Bulldozer vs Nehalem

[pausert20]

I hold my doubts because of lackluster information on how AMD will address leakage (not simple gate leakage but source to drain I.off) which plagued their transition to 65nm and gets worse with further shrinks. If they do it as well as Intel, then their early transition to immersion lithography can help them leapfrog Intel in process quality.

WR, I will need to disagree with your statement that if AMD can transition as well as Intel. From the looks of the Transition from 90nm to 65nm it will just get much worse for AMD on the 65nm to 45nm transition. They don't have Hi K which I believe they will need to get anywhere near Intel's Penryn design in performance. I also don't believe anyone in the industry has better process quality. Intel is the defining manufacturer.

 

[itotallybelieveyou]

AMD cant compete at new vs. old core, so how could they compete at old vs. new architecture?

you never know what can happen.

 

[wickedmonster]

Rumour has it Bulldozer will feature reverse hyperthreading and SSE5 while Nehalem will have full SSE4 and hyperthreading. It will be interesting to see reverse HT vs regular HT.

 

[pausert20]

Wickedmonster, Penryn already has SSE4. I have not heard if Intel plans to provide any additional SSE instructions for Nehalem but yes it will have Hyperthreading enabled in it.

 

[randomizer]

I can tell you firsthand that Nehalem is a monster. It will make you crap your pants.

I believe you, just coz I know that you know :lol:

 

[jaguarskx]

Nothing like paper tiger going at each other.

 

[Kamrooz]

Nehalem is already up and running on windows (the first sample, also runnoing Mac OSX as well). Did anyone watch the intel keynote? Take it with a grain of salt, but they stated there will be a big performance jump, similar to the migration from netburst to core.

The big factor though. AMD delayed bareclona for AGES, they had tons of issues to work out. The main problem is how much time they had to invest with their R&D department to at least get it to where it is today. Personally..I think the Barcelona release may have pushed back their next architecture my a slight margin..How big? I have no idea. But the extremely late release of barcelona will have some consequences in their next architectures development/release.

Penryn is nothing to phenom. By that I mean not a big problem. Penryn pretty much offers a slight increase performance wise to their 65nm brethren (not in certain tasks though thanks to SSE4, where performance in encoding has improved greatly). Needless to say, Intel's 65nm brethren can out pace barcelona at the same clock speed. But Nehalem will be a monster I believe...and AMD's worst nightmare. They are already deep in the hole, now comes the time for intel to fill it up with some dirt to hold down AMD for a bit, how long will it slow down amd?..who knows, but hopefully Bulldozer won't be a second tasting of barcelona...Otherwise I've lost all hope for AMD.

 

[WR]

They don't have Hi K

Hi-K is just a matter of using Hafnium Oxide for the gate insulator and thickening that to reduce leakage; the secret sauce however is in the gate electrode itself. If AMD has something that's both compatible and performs well, then immersion will only make it easier for them to get high yields.

 

[the_vorlon]

For the tiny little bit it is worth...

In an interview with the Inquirer, Kirk Skaugen, Intel's Digital Enterprise Group VP and GM of Server Platforms Group said that the CPU core performance jump from the same process Core 2 (Penryn) to Nehalem would be higher than the jump for Netburst to Core 2 itself.

IE: Nehalem will exceed Core2, by a margin greater than Core2 exceeded Netburst.

~~presumablely~~ Kirk is in a position to actually know, and since Nehalem is up and actually running (likely at low clock speeds, but at least functional) this postulation by Kirk should be based upon real silicon and is not "simulated benchmarks" or something equally daft.

Of couse Phenom was going to be 40% faster than Conroe.. so I guess you never do actually know till silicon comes out...


http://www.theinquirer.net/gb/inquirer/news/2007/10/16/idf-taipei-nehalem-real-big

 

[pausert20]

WR, so you are saying that IBM will have their version of Hi-K and metal gates ready for AMD when they bring out their 45nm process???

I'm betting they won't. I'm betting they might get some processor out by the time Nehalem comes out in 45nm but without Hi-K and metal gates.

 

[yomamafor1]

BullDozer is a TOTAL REDESIGN @ 45nm. It will have SSE5 and probably much more IPC. Montreal will be MCM with more L3.

You also said K10 was a TOTAL REDESIGN... which turned out to be a slightly modified K8, with a shared L3. You also said 1Ghz Barcelona will outrun any Core 2..

Although the rule of thumb is never introduce a new architecture on a new node, but we'll see how AMD does. Given how they executed the last 2 years, I don't see them turnover soon.

Maybe I'll put this into my sig, and we'll wait for next year.

 

[yomamafor1]

WR, so you are saying that IBM will have their version of Hi-K and metal gates ready for AMD when they bring out their 45nm process???

I'm betting they won't. I'm betting they might get some processor out by the time Nehalem comes out in 45nm but without Hi-K and metal gates.

