• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Intel's 65nm

Lazn_Work said:
AMD's 90nm is actually licensed from IBM, they do not have thier own version.

http://www.internetnews.com/bus-news/article.php/3412251

I remember when AMD first tried to go to 90nm, they had to pay IBM ~$50 million or so to come bail them out becuase they could not get it to work, now they are conitinuning to pay $11 million on just licensing to IBM to be able to continue to use that tech, and in addition AMD is paying IBM $250-$280 million to help them with 65nm.

Not that this is a bad idea, IBM has great Tech (arguably better than Intel right now), but to crow about how great AMD's are at 90nm is a little silly when they do not even have their own 90nm process.

==>Lazn
Click to expand...
Ok, how about I put it this way:

I wish Intel would get off their asses and buy some good 90nm tech from IBM like AMD did instead of wasting my time on power-hungry Intel-fabbed 90nm prescott POSes.

Really, it doesn't matter who they baught it from. Companies buy and sell IP all the time. This is nothing new. What matters is that Intel doesn't think it's worth spending a little money to learn how to make good CPUs for their customers. Instead, they're willing to push over-heating crap out the door to customers because it's cheaper. I'm glad AMD did it right.
 
visaris said:
Ok, how about I put it this way:

I wish Intel would get off their asses and buy some good 90nm tech from IBM like AMD did instead of wasting my time on power-hungry Intel-fabbed 90nm prescott POSes.

Really, it doesn't matter who they baught it from. Companies buy and sell IP all the time. This is nothing new. What matters is that Intel doesn't think it's worth spending a little money to learn how to make good CPUs for their customers. Instead, they're willing to push over-heating crap out the door to customers because it's cheaper. I'm glad AMD did it right.
Click to expand...

DOTHAN DOTHAN DOTHAN DOTHAN, have we not been through this?
 
visaris said:
Ok, how about I put it this way:

I wish Intel would get off their asses and buy some good 90nm tech from IBM like AMD did instead of wasting my time on power-hungry Intel-fabbed 90nm prescott POSes.

Really, it doesn't matter who they baught it from. Companies buy and sell IP all the time. This is nothing new. What matters is that Intel doesn't think it's worth spending a little money to learn how to make good CPUs for their customers. Instead, they're willing to push over-heating crap out the door to customers because it's cheaper. I'm glad AMD did it right.
Click to expand...

They are also interested in reinventing the wheel to make those proccessors work [ I.E. Their propriatary mobo format :p ]
 
Scali said:
That's a fallacy.
Sure, not all transistors are the same (the ones you're talking about are often found in power amplifiers, and have very little to do with the ones you find in CPUs), but it is obviously safe to assume that the transistors in the Northwood and the Prescott are similar (if not identical), and even the ones in Prescott and Venice are similar. They perform the same kind of tasks, work at the same kinds of voltage and temperatures etc... So transistorcount is obviously not completely useless, considering that the other factors vary very little.
Click to expand...
What youre not taking into account is the transistors on Athlons are doing a different job, memory controller for one, different amounts of cache etc. all of which increase transistor count Sure northwood->prescott the comparison holds, but increasing transistor count isnt an excuse for procs to be significantly hotter, AMD managed to do it going to 64 bit. And if its a fair comparison, then why is dothan cooler than prescott, despite having more transistors?

I think basically were of the same opinion, prescott is fast, and flawed, dothan is sweet as anything, and Venice is good, but took a long time to get right.

and @Lazn, who cares if AMD licence the manufacturing process, as far as im aware the majority of the design work is still AMD, and i still hold that intel will have spent at least as much on r&d, just internally rather than outsourcing. Both of which are valid ways of doing business and have no impact on the performance etc of the cores at hand, i arent talking AMD vs intel, thats so apples to, well, i wont say oranges, but maybe apples to something slightly similar to apples, how about apple crumble and apple pie?

I wasnt crowing about AMD being better than intel, nor was i suggesting that AMD 90nm chips are better than intel ones, thats an entirely end user judgement.
 
Scali said:
Technically the move from Northwood to Prescott is quite similar to the move from Athlon to Athlon64. In both cases, cache was increased, 64-bit mode was added, and the pipeline was extended by a few stages. The main difference is that the Prescott doesn't have an onboard memory controller.
Click to expand...

