Can AMD salvage QFX with an in-house chipset?

[elbert]

I think AMD is keeping the pressure on Intel until they get to K8L for their stock prices justification. Without the 4X4 Intel could of had a very profitable year and AMD having a flat year would have killed their stock.

This is where I will respectfully disagree -- it isn't Intel currently underpressure from a performance prespective.... and the marketshare trend is turning around.

http://www.economist.com/business/displaystory.cfm?story_id=8355132


While the MSS trend has been in AMD's favor, and the pressure was on Intel in the last quarter of 2005/1H 2006, it would appear they have turned the corner.... C2D has, so far, caused AMD to drop prices on the DT > 60%, offer rebates on most all remaining stock, and basically rush out a 4x4 solution.

Now, this is just what it appears --- it may trend off lower or trend off higher. But, after studying the dynamic of the MSS for DT, Sever, and mobile it is beginning to look like Intel's MSS trend is turning around. AMD did not take share from Intel in Q3, and if you scrub for the Via affect, Intel gained in both sever and DT (AMD had a great mobile showing however). It will take a few quarters to see how much pressure comes from one side or the other.... but AMD is currently, from a technical perspective, on the defensive.
Intel kind of is under pressure to stay ahead of AMD. If Intel cant hold the lead for a while Intel could become the number 2 CPU maker. What I was point out is AMD is trying to justify their low earings and the only way they can is if Intel has low earnings. Intel releases low earnings then AMD's low earnings are justifyed. On the other hand if Intel release high earnings then AMD's low earnings could cause their stock to be down graded.

This is part of the reason AMD released 4X4 which couldnt beat C2Q. They could have waited for 65nm FX's so the power numbers would look better. 4X4 will force Intel to cut C2D prices instead of charging higher prices for C2Q.

P.S. I truly enjoy talking tech with you. You have a great understanding of technology.

 

[elbert]

I dont agree with BM because I dont think AMD has to salvage QFX for any chipset. Lets look at what AMD set out to do with 4X4. Beat C2D and even the lowest FX-70 does just that at a lower current price than the top C2D. Did 4X4 loss against the C2Q? Its a mobo so it really must be judged for what it was designed. It offers about the same GHz per GHz increase as C2Q with C2Q getting the best percentage on games and the 4X4 on multitasking.

What are the downfalls of both implementation of quad core? The 4X4 increases memory latency and costs more on the motherboard side. The C2Q6700 losses about 600MHz off its top OC over C2D's 4GHz and about 370MHz off its top released C2D6800.

What are the gains for both implementations of quad core? The FX-74 gains in refinement to give an added 200MHz over AMD's top FX-62. The 4X4 has a better upgrade path as its set for 2X-Agena FX's. The C2Q motherboard is cheaper than the 4X4 and one isnt forced to upgrade if Agena FX's performance is worse than expected.

Which company losses here? Both companies loss because this is a tech war. Intel losses due to having to modify a core well short in the cores life cycle and its top C2D6800 falling to 6th over all in about 4 months. AMD losses because the 4X4 and FX's will cut into its opty sales and could be a bad return on investment in the short term.

Which company wins? Neither because only we the consumer have won with cheaper faster CPU's.

Interesting perspective and a very logical argument given the timing, though I would disagree (respectfully of course 😀 ) on the upgrade path option for AMD. Due to the wafer usage and binning, I dont think AMD will be able to price Agena agressively enough to make selling them in box sets of 2 an attractive option for all but a select few consumers with excessive money to burn.
Why do think you must buy a box set? Agena FX may work with only 1 CPU on some 4X4's. Technology that grows with your needs comes to mind.

Another interesting point, but if you only want to upgrade to one CPU, why bother buying a 4x4 in the first place?
Technology that grows with your needs comes to mind.

 

[JonathanDeane]

Buy a Hummer (H1) and a 4X4 !!! Sounds like the same customer experience to me

 

[turpit]

I dont agree with BM because I dont think AMD has to salvage QFX for any chipset. Lets look at what AMD set out to do with 4X4. Beat C2D and even the lowest FX-70 does just that at a lower current price than the top C2D. Did 4X4 loss against the C2Q? Its a mobo so it really must be judged for what it was designed. It offers about the same GHz per GHz increase as C2Q with C2Q getting the best percentage on games and the 4X4 on multitasking.

What are the downfalls of both implementation of quad core? The 4X4 increases memory latency and costs more on the motherboard side. The C2Q6700 losses about 600MHz off its top OC over C2D's 4GHz and about 370MHz off its top released C2D6800.

