Nvidia announces x86 chip *edit: not true*

[Idontcare]

Originally posted by: IntelUser2000
You are stupid.

Is this really necessary?

 

[apoppin]

Originally posted by: IntelUser2000

Originally posted by: apoppin

One people say that Intel will fail with GPUs but they are basing it on their IGP status. Well unlike their IGP Larrabbee is actually meant to make money so they'll put much more resources towards it and care how it turns out.

Here is where i stopped reading

intel makes huge money off their IG .. all of their products are meant to make money ,, especially P4; it did - before their engineers failed to meet their marketing department's unrealistic dreams for it

all i will say is that Larrabeast's success will depend on how well it PERFORMS in 3d aps
-compared directly to AMD's and Nvidia's graphics

You are stupid. Sure they make money off their IG, but they charge a whole $4-5 more than their non-IGP variants. It's going to be nothing like Larrabbee which is a entirely different product and seperate segment itself(unlike IG which comes with a chipset and you don't even need to use it).

They dont' need to waste silicon making an IGP fast as mid-level discrete graphics for $4 when they can do that on lower-end discrete devices which will make real money.

There is nothing you said that is particularly intelligent here. Intel already makes excellent money on their IGP because of the huge volume of their sales; they know how to make the very best out of a slim margin.

You have ZERO clue how Larrabeast will perform or it it will *ever* provide a RoI - "real money" for Intel; unlike their IG which is a solid moneymaker for Intel

If Nvidia and AMD's graphics blow it away, it will be a failure - the way their marketing plays it up, now

 

[Idontcare]

Originally posted by: apoppin
Intel already makes excellent money on their IGP because of the huge volume of their sales; they know how to make the very best out of a slim margin.

We know Intel pretty much designs their products to be capable of generating 50% GM's as a minimum acceptable threshold, with the target GM's around 55-58% and stretch goal of 60% set internally.

I doubt they are operating their IGP division on a slim-margin business status. If the ASP for an IGP chipset is $4 more than the ASP for a non-IGP chipset then I've got to believe $2 of that $4 delta is gross margin profits or Intel wouldn't bother.

Originally posted by: apoppin
If Nvidia and AMD's graphics blow it away, it will be a failure - the way their marketing plays it up, now

When Intel first entered the graphics market over a decade ago they did so with similar fanfare and hype regarding their intentions of pwning the market. By their marketing dept's ambitions those efforts failed, but Intel surely wouldn't call the IGP a failure given the quantity of profits it has added to its bottom line unlike NV and AMD graphic's divisions lately.

If NV's and AMD's flagship products perform better than Larrabee at time of release then surely Intel would know this to be the case before they actually released Larrabee, and since Intel appears to know how to make billions in profits while NV and AMD do not it would then beg the question at that time as to what is to be labeled a failure versus success and why are we (people who are equally incapable of generating billion dollar profits despite our collective business savvy and superior understanding of the graphics industry) so intent on claiming this distinction be made from our couches.

My expectation for Larrabee is that it simply won't be released until it has been iterated internally as many times as it takes before it is rendered in a form capable of blowing away the competition at its time of release. This means if 45nm doesn't have the xtor budget or thermal budget to get enough cores at high enough clockspeeds then 45nm larrabee simply won't be released to the general public, would still make its way to seeding the developers with working silicon models as well as populating a few high-profile render houses to generate as much PR as possible.

The focus for commercial release of Larrabee would then shift to 32nm, etc until Intel manages to outstrip NV and AMD in the performance race (and in the marathon of process technology development, such a race handedly favors Intel).

This is all just my opinion obviously, nothing special about it. But I do have one question, is there anything about Larrabee that we know of which precludes it from becoming the IGP replacement for the existing integrated graphics core post-45nm if for some reason Larrabee does "fail" at being a discreet GPU?

