K8L and 2007

[caligula]

Mark these words: with their new K8L architecture, AMD won't be able to release mid-range processors with the same price/performance like E6300 for example (which will be at a lower price than now and will get a special price cut as a "coupe de grace" or last nail in the coffin for AMD). A quick salvage for them would be to finally introduce 2MB/cache/core - and even with this they won't get even unless their cache is still accessed at half the width Conroe has. They will first release K8L on high-end and server market, where, especially on the latter, is Intel land. Poor OS support, corporate holdback (remember how many years took AMD64 to become adopted in server environments), high prices, small number of initial lab rats willing to buy will make their release as "hey, really nice but who's buying?". Ridiculously trying to save their smashed pride with throwing "leaked previews" of K8L infrastructure, which mainly introduces nifty gizmos like offloading engines for XML (which will stay offline for most of the time) was definitely a bad move. Who cares you say that you will drive a S-klasse Mercedes in 10 years from now on, when you still have your '99 Honda Civic and you get smoked by a guy in a Maseratti ? Even worse, lots of people start driving Maserattis
But, until late 2007 AMD has big chances that it will be history.
If they don't come up with a new something (I dunno, but they should do something), they will lose the train for good. like ATI does for the last 3 generations of GPUs.
Sorry AMD fans, but this is no soccer, when it comes to figures you lost.

[largon]

So, caligula...

You think K8L won't gain enough to compete Conroe/Kentsfield from:
- 33% wider execution, compared to K8
- doubled instruction fetch width
- doubled FPU
- doubled L1 width
- enhanced branch predictor
- 2MB+ L3
- OoO
- nifty gizmos like offloading engines for XML, lol

Also, you honestly think AMD won't be able to supply enough 65nm QC / DC K8L cpus whose die sizes smaller than current DC dies out of their fully 65nm ramped Fab36 with double the capacity compared to Fab30?

[BlackX]

So, caligula...

You think K8L won't gain enough to compete Conroe/Kentsfield from:
- 33% wider execution, compared to K8
- doubled instruction fetch width
- doubled FPU
- doubled L1 width
- enhanced branch predictor
- 2MB+ L3
- OoO
- nifty gizmos like offloading engines for XML, lol

Also, you honestly think AMD won't be able to supply enough 65nm QC / DC K8L cpus whose die sizes smaller than current DC dies out of their fully 65nm ramped Fab36 with double the capacity compared to Fab30?

[arisythila]

The majority of servers are 2 core . Woodcrest beats opties in 2 core server. Soon enough intel will have 4 cores with tigerton . that will beat amd in the 2x 4 core server market. Your optimistic but its unfounded.

In 64 bit mode? Are you sure Intel's Woodcrest will wipe the floor with AMD's Opteron, in 64 bit mode? a Majority of servers in the US, run Linux, or some forum...(Unix, Solaris,ect ect) Not windows.

I think your thinking about 32 bit mode. hehehe silly turtle.

~Mike

[FghtinIrshNvrDi]

largon, if this site had a reputation system, you'd have a point or two flyin your way. Thanks for using actual plans, goals, and outlines put out by AMD.

Ryan

[grimREEFER]

In 64 bit mode? Are you sure Intel's Woodcrest will wipe the floor with AMD's Opteron, in 64 bit mode? a Majority of servers in the US, run Linux, or some forum...(Unix, Solaris,ect ect) Not windows.

I think your thinking about 32 bit mode. hehehe silly turtle.

~Mike

it would beat it in both 32bit and 64 bit modes...

[FghtinIrshNvrDi]

That's what I noticed, and thought was pretty darn cool! AMD has had this architecture out for a long time, and yet it's virtually neck-in-neck when compared clock-to-clock. I don't really care who's winning, I just think it's cool that AMD competes with the aging A64 architecture against the new Intel 'superchip'. Anyone else, or am I just impressed to easily?

Ryan

[SunFlowerSeeds]

I always thought EM64T is similar to AMD64. Apparently it is not the case as preliminary benchmark show not performance gain in 64bit applications compared to it 32bit counterpart using EM64T. In contrast, AMD64 show some improvement in 64bit application than 32bit application.

Actually, hate to burst your bubble, but Intel's 64-bit performance drop (relative to 32-bit) is because of a bug in the way EM64T is implemented. IF they ever decide to fix it, we should see a performance boost. IF . . .

[FghtinIrshNvrDi]

EM64T is an emulation of 64 bit processing. It runs 64 bit info through 32 wide hardware, but has the same end-result. It actually should be hurting performance in 64 bit operation.