What are the gains for both implementations of quad core? The FX-74 gains in refinement to give an added 200MHz over AMD's top FX-62. The 4X4 has a better upgrade path as its set for 2X-Agena FX's. The C2Q motherboard is cheaper than the 4X4 and one isnt forced to upgrade if Agena FX's performance is worse than expected.

Which company losses here? Both companies loss because this is a tech war. Intel losses due to having to modify a core well short in the cores life cycle and its top C2D6800 falling to 6th over all in about 4 months. AMD losses because the 4X4 and FX's will cut into its opty sales and could be a bad return on investment in the short term.

Which company wins? Neither because only we the consumer have won with cheaper faster CPU's.

Interesting perspective and a very logical argument given the timing, though I would disagree (respectfully of course 😀 ) on the upgrade path option for AMD. Due to the wafer usage and binning, I dont think AMD will be able to price Agena agressively enough to make selling them in box sets of 2 an attractive option for all but a select few consumers with excessive money to burn.
Why do think you must buy a box set? Agena FX may work with only 1 CPU on some 4X4's. Technology that grows with your needs comes to mind.

Another interesting point, but if you only want to upgrade to one CPU, why bother buying a 4x4 in the first place?
Technology that grows with your needs comes to mind.

Yes but again, why bother in investing in a dual socket platform now, only to place a single proc in it later? Especially when the prices are astronomical (figuratively speaking of course) now and bound to decrease with time.

You can reply with "Technology that grows with your needs comes to mind" again, and it is a valid point, however, it may better suited to the server arena rather than the DTPC, especially considering the current rate
of developement.

My point is, for the home user, investing in the expensive dual socket platform now (which for the time will not be significantly challenged by typical software) at a very steep entry price, does not make sence when, as all they always do, prices will drop later, and by that time, I cant help but believe single socket solutions of equivelent or better performance will probably be available at a lower cost. Especially considering current procs and that current benchmarks show the single socket outperforming multi socket right now.

Additionally, I cant really see the main stream OEMs enthusiastically getting behind personal use dual socket systems, and with the exception of a few hold outs, the DIYers either if an equivilent or better performing system is available in single socket and costs signifcantly less. Thinking about expandability 1 or 2 years down the road with a setup thats future is far from certain right now does not engender feelings of security either.

Just my thoughts

 

[BaronMatrix]

And what graphics card were they using?

Anand's tests used 1 8800GTX. He got CnQ working and idle dropped dramatically.

 

[elbert]

Simply to buy 1 for increased performance and later another for double the performance at a great savings. Its true the motherboard is costly as any 2 socket from Intel or AMD would be. The price however of 2 motherboards and 2 types of memory would be more over the same time frame. On a 2 socket motherboard your reuse of both CPU and memory would give added saves.

 

[turpit]

Simply to buy 1 for increased performance and later another for double the performance at a great savings. Its true the motherboard is costly as any 2 socket from Intel or AMD would be. The price however of 2 motherboards and 2 types of memory would be more over the same time frame. On a 2 socket motherboard your reuse of both CPU and memory would give added saves.

Yes but the same argument of resuing RAM, PSU HDD etc will hold true for any single socket mobo as well, be it dual, quad or octo core CPU supporting, or whatever else the future holds fors us. My ASUS P5WDH supports C2Q, at a significantly lower investment cost than a 4x4 mobo, PSU and RAM. so I can easily upgrade to better performance than QFX at a signifcantly lower outlay.


Im wondering, are you thinking that multi socket is the next step beyond multicore? If so, then your point is definative, and mine moot. However, I would still debate that waiting vs investing now is the wiser route. On the other hand, if single socket is going to be around for some time, and both AMD and intel progress to octo core as I suspect they will, then the multi socket mobo may not be the most cost effective or best performing way way to access future expandability.

 

[BaronMatrix]

My ASUS P5WDH supports C2Q, at a significantly lower investment cost than a 4x4 mobo, PSU and RAM. so I can easily upgrade to better performance than QFX at a signifcantly lower outlay.

I hesitate to agree fully. Since QFX was designed to be FX, it will take a few months to get more mobos. I figure there will turn out to be at least 5 by March.

The key is for AMD to actually support the platform and not act like it was a failed attempt to beat C2Q. In my mind it was the best way to continually distinguish FX. There will be an X4 next year for the single socket desktop.

Once the Sus mobo goes out to retail ( the initial word was it wouldn't for the first few months), we will see what it is running. Asus" STriker for C2Q is currently $400 so th epinnacle of mobos should be similarly priced.