If scaled down to the appropriate number of cores so as to be manufacturable in 32nm process technology whilst adding only $2 to the chipset production cost and generating $4 in sales (hitting the 50% GM requirement) would the performance of such a Larrabee-based IGP exceed that of the existing/planned 45nm IGP in Clarkdale by enough margin as to render it the presumptive successor of the IGP program at Intel?

(so even a failure turns out to be a winner, no less a failure than the past decade of profits from the existing graphics "failure")

 

[apoppin]

i am saying "slim" compared to Discreet Graphics
- i already said that intel maximized their profit expertly and makes big bucks there

"slim" in that many others have tried and failed to make a profit in IG - where intel excels


--

Now your scenario can also be applied to the P4 and NetBust

we know how that turned out


AMD blew it away
- and yet it made huge bucks for intel


My prediction: Intel will come out with decent IG. [yes that IS a "winner"; i never denied] - i am saying the *rest* of their predictions are Smoke and Mirrors - Just Like with NetBust - down to the Same Marketing Spiel and the same people it seems. Intel will never be to compete profitably in the high end as Nvidia and ATi can keep raising the bar
/end prediction

 

[Idontcare]

Originally posted by: apoppin
Now your scenario can also be applied to the P4 and NetBust

we know how that turned out

Yes we do, core debuted and the rest has been history for almost three years now.

What's to keep larrabee from skipping the neturst stage with larrabee and going straight to the core phase?

Originally posted by: apoppin

My prediction: Intel will come out with decent IG. [yes that IS a "winner"; i never denied] - i am saying the *rest* of their predictions are Smoke and Mirrors - Just Like with NetBust - down to the Same Marketing Spiel and the same people it seems. Intel will never be to compete profitably in the high end as Nvidia and ATi can keep raising the bar
/end prediction

I'm by no means attempting to argue that success is guaranteed, failure is certainly an option, but I am at odds to understand why you work so hard to generate the impression that you believe failure is guaranteed and success is not an option. Why are you so certain of this outcome, and why do you seemingly want this outcome to come to pass? Or am I just getting a completely inaccurate perception of this from your posts?

(who doesn't want a successful third contender in the discreet GPU world? except employees and shareholders of the existing one's of course...but you are neither to my understanding)

 

[apoppin]

want it to pass?

i don't see a negative prediction as "wanting" anything

i am giving my own picture of what i see; based on my experience and on intel's past recent history

imo - intel is not going to fail with Larrabeast - no more than they "failed" with P4
- i think they will get much improved IG - almost a guarantee

and a "shot" at high-end graphics

if i get my way and my finances together; i will be a shareholder of all 3 companies

 

[Idontcare]

Fair enough, I'm prepared to consign my perception of your posting preference for a "larrabeast" failure as simply attributable to a misunderstanding on my part then.

Let me ask you a potentially more cerebral question - of what you know of Larrabee's ISA and architecture in relation to what you know of NV's and AMD's GPU ISA and architecture circa 2010 timeframe, what is it about Larrabee that is setting of the red-flags and generating confidence in yourself that Larrabee's failure as a discreet GPU is nearly a sure thing?

I don't follow the architecture/ISA details of the GPU productscape with enough depth to have the confidence in myself to identify the achilles heel of Larrabee, but you do seem to spend a fair amount of your energy thinking about all things GPU so I'd really be appreciative of your insights if you've the time/desire to educate me on it.

 

[apoppin]

"failure as a discreet GPU, a sure thing"?
:Q

let's say, 'failure to meet expectations' - shall we?

in a nutshell, It is gonna be very difficult for Larrabeast to drag x86 overhead along as easily as they - or their supporters -claim. To do so with an economy of cores that makes economic sense is what remains to be seen to be believed.
- Hell, YES .. i *hope* so ... but i think Nvidia and AMD will also be much further along circa 2010 timeframe than they are now making for a difficult moving target for intel to hit

 

[magreen]

Forgive my trolling:

