmaking ever. It must have looked dåmn impressive in the slideshows IBM used to get AMD into this mess.

[red]

It's ok, http://seekingalpha.com/article/41686 they are evaluating bulk instead of SOI for 32nm

[nn_step]

It's ok, http://seekingalpha.com/article/41686 they are evaluating bulk instead of SOI for 32nm

They always evaluate before the commit to something. You don't spend hundreds of millions of dollars without spending money to evaluate if it is worth it and if there is something better.

[onewingedangel]

SOI's main problem is that it doesn't scale in the same way as traditional bulk processes. At the moment getting the process down right is very expensive and time consuming for each new process. The question is whether this will continue down the generations or whether the SOI scaling can be 'solved' making it far easier to implement in subsequent generations.

Clearly AMD don't want to have to expend so much time and effort on implementing SOI for each node especially when its benefits are far from clear, but if it is a design issue that can be solved they clearly don't want to throw the baby out with the bathwater and discard a potential advantage over the competition.

IBM don't seem to be having as many issues with their '65nm' power6 chips(which use a combination of 90 and 65nm logic design), so perhaps there is a solution for AMD.

[ted3]

You don't spend hundreds of millions of dollars without spending money to evaluate if it is worth it and if there is something better.

Oh THATS why they went for SOI, it was carefully evaluated before spending any money? I kind of doubt SOI was ever evaluated by AMD engineers, suits and ties struck a deal after watching a slideshow from IBM is my guess.

[[XC] leviathan18]

Oh THATS why they went for SOI, it was carefully evaluated before spending any money? I kind of doubt SOI was ever evaluated by AMD engineers, suits and ties struck a deal after watching a slideshow from IBM is my guess.

so y ou dont have any idea how to run a business

[ted3]

so y ou dont have any idea how to run a business

What on earth are you talking about? Post count?

On the IBM and SOI, there were reports that IBM couldnt make so many Cells because of low yields, not enough chips could reach desired clocks. That was last year, since then they had the same improvement as AMD i guess (which means "not much if any" if one were to look at AMDs 65nm offerings).

[xlink]

so basically SOI is good for laptops and LV servers and that's it?

[Pinnacle]

That is right, you drop the Voltage on AMD's 65nm chips. Their power consumption drops like a brick. My backup rig doesn't use or need a heatsink and it is running WCG 24/7

You wanna snap a quick pic of that and show us? make sure the power is turned on

[madcho]

SOI will win

[steelballzz]

You wanna snap a quick pic of that and show us? make sure the power is turned on

waits for picture....

[[XC] leviathan18]

What on earth are you talking about? Post count?

On the IBM and SOI, there were reports that IBM couldnt make so many Cells because of low yields, not enough chips could reach desired clocks. That was last year, since then they had the same improvement as AMD i guess (which means "not much if any" if one were to look at AMDs 65nm offerings).

so you really think amd adopted SOI without testing it first.... they tested it look at what they achieved with 90nm the problem is they need more money and time to make a better process and tweak the design to take advantage...

and trust been here long enough to dont care about post count....

[Sparky]

Hard to believe that some AMD supporters still think AMD 65nm is good, its been obvious from the first day brisbanes came out that the process is not at all up to what IBM/AMD has claimed and unfortunately it will continue to be the limiting factor also for K10.

Why are people saying AMD's 65nm sucks? The brisbanes seem to be doing pretty well, hitting 3GHz on them isn't rare. I've seen a few of them even higher. Seem to be fine to me Yeah still K8 architecture but a die shrink doesn't change that really.