OF course you can't get a generic board yet for QFX but why would you put a $1000+ CPU on a "cheap" mobo?

 

[apache_lives]

Oh yeah, I won't be buying because current G80s are too expensive and hot. I may wait for R600.

ROFLMAO!!!!!!!

So Baron, you wont buy a "too hot and expensive" video card, BUY YOU WILL BUY AN OVERPRICED, HOT (AND SLOWER) AMD 4X4 SYSTEM?

:lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:

 

[BaronMatrix]

Oh yeah, I won't be buying because current G80s are too expensive and hot. I may wait for R600.

ROFLMAO!!!!!!!

So Baron, you wont buy a "too hot and expensive" video card, BUY YOU WILL BUY AN OVERPRICED, HOT (AND SLOWER) AMD 4X4 SYSTEM?

:lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:


Of course. Just like the idiots who buy Opteron 2xxx.

 

[BaronMatrix]

Oh yeah, I won't be buying because current G80s are too expensive and hot. I may wait for R600.

ROFLMAO!!!!!!!

So Baron, you wont buy a "too hot and expensive" video card, BUY YOU WILL BUY AN OVERPRICED, HOT (AND SLOWER) AMD 4X4 SYSTEM?

:lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:

It is called Baron nLogic®.... and yes it is funny.



So how does it feel to have a wife and kid and

NO LIFE?

 

[mesarectifier]

So how does it feel to have a wife and kid and

NO LIFE?

Is that not a contradiction in terms, Baron?

 

[BaronMatrix]

So how does it feel to have a wife and kid and

NO LIFE?

Is that not a contradiction in terms, Baron?

Exactly.

 

[gOJDO]

 

[BaronMatrix]

At least I'm not a stalker.

 

[turpit]

My ASUS P5WDH supports C2Q, at a significantly lower investment cost than a 4x4 mobo, PSU and RAM. so I can easily upgrade to better performance than QFX at a signifcantly lower outlay.

I hesitate to agree fully. Since QFX was designed to be FX, it will take a few months to get more mobos. I figure there will turn out to be at least 5 by March.

The key is for AMD to actually support the platform and not act like it was a failed attempt to beat C2Q. In my mind it was the best way to continually distinguish FX. There will be an X4 next year for the single socket desktop.

Once the Sus mobo goes out to retail ( the initial word was it wouldn't for the first few months), we will see what it is running. Asus" STriker for C2Q is currently $400 so th epinnacle of mobos should be similarly priced.

OF course you can't get a generic board yet for QFX but why would you put a $1000+ CPU on a "cheap" mobo?

"Cheap" is a relative term. While the example of the P5WDH is "cheap" relative to the Striker Extreme (at $222 US vs $331 US) , I would hardly refer to the P5WDH as a "cheap" board in terms of cost, performance, expandability or capability, and, as I pointed out, it is upgradable to C2Q making it a much more cost effective route to quad core, if thats what a person wants.

$331 striker
http://www.ajump.com/ajump/product.asp?pf%5Fid=4940163&dept%5Fid=5109

$222 P5WDH
http://www.zipzoomfly.com/jsp/ProductDetail.jsp?ProductCode=247085&prodlist=pricegrabber

As for AMD supporting QFX, from a PR perspective, I dont see that they have much choice as failing to support it would publically label it exactly as you stated, "a failed attempt to beat C2Q".

Unfortunately for AMD, being as resource limited as they are, relative to Intel in terms of both production capacity and R&D, devoting any portion of their limited resourses to pursuing what at this moment is at best a "niche" market may not be the wisest investment. Especially if that investment fails to produce what must have been their intended results, winning back the performance crown and reclaiming the related PR.

IMO AMD would be much better served concentraing their resources on their next gen Uarch rarther than pursuing further optimizations or expansions of K8. While refining their already strong Uarch worked spectacularly for 3 years, C2Ds Uarch changed that, and no amount of optimization or expansion will reverse that fact, or even delay the inevitible. AMD really should have poured every ounce of themselves into K8L and 65nm.

One of the things I find truly ironinc about 4x4, is that so many people refer to C2Q as a "glued" chip, when in effect, all QFX really is, is a "glued" mobo. Intel used 2 C2Ds to create its quad core where as AMD used 2 motherboards "glued" together to achieve a "quad core" system.

 

[BaronMatrix]

IMO AMD would be much better served concentraing their resources on their next gen Uarch rarther than pursuing further optimizations or expansions of K8. While refining their already strong Uarch worked spectacularly for 3 years, C2Ds Uarch changed that, and no amount of optimization or expansion will reverse that fact, or even delay the inevitible. AMD really should have poured every ounce of themselves into K8L and 65nm.