DISCREET/DISCRETE

The more common word is ?discreet,? meaning ?prudent, circumspect?: ?When arranging the party for Agnes, be sure to be discreet; we want her to be surprised.? ?Discrete? means ?separate, distinct?: ?He arranged the guest list into two discrete groups: meat-eaters and vegetarians.? Note how the T separates the two Es in ?discrete.?

http://www.wsu.edu/~brians/errors/discreet.html

 

[Keysplayr]

Originally posted by: apoppin
"failure as a discreet GPU, a sure thing"?
:Q

let's say, 'failure to meet expectations' - shall we?

in a nutshell, It is gonna be very difficult for Larrabeast to drag x86 overhead along as easily as they - or their supporters -claim. To do so with an economy of cores that makes economic sense is what remains to be seen to be believed.
- Hell, YES .. i *hope* so ... but i think Nvidia and AMD will also be much further along circa 2010 timeframe than they are now making for a difficult moving target for intel to hit

Well, that was a "safe" comment at least. Nobody knows guys. It might suck, it might be meh, it might be mainstream, or it might rock. But I honestly don't think Intel, even with all it's mighty 3dlabs and other 3D company buyouts along with their engineers and programmers will come close to Nvidia or AMD when it comes to graphics.

My prediction? The comparable performance of the Volari V8 paired against it's at the time competitors. It will be great for certain games, completely tank in others, and not even work with the rest. At least with Larrabee's first incarnation.

Why do I feel this way? Because Nvidia and AMD were, and always have been light years ahead of intel when it comes to graphics. NV and AMD already have most of the top viable graphics engineers and top tier programmers. If you graduated college with top honors with a degrees up the wazoo in this field, who would you send your resume out to first?
Intel? I think not. Nvidia or AMD comes first. When they don't get hired, then they go to the next best. Yeah, I'm guessing a lot here, and of course could be wrong. I guess I just wanted to shovel my opinion out there along with the rest of everyone else's.

/nite dudes and dudettes.... sleep time.

 

[Idontcare]

Originally posted by: keysplayr2003

Originally posted by: apoppin
"failure as a discreet GPU, a sure thing"?
:Q

let's say, 'failure to meet expectations' - shall we?

in a nutshell, It is gonna be very difficult for Larrabeast to drag x86 overhead along as easily as they - or their supporters -claim. To do so with an economy of cores that makes economic sense is what remains to be seen to be believed.
- Hell, YES .. i *hope* so ... but i think Nvidia and AMD will also be much further along circa 2010 timeframe than they are now making for a difficult moving target for intel to hit

Well, that was a "safe" comment at least. Nobody knows guys. It might suck, it might be meh, it might be mainstream, or it might rock. But I honestly don't think Intel, even with all it's mighty 3dlabs and other 3D company buyouts along with their engineers and programmers will come close to Nvidia or AMD when it comes to graphics.

My prediction? The comparable performance of the Volari V8 paired against it's at the time competitors. It will be great for certain games, completely tank in others, and not even work with the rest. At least with Larrabee's first incarnation.

Why do I feel this way? Because Nvidia and AMD were, and always have been light years ahead of intel when it comes to graphics. NV and AMD already have most of the top viable graphics engineers and top tier programmers. If you graduated college with top honors with a degrees up the wazoo in this field, who would you send your resume out to first?
Intel? I think not. Nvidia or AMD comes first. When they don't get hired, then they go to the next best. Yeah, I'm guessing a lot here, and of course could be wrong. I guess I just wanted to shovel my opinion out there along with the rest of everyone else's.

/nite dudes and dudettes.... sleep time.

Yeah I was hoping for a more substantial rationale based on scrutiny of some of the existing architecture and ISA expectations out there for Larrabee, like this for example:

http://pc.watch.impress.co.jp/...2009/0305/kaigai01.pdf

Saying "they will suck because Intel is doing something new and all things new at Intel are doomed to suck" is an acceptable argument/position/opinion but it is not a very substantive one.

Equally so, saying "they will pwn all who exist today because their name is Intel and all that is Intel will pwn by definition" is an acceptable opinion but again lacks the depth necessary to sustain conversation and ongoing analyses.