If AMD's own roadmap serves any ring of truth, the next 45nm Deneb will come out in H208, which coincides with Nehelam's launch. Also, according to AMD, HK won't be implemented until the 2nd generation of 45nm, which is likely Bulldozer.

2008~2009 will be the bloodiest year for AMD. Personally I really don't see how they can survive it.

 

[vonhell]

if the Bulldozer is to appear 2009/10 - it might not compete with Nehalem at all...
Nehalem - Q4 2008
Westmere (Nehalem shrink + improvements (?) ) - H2 2009
Sandy Bridge - new uArch - 2010

if we cannot reliably compare a technology existing on paper with technology available only @ intel's labs - how are we to compare 2 uArchs that we have absolutely no info about as of now ?

 

[wickedmonster]

http://chip-architect.com/news/2001_10_02_Hammer_microarchitecture.html

Here's a link to several patents that AMD intended for the original K8 but never made it to silicon. Maybe at that time, process tech was not good enough to include all those things in there. With 45nm, AMD might be able to better include them. Certainly if Bulldozer includes all those upgrades, it will be a monster performer.

 

[yomamafor1]

if the Bulldozer is to appear 2009/10 - it might not compete with Nehalem at all...
Nehalem - Q4 2008
Westmere (Nehalem shrink + improvements (?) ) - H2 2009
Sandy Bridge - new uArch - 2010

if we cannot reliably compare a technology existing on paper with technology available only @ intel's labs - how are we to compare 2 uArchs that we have absolutely no info about as of now ?

IMO, for AMD, coming early is a lot better than coming late. If they can launch 45nm Deneb before Nehalem, then AMD can at least secure some foothold on the market. If they come in after Nehalem, it'll be a lot more difficult for them to persuade buyers from staying with AMD. Afterall, Nehalem will eliminate AMD's architectural advantage at that point, with some major improvements.

According to those with first hand experience with Nehalem, it seems like Nehalem will once give us about 20~40% increase in performance compared Core 2.

The thing is, we cannot compare Nehalem with Bulldozer, as we have no information on how Bulldozer will perform, or what features does it have. However, given K8=> K10's transition, and if this continues, I don't have too high hopes for AMD.

But of course, I'm a d@mn Intel fanboy 😀

 

[jkflipflop98]

I believe you, just coz I know that you know :lol:

Thanks kittah =)

I got to play around with a nehalem system the other day in the lab. I've been building chips for a while now, and I've never seen anything as cool as this.

 

[pausert20]

JK, have you played with one of the Wolfdale processors. From what I have heard you can practically run it with a passive heatsink because it is so cool running.

I'm wondering why Intel does not match AMD's lower 45W DTP for desktop.

 

[monsterrocks]

IMO, for AMD, coming early is a lot better than coming late. If they can launch 45nm Deneb before Nehalem, then AMD can at least secure some foothold on the market. If they come in after Nehalem, it'll be a lot more difficult for them to persuade buyers from staying with AMD. Afterall, Nehalem will eliminate AMD's architectural advantage at that point, with some major improvements.

According to those with first hand experience with Nehalem, it seems like Nehalem will once give us about 20~40% increase in performance compared Core 2.

The thing is, we cannot compare Nehalem with Bulldozer, as we have no information on how Bulldozer will perform, or what features does it have. However, given K8=> K10's transition, and if this continues, I don't have too high hopes for AMD.

But of course, I'm a d@mn Intel fanboy 😀

I don't think a transition is ever gonna happen like that again at AMD; at least not for long time. AMD maay have some stupid people in it, but I think they learned first hand what not to do. Not that I am saying that the transition will be flawless, but rather it will be neater than the last. Plus is the drop old Hector, we might get someone in his position that will help bring the company back into trust with it's customers....cast I don't believe much they say about any of their products right now. But you are right, I think if they don't get their act together; 2008-2009 are gonna be the bloodiest days AMD have every seen.

 

[jkflipflop98]

JK, have you played with one of the Wolfdale processors. From what I have heard you can practically run it with a passive heatsink because it is so cool running.

I'm wondering why Intel does not match AMD's lower 45W DTP for desktop.

Because AMD's new little power naming scheme is the average of what they think they will use. Intel's power measurement is taken when we turn on every transistor on the chip and measure how much heat you would ever have to remove. Positively worst case scenario.

Granted, no one will ever use 100% of a chips transistors at any one time. Hell, if you use 50% at once, you're doin pretty good. But, better safe than sorry. If you actually measure the power draw of a 125W rated 45nm processor, you're likely to discover that you're only using .25 to .5 of that at full load.

Pauser, you can run some Conroes on passive if you want. The new HK/MG really cuts down on the heat (obviously) so you'll be able to run higher clocks than conroe and keep it cooled passively if you like. Or you can put a fan on the puppy and crank the clocks. Your call.

 

[pausert20]

Because AMD's new little power naming scheme is the average of what they think they will use. Intel's power measurement is taken when we turn on every transistor on the chip and measure how much heat you would ever have to remove. Positively worst case scenario.