Also, do not forget about the new instructions added to the prescott, as well as improved branch prediction and the fact that the entire architecture changed. For the move to 90nm from 130nm AMD didn't touch the architecture.
 
Herulach said:
What youre not taking into account is the transistors on Athlons are doing a different job, memory controller for one, different amounts of cache etc. all of which increase transistor count Sure northwood->prescott the comparison holds, but increasing transistor count isnt an excuse for procs to be significantly hotter, AMD managed to do it going to 64 bit. And if its a fair comparison, then why is dothan cooler than prescott, despite having more transistors?

I think basically were of the same opinion, prescott is fast, and flawed, dothan is sweet as anything, and Venice is good, but took a long time to get right.

and @Lazn, who cares if AMD licence the manufacturing process, as far as im aware the majority of the design work is still AMD, and i still hold that intel will have spent at least as much on r&d, just internally rather than outsourcing. Both of which are valid ways of doing business and have no impact on the performance etc of the cores at hand, i arent talking AMD vs intel, thats so apples to, well, i wont say oranges, but maybe apples to something slightly similar to apples, how about apple crumble and apple pie?

I wasnt crowing about AMD being better than intel, nor was i suggesting that AMD 90nm chips are better than intel ones, thats an entirely end user judgement.
Click to expand...


I think we're absolutely forgetting the fact that the transistors must switch at a greater speed on the prescott as compared to the dothan and 90nm AMDs... I am willing to bet that the 90nm @ 3.2-3.4Ghz on the stock cooler would be about the same temperatures as the prescotts, wouldn't you?
 
ScHpAnKy said:
DOTHAN DOTHAN DOTHAN DOTHAN, have we not been through this?
Click to expand...
Good point well made :)
I think we're absolutely forgetting the fact that the transistors must switch at a greater speed on the prescott as compared to the dothan and 90nm AMDs... I am willing to bet that the 90nm @ 3.2-3.4Ghz on the stock cooler would be about the same temperatures as the prescotts, wouldn't you?
Click to expand...
I arent forgetting that, the point about dothan/prescott transistor counts was to illustrate the fact that it is a pretty useless number when it comes down to how hot a processor is gonna run, or how well it performs.
 
Herulach said:
What youre not taking into account is the transistors on Athlons are doing a different job, memory controller for one, different amounts of cache etc. all of which increase transistor count Sure northwood->prescott the comparison holds, but increasing transistor count isnt an excuse for procs to be significantly hotter, AMD managed to do it going to 64 bit. And if its a fair comparison, then why is dothan cooler than prescott, despite having more transistors?
Click to expand...

I don't think I said you can directly compare Athlons, Dothans and Pentium 4.
The point was only the Northwood -> Prescott transition, and why the Prescotts didn't run cooler, despite being 90 nm.
 
Scali said:
I don't think I said you can directly compare Athlons, Dothans and Pentium 4.
The point was only the Northwood -> Prescott transition, and why the Prescotts didn't run cooler, despite being 90 nm.
Click to expand...
Good, so we agree on that then. Misread you, although i could dredge up banias > dothan, i wont, cos the transistor jump there was mainly in cache, not core logic.
 
Herulach said:
Good, so we agree on that then. Misread you, although i could dredge up banias > dothan, i wont, cos the transistor jump there was mainly in cache, not core logic.
Click to expand...

I never said it was the only reason either. I just said it was a big factor.
 
This may be a shock to some, but ATX was not always the standard format either.
There used to be AT before that.
BTX has some distinct advantages, so I don't see why we shouldn't move over to BTX, just like we moved over to ATX in the past.
 
Herulach said:
I think he probably meant BTX.
Click to expand...

I'm sure he did, I was just making the point that Intel was the driving force behind the beloved ATX as well. I guess they were "reinventing the wheel to make those proccessors work [ I.E. Their propriatary mobo format ]" back then too.
 
visaris said:
I wish Intel would get off their asses and buy some good 90nm tech from IBM like AMD did instead of wasting my time on power-hungry Intel-fabbed 90nm prescott POSes.
Click to expand...
Like mentioned above, it's not intel's process that's the problem. 90nm is doing great on the P-M Dothan.