I don't know what Larrabee is supposed to have under the hood when it debuts relative to what NV and AMD are expected to have. But I do know what NV and AMD have access to process technology wise in comparison to what the Intel Larrabee engineers have access to and in that vein of thinking I am compelled to consider that even if Larrabee's ISA sucks it will have the wind at its back (as x86 has for 20yrs) thanks to the leading edge process technology with which Intel will produce Larrabee.

But as I've said before I don't see where Intel is doing anything with Larrabee ISA-wise that AMD can't do. NV is not able to pursue Intel into this rabbit-hole, but presumably if it is worth pursuing then AMD would already well know if Larrabee approach is superior to the traditional SPE way of doing things that NV and ATI have been doing.

 

[tommo123]

intel arent doing anything better than AMD/ Nvidia are doing right now - they are simply creating a graphics accelerator from individual x86 cores right? AMD doesnt need to do this since they have ATi and its experience in creating real GPUs.

personally, i think intel doesn't care about graphics as such. the only thing that bothers them is CUDA. I think they are afraid of where that may lead. If something is going to suplant the CPUs importance in a PC/server/whatver, then they want it to be something they create. I'm guessing that larabee will be acceptable for graphics performance, but not where the gamers go. I am also guessing though that it will be a challenger to nvidia via open CL.

 

[IntelUser2000]

Originally posted by: Idontcare

Originally posted by: apoppin
Intel already makes excellent money on their IGP because of the huge volume of their sales; they know how to make the very best out of a slim margin.

We know Intel pretty much designs their products to be capable of generating 50% GM's as a minimum acceptable threshold, with the target GM's around 55-58% and stretch goal of 60% set internally.

I doubt they are operating their IGP division on a slim-margin business status. If the ASP for an IGP chipset is $4 more than the ASP for a non-IGP chipset then I've got to believe $2 of that $4 delta is gross margin profits or Intel wouldn't bother.

Originally posted by: apoppin
If Nvidia and AMD's graphics blow it away, it will be a failure - the way their marketing plays it up, now

When Intel first entered the graphics market over a decade ago they did so with similar fanfare and hype regarding their intentions of pwning the market. By their marketing dept's ambitions those efforts failed, but Intel surely wouldn't call the IGP a failure given the quantity of profits it has added to its bottom line unlike NV and AMD graphic's divisions lately.

If NV's and AMD's flagship products perform better than Larrabee at time of release then surely Intel would know this to be the case before they actually released Larrabee, and since Intel appears to know how to make billions in profits while NV and AMD do not it would then beg the question at that time as to what is to be labeled a failure versus success and why are we (people who are equally incapable of generating billion dollar profits despite our collective business savvy and superior understanding of the graphics industry) so intent on claiming this distinction be made from our couches.

My expectation for Larrabee is that it simply won't be released until it has been iterated internally as many times as it takes before it is rendered in a form capable of blowing away the competition at its time of release. This means if 45nm doesn't have the xtor budget or thermal budget to get enough cores at high enough clockspeeds then 45nm larrabee simply won't be released to the general public, would still make its way to seeding the developers with working silicon models as well as populating a few high-profile render houses to generate as much PR as possible.

The focus for commercial release of Larrabee would then shift to 32nm, etc until Intel manages to outstrip NV and AMD in the performance race (and in the marathon of process technology development, such a race handedly favors Intel).

This is all just my opinion obviously, nothing special about it. But I do have one question, is there anything about Larrabee that we know of which precludes it from becoming the IGP replacement for the existing integrated graphics core post-45nm if for some reason Larrabee does "fail" at being a discreet GPU?

If scaled down to the appropriate number of cores so as to be manufacturable in 32nm process technology whilst adding only $2 to the chipset production cost and generating $4 in sales (hitting the 50% GM requirement) would the performance of such a Larrabee-based IGP exceed that of the existing/planned 45nm IGP in Clarkdale by enough margin as to render it the presumptive successor of the IGP program at Intel?