The core2 duo's don't have emulated 64 bit, do they? That was a Pentium 4 thing

Ryan

[SunFlowerSeeds]

So EM64T in Conreo not similar to EM64T in P4? Why the benchmark score tend to lower in 64bit than 32bit in Conreo? Please enlighten me.

EM64T is an emulation of 64 bit processing. It runs 64 bit info through 32 wide hardware, but has the same end-result. It actually should be hurting performance in 64 bit operation.

The core2 duo's don't have emulated 64 bit, do they? That was a Pentium 4 thing

Ryan

[FghtinIrshNvrDi]

Hey, I'm not gonna lie. I don't know for sure if Conroe is emulated 64 bit, or true 64 bit operation. I was inclined to think that it was true 64 bit, this far down the timeline. Maybe they dropped due to the bug that Brent brought up.

Ryan

[Bloody_Sorcerer]

Mark these words: with their new K8L architecture, AMD won't be able to release mid-range processors with the same price/performance like E6300 for example (which will be at a lower price than now and will get a special price cut as a "coupe de grace" or last nail in the coffin for AMD). A quick salvage for them would be to finally introduce 2MB/cache/core - and even with this they won't get even unless their cache is still accessed at half the width Conroe has. They will first release K8L on high-end and server market, where, especially on the latter, is Intel land. Poor OS support, corporate holdback (remember how many years took AMD64 to become adopted in server environments), high prices, small number of initial lab rats willing to buy will make their release as "hey, really nice but who's buying?". Ridiculously trying to save their smashed pride with throwing "leaked previews" of K8L infrastructure, which mainly introduces nifty gizmos like offloading engines for XML (which will stay offline for most of the time) was definitely a bad move. Who cares you say that you will drive a S-klasse Mercedes in 10 years from now on, when you still have your '99 Honda Civic and you get smoked by a guy in a Maseratti ? Even worse, lots of people start driving Maserattis
But, until late 2007 AMD has big chances that it will be history.
If they don't come up with a new something (I dunno, but they should do something), they will lose the train for good. like ATI does for the last 3 generations of GPUs.
Sorry AMD fans, but this is no soccer, when it comes to figures you lost.

please point out which part of that intel hadn't been struggling with for 3 years. somehow, even whne a firm produces weaker chips, they still keep going and growing!!! omg. sure, AMD will have to cut back and maybe cut into profit margins a little/lot for 6 months to a full year, but so what? intel had to do it for 3 years while amd reaped massive revenues off chips with massive profit margins (c'mon, an opteron die is <100 bucks a pop, amd told us so. said chips [ie FXes] are 1000+ bucks. do the math).
amd will not "be history". they'll simply be fighting from the weaker side for a while; life moves on. they won't "lose the train for good"; even intel didnt manage to do that when the monstrosity known as netburst was born.

durr.

[caligula]

So, caligula...

You think K8L won't gain enough to compete Conroe/Kentsfield from:
- 33% wider execution, compared to K8
- doubled instruction fetch width
- doubled FPU
- doubled L1 width
- enhanced branch predictor
- 2MB+ L3
- OoO
- nifty gizmos like offloading engines for XML, lol

Also, you honestly think AMD won't be able to supply enough 65nm QC / DC K8L cpus whose die sizes smaller than current DC dies out of their fully 65nm ramped Fab36 with double the capacity compared to Fab30?

Look, I don't say that that's not true and I haven't expected less from them for a CPU that will be released at least 1 year from now on lol. But, by the time it will show up, Intel will be prepared with cost cuts and maybe a new family of processors - I really don't think that their R&D is put to sleep - like AMD's looked like. If it was for AMD to keep the pace, they would have released it now.
To translate your words:
"In 10 years I will drive my S-Klasse!"
"WOOOW" - the audience
"And I will have a v12 engine by then!"
"WOOOW"
"And all leather inside!"
"WOOOW"
"And three spare wheels" - for XML lol
"WOOOW?"
But, you know, a v12 from today may be the standard engine in 10 years.

And about those gizmos, I still believe they're crap. Non standard, no one will adopt them immediately. A complicated architecture is not good, watch the CISC/RISC clash where the latter owns due to simplicity.

[informal]

Actually, hate to burst your bubble, but Intel's 64-bit performance drop (relative to 32-bit) is because of a bug in the way EM64T is implemented. IF they ever decide to fix it, we should see a performance boost. IF . . .

That is nonsense.Please read this,it should clear some bugs of those:

http://babelfish.altavista.com/babel...fkaigai288.htm

http://www.aceshardware.com/forums/r...2277&forumid=1

Simply put:Conroe has some trade-offs when running long mode and they are obviously hurting the performance.AMD on the contrary has no such trade-offs and is doing just fine.