The Prescott is just a sloppy processor. If Intel had just tweaked Northwood and shrunk it down to 90nm (strained silicon), you'd probably be seeing Venice-like power consumption on the P4. That never happened because Intel needed x86-64. Quickly. Even on 130nm, 2.6GHz Mobile Pentium 4-M CPUs (Northwood) were only sipping 35W. HT on Northwood CPUs only increased power by 10%. What could have been... ;)
 
JetUsafMech said:
I'm sure he did, I was just making the point that Intel was the driving force behind the beloved ATX as well. I guess they were "reinventing the wheel to make those proccessors work [ I.E. Their propriatary mobo format ]" back then too.
Click to expand...
Fair point, but ATX does offer considerable advantages over AT, including the whole new PSU thing. Where as BTX is nothing new really, just moving things around, to a position that is (maybe deliberatly) difficult to adapt AMD chipsets/boards to, and has no real benefits for cooling over atx assuming your case airflow is right.
 
Herulach said:
Fair point, but ATX does offer considerable advantages over AT, including the whole new PSU thing. Where as BTX is nothing new really, just moving things around, to a position that is (maybe deliberatly) difficult to adapt AMD chipsets/boards to, and has no real benefits for cooling over atx assuming your case airflow is right.
Click to expand...

I was under the assumption (and I may be wrong) that another tangible benefit of BTX was the simplicity of routing the traces during mobo manufacture, a more 'direct route' if you will. Lessens the amount of material used, thus reducing cost. How this would affect AMD chipsets/boards, I have no idea, but I wouldn't think it would be that difficult.
 
JetUsafMech said:
I was under the assumption (and I may be wrong) that another tangible benefit of BTX was the simplicity of routing the traces during mobo manufacture, a more 'direct route' if you will. Lessens the amount of material used, thus reducing cost. How this would affect AMD chipsets/boards, I have no idea, but I wouldn't think it would be that difficult.
Click to expand...
IIRC thats the problem, because of teh way A64s work they need traces routing a different way than P4s, because the ram traces have to go to the CPU, not the northbridge.
 
Herulach said:
I dont recall saying he was wrong, i readily admitted that dothan is good, my point was entirely about chipsets, to quote myself:

If amd outsourced venice who cares? Who cares also how much they spent, im sure intel spent a comparable amount on the prescott shrink.

I know he said 90nm, he was also commenting on AMDs first attempt at 90nm, which, as i admitted was dissapointing, and comparing it to intels first attempt which was prescott.

As far as transistor count, what the hell difference does that make? I have transistors that need heatsinks as bug as the one on my chipset to function right, and i have ones that dont need any cooling at all, comparing transistor count, as everyone who has gone through the statistics of anything knows, is completly useless.

Venice isnt merly a core shrink, it does add features, SSE3, improved memory controller, admittedly not as many as northwood - prescott, but still enough to not call it a die shrink.

By the same argument Venice proves AMDs 90nm process is good, although id be very surprised if Dothan and Prescott are made at the same fabs.

Whos fault is it prescott has layout problems? Thats no excuse for anything, if intel had designed the thing without said problems in the first place then we wouldnt be having this discussion would we?


maybe Mhz to power draw is the same, ive never seen any reliable info on it (the article you linked compares a P-M to older A64s no doubt on full size mobos as opposed to microatx ones).

And comparing to prescott, well of course its not, look at the kind of temps people with 3 ghz prescotts get, and then look at the people with overvolted Venices running at nearly 3Ghz running the same cooling, guess which is gonna run hotter?
Click to expand...

No, I was just replying to him saying AMD got it right the first time. I didn't mean *just you here either.

For the record, IBM is the process King! IBM came up with CU, AMD and Intel got it from them. IBM came up with Strain Silcon, SOI, Finished the first multi core processor, finished many other industry firsts. The difference is Intel talked about IBM's help, AMD didn't until they were forced to sheesh! Give credit where the hell it is due, in this case, IBM!

Tranny count! What the hell difference does it make? You're kidding right? It's not just the L2 hell the first Hotscotts had only a 1MB L2. The core has 11 more stages, 4 more K of L1, more for TLB, X86-64, EIST and other crap. But more bad news than that, it is the core's layout that has hot spots in hard to cool places and etc.. One more layer helped but it's still hot.

AMD's lower overall clock, most of trannies going to Memory controller, Crossbar, HTT links and yes 128K plus a 1024K of cache means the real core has a LOT fewer transistors. Though it does a great JOB with those fewer **CORE** trannies. These fewer are doing all of the hard/leg work. Hell yes I'd like to see what a 1MB L2 Tualatin at 90nm would do! I know danged well someone at Intel proposed a 90nm tweaked Northwood. Now if someone would just stop fighting the integrated memory controller and give it a shot.....