One of the things I find truly ironinc about 4x4, is that so many people refer to C2Q as a "glued" chip, when in effect, all QFX really is, is a "glued" mobo. Intel used 2 C2Ds to create its quad core where as AMD used 2 motherboards "glued" together to achieve a "quad core" system.

But that's what they are doing with this. They are providing the platform forst to distinguish the quad cores better. I mean if QFX is any indication of what Agena (single socket desktop quad) can do, then they will need to use more than an unlocked multiplier to differentiate X4 and QFX.

Because the concept of two sockets already exists in the server wksta market it is innovative to move the paradigm to the desktop, epecially since Agena FX isn't due until Aug/Sep and they need a way to increase the perf of FX in the meantime. Not because of Core 2 but because 90nm FX has little to no headroom left.

 

[turpit]

IMO AMD would be much better served concentraing their resources on their next gen Uarch rarther than pursuing further optimizations or expansions of K8. While refining their already strong Uarch worked spectacularly for 3 years, C2Ds Uarch changed that, and no amount of optimization or expansion will reverse that fact, or even delay the inevitible. AMD really should have poured every ounce of themselves into K8L and 65nm.

One of the things I find truly ironinc about 4x4, is that so many people refer to C2Q as a "glued" chip, when in effect, all QFX really is, is a "glued" mobo. Intel used 2 C2Ds to create its quad core where as AMD used 2 motherboards "glued" together to achieve a "quad core" system.

But that's what they are doing with this. They are providing the platform forst to distinguish the quad cores better. I mean if QFX is any indication of what Agena (single socket desktop quad) can do, then they will need to use more than an unlocked multiplier to differentiate X4 and QFX.

Because the concept of two sockets already exists in the server wksta market it is innovative to move the paradigm to the desktop, epecially since Agena FX isn't due until Aug/Sep and they need a way to increase the perf of FX in the meantime. Not because of Core 2 but because 90nm FX has little to no headroom left.


But why bother? Personally, I dont think 4core needs any validation be it AMDs solution or Intels. That goes back far enough now that I cant even remember specifically when or where to find the proof, so you can take this with an entire bag of salt, but when the entire industry believed that CPUs were closing in on an insurmountable MHz barrier, the solution they were looking at was multiple cores. Here they are. And I hope OFX is not any indication of Agenas capabilities. Personally, I hope AMD has much greater goals/success in mind for Agena. For the time being, the old tactic of refinement will not serve them, they need to get K8L and 65nm out the door. Working to validate a concept with old tech is stealing resources from developing the new tech, and is not nessasarily a practical approach, nor may it provide conclusive results.

From a perspective of how to implement multple cores, whether it be 2 dual core on 2 sockets or 4 cores on a single socket, (or 4x2 or 8x1 etc) it falls into the catagory of 6 of one, half dozen of another. There is nothing that a dual socket mobo can do that a single cant. Obviously 2 CPUs in 2 sockest will perform better than a single socket/CPU, but when you start mutliplying cores, the situation now becomes dynamic and devolves to hardware I/O capabilty. Going the multi socket route simply drives prices up as AMD has currently implemented it. Remember why dual socket server mobos exist in the first place...to put more power in a smaller space. Multicore is simply a further extension of that, more power with a smaller footprint.

Now, if AMD chooses to go the multi socket route and Intel choose to go the single socket route, thats their descision, but again, AMDs current "solution" to achieve quad core is not proving to be as effective as Intels, even if people choose to disdain Intels "glued" approach.

Futhermore, while 4x4 will be upgarable to acheive 4cores per socket, assuming that 4 cores per die ( or 4x2 or whatever iteration you want to play with)are the max is a bit premature. It is also a bit of a stretch to assume Intel will not have an octocore, or some equivelent, even if that itself is multisocket. I think Intel learned its lesson, and I dont think they are going to let AMD beat them to a DT 8 core system be it 8x1 or 4x2.

 

[elbert]

Simply to buy 1 for increased performance and later another for double the performance at a great savings. Its true the motherboard is costly as any 2 socket from Intel or AMD would be. The price however of 2 motherboards and 2 types of memory would be more over the same time frame. On a 2 socket motherboard your reuse of both CPU and memory would give added saves.

Yes but the same argument of resuing RAM, PSU HDD etc will hold true for any single socket mobo as well, be it dual, quad or octo core CPU supporting, or whatever else the future holds fors us. My ASUS P5WDH supports C2Q, at a significantly lower investment cost than a 4x4 mobo, PSU and RAM. so I can easily upgrade to better performance than QFX at a signifcantly lower outlay.