[DoubleZero]

Actually, hate to burst your bubble, but Intel's 64-bit performance drop (relative to 32-bit) is because of a bug in the way EM64T is implemented. IF they ever decide to fix it, we should see a performance boost. IF . . .

Link? Proof? If they decide to fix it? OMG lmao...

[alayashu]

IF they ever decide to fix it, we should see a performance boost. IF . . .

makes you wonder why Yonah didnt have 64b implemented on it
because it's emulation stinks in case u didn't figure out lol

btw Caligula, for a guy that doesn't overclocks chips (most people in the world actually), what do you thing he would choose, an X2 3800 @ 149$ or an E6300 @ 185$ (both equipped with DDR2 800) when the E6300 is only 5-10% faster than the X2 on 32b and much slower on 64b?

[SunFlowerSeeds]

that will depend on how good is vista 64bit.

makes you wonder why Yonah didnt have 64b implemented on it
because it's emulation stinks in case u didn't figure out lol

btw Caligula, for a guy that doesn't overclocks chips (most people in the world actually), what do you thing he would choose, an X2 3800 @ 149$ or an E6300 @ 185$ (both equipped with DDR2 800) when the E6300 is only 5-10% faster than the X2 on 32b and much slower on 64b?

[SunFlowerSeeds]

X2 3800 @ $149 is really band for buck. Check the gaming benchi http://www.xtremesystems.org/forums/...74&postcount=4

[caligula]

X2 3800 @ $149 is really band for buck. Check the gaming benchi http://www.xtremesystems.org/forums/...74&postcount=4

http://www.anandtech.com/cpuchipsets...spx?i=2795&p=2

E6300@183$ (which, with a bit newbie o/c wipes any AMD Dual Core heavily overclocked) is what may be called the bang for the buck.
I wouldn't rely that much on these results, since most GPUs are saturated today even by AMDs. And who's willing NOT to pay 50$ more for an extra 1M of L2 cache and 65nm instead of castrated 1M and 90nm ?

[largon]

E6300@183$ (which, with a bit newbie o/c wipes any AMD Dual Core heavily overclocked) is what may be called the bang for the buck.

QTF.

It's been a while since my last upgrade, thus I've been browsing for a decent board for this cpu...

[`odin]

http://www.anandtech.com/cpuchipsets...spx?i=2795&p=2

E6300@183$ (which, with a bit newbie o/c wipes any AMD Dual Core heavily overclocked) is what may be called the bang for the buck.
I wouldn't rely that much on these results, since most GPUs are saturated today even by AMDs. And who's willing NOT to pay 50$ more for an extra 1M of L2 cache and 65nm instead of castrated 1M and 90nm ?

nah allendale

[caligula]

Figures here: http://www.xbitlabs.com/articles/cpu...duo-e6300.html -> Beat this, AMD zealots.
The processors are not quite released in stores yet so hold your breath a bit Go to the Intel thread and you'll find a lot. Hint: DFI Infinity 975x, Gigabyte GA-965P-DS3 and (a nice new entry) soon-to-be-released Biostar T-Force P965 deluxe -DFI LP bios like. But enough offtopic.

[PetNorth]

My curiosity: Where have you been hidden, you Intel zealots, the last three years?

[arisythila]

http://www.anandtech.com/cpuchipsets...spx?i=2795&p=2

E6300@183$ (which, with a bit newbie o/c wipes any AMD Dual Core heavily overclocked) is what may be called the bang for the buck.
I wouldn't rely that much on these results, since most GPUs are saturated today even by AMDs. And who's willing NOT to pay 50$ more for an extra 1M of L2 cache and 65nm instead of castrated 1M and 90nm ?

Ummm AMD doesnt benefit very much from 1 mb of cache compared to Intel.

So who cares about Cache? Oh yeah, you intel cats do.

~Mike

[caligula]

My curiosity: Where have you been hidden, you Intel zealots, the last three years?

When no arguments are to be said, history is called upon™
ROTFLMAO indeed.
No, I'm not an Intel zealot. When AMD had its time, I used to roll one of the first Oracle/AMD64 servers ever. Used to work with Linux64 at the times it was a bloody unstable patchset. Yet still, I am really amused where humiliation and frustration can lead when you see your team lose.
I kinda support no one. Or - uh - I'm a Power5+ fan. I love my IBM p595. And when the time comes, I will happily replace the damn wardrobe with whatever is at its best at that time.
On the other hand, you guys should wait to go past 18 and stop trolling and insulting with "cats" and "flees".