Donnie27
 
Donnie27 said:
IBM came up with Strain Silcon
Click to expand...

I'm quite sure Intel had that one first. Strained Silicon is not even the 'official' name of the technology that AMD uses. That is called DSL, which I suppose is IBM's name for it.
I'm also quite sure that Intel's production technology is slightly ahead of IBM's, but we had that discussion already, some time ago.

Tranny count! What the hell difference does it make? You're kidding right? It's not just the L2 hell the first Hotscotts had only a 1MB L2. The core has 11 more stages, 4 more K of L1, more for TLB, X86-64, EIST and other crap. But more bad news than that, it is the core's layout that has hot spots in hard to cool places and etc.. One more layer helped but it's still hot.
Click to expand...

Those are all transistors yes :)
It's just that L2 cache runs at a lower clockspeed, and you can pull other tricks on cache, so transistor count doesn't affect power/heat dissipation that much there. So using L2 cache as an example that transistor count does not matter is a fallacy. Transistor count inside the execution core does matter, and that's basically exactly what you just said.

Now if someone would just stop fighting the integrated memory controller and give it a shot.....
Click to expand...

Who says they're fighting it?
 
Scali said:
I'm quite sure Intel had that one first. Strained Silicon is not even the 'official' name of the technology that AMD uses. That is called DSL, which I suppose is IBM's name for it.
I'm also quite sure that Intel's production technology is slightly ahead of IBM's, but we had that discussion already, some time ago.
Click to expand...

Naw, I remember reading two things. One that IBM and another company fought in court over the Patent for Strained Silicon, IBM won. IBM first talked about it as far back as 1999. Yet maybe you're right though. IBM and Intel agreed to not sue eachother as well.

http://www.research.ibm.com/resources/press/strainedsilicon/
http://news.com.com/2100-1001-976120.html?tag=cd_mh

Yes I do remember the process called Dual Stress Liner but that's a newer version of Strained Silicon and the name change is just PR, IMHO.

Scali said:
Those are all transistors yes :)
It's just that L2 cache runs at a lower clockspeed, and you can pull other tricks on cache, so transistor count doesn't affect power/heat dissipation that much there. So using L2 cache as an example that transistor count does not matter is a fallacy. Transistor count inside the execution core does matter, and that's basically exactly what you just said.

Who says they're fighting it?
Click to expand...

Yes:) and Yes:) If you removed the L1 and L2, Prescott has many more trannies than Northwood.

I just got through reading about advance sleep and standby states for the newer L2s Intel uses. Yes L1 runs faster rhan L2 and many folks even forget that Intel ALUs run at twice the speed of the rest of the core.

Donnie27
 
perplex said:
first 65nm Intels will be Q1 2006 right?
Click to expand...

The rumor mill says Q4-05 before Christmas. Common sense says the end of Q1-06. That's not counting Yonah that might launch this Fall. If AMD has good results, you might see Pressler and the rest sooner than most folks think.

Donnie27
 
It seems that Cedar Mill's power consumption will be around 65W. I skipped over the 8xx series dual core processors due to the heat that they produced and I can only hope that the 65nm 9xx series dual core chips run a lot cooler and consume less power.
 
Yonah = 65nm (Mobile Dual Core). I feel that Yonah, Cedar Mill, and Presler all look promising. If Intel can keep Prelser's power comsumption down in idle, as well as full load conditions then they may have a winner on their hands.
 
The Doc said:
Yonah = 65nm (Mobile Dual Core). I feel that Yonah, Cedar Mill, and Presler all look promising. If Intel can keep Prelser's power comsumption down in idle, as well as full load conditions then they may have a winner on their hands.
Click to expand...

I haven't been keeping track of Intel's roadmaps, so please excuse my ignorance.

Is Intel planning to release desktop processors based on all these promising mobile cores? I'd be damned if they didn't.
 
Cedar Mill = Desktop (65nm) 65W power consumption
Presler = Desktop (Dual Core 65nm) power consumption = ???
Yonah = (Mobile Dual Core 65nm) power consumption = ???