(so even a failure turns out to be a winner, no less a failure than the past decade of profits from the existing graphics "failure")

OK I was probably too harsh on my previous post, I apologize.

Here are the die sizes for Intel IGPs:
G45: 97.848mm2
G35: 124.45mm2
G965: 101.76mm2
945G: 101.76mm2

A year or two ago I was reading PCWatch article about the G965 chipset(trying to desperately find out performance/specifications) it said that 70% of the die size is taken up by the integrated graphics processor alone.

Back here I estimated Clarkdale/Arrandale GPU+MC+PCI-Express controller portion to be approximately 160mm2: http://forums.anandtech.com/me...ght_key=y&keyword1=G45

Significant amount of the die size for GMCH is taken up by the graphics. They aren't making a lot of money off this. GMCHs are made in a way that is strategically important to Intel, not directly to make money. If people buy Intel's GMCH, they buy their CPUs and service.

Unlike Intel IGPs which are rumored to have lack of driver programmers, Larrabbee is showing significantly better support, with 500 heads working on driver programming division alone: http://forum.beyond3d.com/showthread.php?t=52424

Commitment on the Larrabbee compared to IGPs are on entirely different level, on the levels of their CPUs.

UPDATE: Intel chipset pricing for 4-series

They used to have small price premium with IGP chipsets but not with 4-series:
http://www.dvhardware.net/article25562.html

X48: $70
Q45: $44
Q43: $40
P45: $40
G45: $40
G43: $34
P43: $34

 

[Idontcare]

Originally posted by: IntelUser2000
Back here I estimated Clarkdale/Arrandale GPU+MC+PCI-Express controller portion to be approximately 160mm2: http://forums.anandtech.com/me...ght_key=y&keyword1=G45

Thanks for drawing my attention to that other thread, it had not registered with me that you had posted in it. I've replied now.

Originally posted by: IntelUser2000
Significant amount of the die size for GMCH is taken up by the graphics. They aren't making a lot of money off this. GMCHs are made in a way that is strategically important to Intel, not directly to make money. If people buy Intel's GMCH, they buy their CPUs and service.

Interesting, so its more of a sales volume enhancement endeavor whilst minimizing the downward pressure on GM's as much as possible.

The old question of which is better - 1m units sold at 59% GM or 10m units sold at 57% GM?

For equal or slightly lowered GM they manage to substantially boost revenue (and thus EPS).

Originally posted by: IntelUser2000
Unlike Intel IGPs which are rumored to have lack of driver programmers, Larrabbee is showing significantly better support, with 500 heads working on driver programming division alone: http://forum.beyond3d.com/showthread.php?t=52424

Commitment on the Larrabbee compared to IGPs are on entirely different level, on the levels of their CPUs.

Otellini definitely runs the ship with a different approach than Barret. Under Barrett (CEO 1998-2005) Intel suffered thru Netburst/P4 and Itanium not too mention the graphics storm that never occurred. Under Otellini they've reassembled their CPU line-up with tick-tock, skipping 45nm Itanium and leap-frogging to 32nm, and are organized to bring Larrabee to the market.

 

[apoppin]

Otellini definitely runs the ship with a different approach than Barret. Under Barrett (CEO 1998-2005) Intel suffered thru Netburst/P4 and Itanium not too mention the graphics storm that never occurred. Under Otellini they've reassembled their CPU line-up with tick-tock, skipping 45nm Itanium and leap-frogging to 32nm, and are organized to bring Larrabee to the market.

You hit the nail on the head for me. Intel is "politically" driven. They get their orders from the top down and the entire company shifts to accomplish their goals.
- they are beyond ambitious now. And their timing is set up for "failure" with Larrabeast; i do not believe they learned from their last foray into discrete graphics
- their "pattern" is to me repeating the same way they approached it the last time

i am just looking at it in different way than you are - their goals are too high and their PR hype-machine is in full swing .. it is like with NetBust ..
[my view is more metaphysical]