Granted, no one will ever use 100% of a chips transistors at any one time. Hell, if you use 50% at once, you're doin pretty good. But, better safe than sorry. If you actually measure the power draw of a 125W rated 45nm processor, you're likely to discover that you're only using .25 to .5 of that at full load.

Pauser, you can run some Conroes on passive if you want. The new HK/MG really cuts down on the heat (obviously) so you'll be able to run higher clocks than conroe and keep it cooled passively if you like. Or you can put a fan on the puppy and crank the clocks. Your call.

Thanks JK. Looks like us enthusiasts and Intel will be having a party come 1st QTR 2008. 😀

 

[randomizer]

Nehalem's tiny 32nm transistors won't be able to take the crushing force of my 180nm monsters! All 37 million of them.

 

[yipsl]

I cant wait until AMD releases this chip...(sarcasm) Hopefully Intel has some thing better. : lol :

FUD????

I'll admit that AMD is failing to win the performance crown now. I do remember these boards when they had it and all the Intel supporters (I was one back then) kept pointing to C2D. The reasons I decided on new AMD builds last year instead of C2D were the following:

AMD bought ATI and I prefer ATI chipsets and graphics cards. I can't wait for hybrid Crossfire boards and, eventually, for Fusion.

AMD CPUs were priced low enough that I even built one budget system with an MSI barebones with an Nvidia chipset, but I was disappointed in the 405 chipset, so I might just swap the chip over to a 690G board.

News reports of Intel anticompetitive practices vis a vis AMD in Europe and Asia, as well as America, but they won't prosecute here because it's not as glaring as the rest of our business scandals.

So, if the dual core Phenoms aren't too bad, I'll go for that with hybrid Crossfire, or just get a 6400+.

I'm glad you guys have the warm and fuzzy feeling that you're winning a war, but just keep in mind the FUD of Intel fans anxiously awaiting anything other than Netburst before we even saw whether C2D would compete.

i think AMD will have the next round, why? because they already put out a native quad-core and are tweaking it to make it more efficient and faster, whereas intel is shooting for their first native quad core with Nehalem. Intel will probably run into problems like AMD is right now.

Where the hell do they come up with these names? do they just pick them out of a hat?

I think that Intel will do better with Nehelem because it's 45nm. I don't think AMD will do well with Phenom until 45nm. There are supposedly technical issues with native quad core at 65nm, but I don't recall what they are exactly.

Nehelem is a river in Washington state, I think. Phenom is a lousy name for AMD, they should fire the marketer who came up with it, even if it's a suit at the top.

 

[the yeti]

AMD needs to put 2 6400+ ' s together and call it a quad for now, to raise money and get out of the hole they dug themselves into.


*I dont now what im talking about here really... just that AMD has made the first real quad core, and that could help in the future...

Then they can work on their monster CPU! I think that AMD was just thinking too far ahead by making a native quad core cpu (native does mean that its real quad core, not 2 dual cores together, right?) But because they have more experience in what i think is the definition of NATIVE, they might come out on top again... then they willl fall, and rise in a cycle =]


BUT MOST IMPORTANTLY...

WE CANT LET INTEL TAKE OVER THE CPU INDUSTRY!!!

This would mean bad business for the consumer =[

WE would be forced to pay whatever price intel says for a CPU =0 ZOMG! Then wed all be in trouble!

 

[Joe_The_Dragon]

Intel needs to work on the bus side of things and there dual cpu boards for a long time lagged behind the i/o and chipset choice of the amd dual cpu boards.

And AMD has a dual cpu board with desktop ram and the newer ones will add pci-e 2.0 in all slots with there chipset that will use cross fire and likey there will be a nvidia one with sli and all pci-e 2.0 sli.

Unlike the Intel one with FB-dimms and x16 x16 x16 x16 pci-e 1.1 sli over spilt useing 2 extra nvidia chips each on a pci-e 1.1 x16 bus or cross fire x16 x16 pci-e 2.0 + other pci-e 1.1 slots on a full sever / workstation board that may end costing less then the Skull Trail board and may even have on board hardware SAS raid and more ram slots.

 

[Joe_The_Dragon]

AMD needs to put 2 6400+ ' s together and call it a quad for now, to raise money and get out of the hole they dug themselves into.


*I dont now what im talking about here really... just that AMD has made the first real quad core, and that could help in the future...

Then they can work on their monster CPU! I think that AMD was just thinking too far ahead by making a native quad core cpu (native does mean that its real quad core, not 2 dual cores together, right?) But because they have more experience in what i think is the definition of NATIVE, they might come out on top again... then they willl fall, and rise in a cycle =]


BUT MOST IMPORTANTLY...

WE CANT LET INTEL TAKE OVER THE CPU INDUSTRY!!!

This would mean bad business for the consumer =[

WE would be forced to pay whatever price intel says for a CPU =0 ZOMG! Then wed all be in trouble!

Amd better bus will make a dual dual core or dual quad core on the same cpu work a lot better then intel setup over the same FSB.