From my understanding, Yonah will be the only mobile dual core offering. Yonah, Cedar Mill and Presler will all debute in Q1'06.
 
perplex said:
the Yonah looks incredibly hopeful.
Click to expand...

Shipping means to the VARs and OEM's so they'll have system ready.

Look at these prices ?

The Inquirer said:
Intel has told its partners that the x50 2.16GHz Yonah dual core chip with 2MB cache will be sold at $636, the x40 (2GHz) at $420, the x30 (1.83GHz) $295 and the x20 (1.66GHz) $240. These prices are consistent with a move from the existing Dothan to the newer CPU.

The low voltage (LV) x48 (1.66GHz) will cost $315 at launch, while the x38 (1.50GHz) will be priced at $285. At launch, the single core 1.66GHz Yonah will cost $210.

The next price cuts for Pentium Ms come in late July, which also is the introduction date for the 780, a 2.26GHz/533 processor. There will be significant price cuts made on other members of the Pentium M family in the 7XX series on that day.
Click to expand...

Now if these scale like the single cores for games, then now they'll have their streaming short comings covered, look out! Anyone not named Intel will be in deep DooDoo.

http://www.geek.com/news/geeknews/2005Mar/bch20050309029499.htm

Geek.com said:
This PMEE will be a Yonah core Pentium-M at considerably higher clock speeds--at the expense of power. The processor will dissipate 60 W, almost twice the non-EE Yonah's 31 W, and may run at 2.5 to 3GHz.
Click to expand...

So they say 31W and even a real EE (nothing like the fakes 840D-EE and 3.73) at only 60W, ummmmm!

Donnie27
 
The article doesn't mention whether or not the current 945P, 954G, and 955X chipsets will work with Yonah. I don't want to buy a 955X motherboard now and then upgrade my motherboard again just to support Yonah. :(
 
The Doc said:
The article doesn't mention whether or not the current 945P, 954G, and 955X chipsets will work with Yonah. I don't want to buy a 955X motherboard now and then upgrade my motherboard again just to support Yonah. :(
Click to expand...

It doesn't need to say it because Yonah will be socket 771
 
The Doc said:
The article doesn't mention whether or not the current 945P, 954G, and 955X chipsets will work with Yonah. I don't want to buy a 955X motherboard now and then upgrade my motherboard again just to support Yonah. :(
Click to expand...

Yup, robberbaron is right. Everything I've seen has said socket 771. Intel even showed on the slides at IDF. Just as the 65nm AMD processors will ship with DDR2 and 3 controllers, Socket M and some crazy 1000+ pin config. But as we all know, it only sucks when Intel does it platform upgrades. Going 945 + a cheaper 2.8 is not as bad as paying almost $600 for an X2, LOL!

Intel might screw the upgrade path for Pressler but it should still be Sc-775 unlike the Sc-771 Yonah. But there's always folks like Asus and CT479, maybe there'll be a CT-771 to 775 converter?

Donnie27
 
The Doc said:
True. I never heard of Socket 771 until you just mentioned it.
Click to expand...

Socket 771

They say it is a Server Socket, others say Yonah, we'll only really know when Intel ships it. Socket 775 and was rumored to be a Xeon socket by some as well.

AMD

So, there's too much Flux for Intel or AMD to point fingers.

Donnie27
 
Nasgul said:
Damn it, it's Intel pulling this kind of crap again?. I'm about to get either an ASUS 945P or 955X because from what I've red (not untill now), it seems like LGA 775 will also be the socket for new upcoming processors (but I guess not).

Donnie, after seeing this thread: 3700 San Diego vs. 3200 Venice: Round 2!, I think I'll have to ditch my thoughts on building my first ever AMD based rig, seeing the San Diego and Venice OC'ed to 2.8 benchmarks, I don't see any difference at all against my Pentium 4 650 OC'ed to 4.0GHZ. I thought they were supposed to be way FAST (er than Pentium 4s) but they aren't.

Yep, I was going to sell my 540 on eBay, get the new and improved fanless design ASUS SLI Premium and a San Diego. But now I'm either getting the P5LD2 or P5WD2.
Click to expand...

Yes but what they do there is waste a lot of time. I run apps, not benchmarks for my main use. They run computers like Drag Racers while I want something like a BMW 735 IL. Now the price cuts are talked about for the end of July but I'm really tired of waiting.

I did see better results from others.

Donnie27
 
You must log in or register to reply here.
Back
Top