Im wondering, are you thinking that multi socket is the next step beyond multicore? If so, then your point is definative, and mine moot. However, I would still debate that waiting vs investing now is the wiser route. On the other hand, if single socket is going to be around for some time, and both AMD and intel progress to octo core as I suspect they will, then the multi socket mobo may not be the most cost effective or best performing way way to access future expandability.
The PSU and HDD's are unimportant to the argument as their life span ranges past the usable life span of the motherboard and CPU's. The thing here is you may buy 2 GB's of DDR2 for your current Motherboard then on your new single socket buy 4 GB's of DDR3. Thats a 4GB's on your new system. On a 2 socket Motherboard you could buy 1 CPU now and 2GB's of DDR2 at about the same price. Later buy the same CPU at a very low price and 4GB's of DDR2. Your new system now has 6GB's compared to your new single socket system with only 4GB's.

Given right now the performance of a single FX isnt the same as the C2Q does but by summer the Agena FX should be about an equal to the C2Q. That said Intel should have an answer for 4X4 sometime so yes I see all but the cheapest motherboards being 2 socket. The server side will move to 4 sockets as some point. I agree as with "waiting vs investing now" as I have stated a 4X4 is a bad short term investment. The 65nm may help atleast in price and power requirements but even then a 4X4 on 1207+ is the best choice.

 

[turpit]

Given right now the performance of a single FX isnt the same as the C2Q does but by summer the Agena FX should be about an equal to the C2Q. That said Intel should have an answer for 4X4 sometime so yes I see all but the cheapest motherboards being 2 socket. The server side will move to 4 sockets as some point. I agree as with "waiting vs investing now" as I have stated a 4X4 is a bad short term investment. The 65nm may help atleast in price and power requirements but even then a 4X4 on 1207+ is the best choice.

Ahh, well, there is the root of the discussion. Your are of the opinion that multisocket is the next step. My opinion is otherwise, at least for the next few years, so for us to continue debating that portion of the converstaion is circular.

However, as far as upgradability, I dont see single socket mobos being any less "upgradable" than multi socket. Yes, you can put 8 sticks of 2GB Ram in the 4x4 vs 4 sticks in the single, but if you are upgrading incrementally over time as your post seems to be saying, that seems to presume that mobo/chipset developement will stagnate. While Ive seen the "socketed" chipset concept batted about, I dont really see that as a viable option. So you seem to be of the mind that buying a mobo and expecting to be able to upgrade the hardware over a 4-6 year period based on capacity is more cost effective that upgrading a lower capacity system every 2-3 years. I would disagree with that based on the current rate of tech advance. Additionally, I see no reason why in the future, either near or far, mobo manufacturers cant or wont simply ad more I/O interfaces to support more system hardware. As such a single socket system designed for octocore should be just as incrementaly expadable as a dual socket system supporting quadcores.

Since I dont have any info as to what route Intel or AMD are going to take, the above argument also falls into the catagory of circular.

We are just going to have to wait and see which way the manufacturers go.

 

[BaronMatrix]

And I hope OFX is not any indication of Agenas capabilities. Personally, I hope AMD has much greater goals/success in mind for Agena. For the time being, the old tactic of refinement will not serve them, they need to get K8L and 65nm out the door. Working to validate a concept with old tech is stealing resources from developing the new tech, and is not nessasarily a practical approach, nor may it provide conclusive results.

What I meant was that Agena will use one socket and 4 RAM slots so it won't be subject TO ANY LATENCY PENALTIES. WIth the "assumed" increased perf we can say that Agena will be to K8 what C2Q is to Core Duo.

I want 2 sockets without ECC and so do a lot of people. Thsi is a good step. If MAD keeps it as a staple platform then there will be new chipsets and mobos ( the first ones ALWAYS incurs R&D costs) that will be cheaper. Intel thought that Itanium would be the desktop 64bit OS but it turned out to be AMD64.

I think the high-end will follow this trend for 2008 since BullDozer will at least match Core 2 perf in most areas based on different analyses.

 

[BaronMatrix]

And I hope OFX is not any indication of Agenas capabilities. Personally, I hope AMD has much greater goals/success in mind for Agena. For the time being, the old tactic of refinement will not serve them, they need to get K8L and 65nm out the door. Working to validate a concept with old tech is stealing resources from developing the new tech, and is not nessasarily a practical approach, nor may it provide conclusive results.

What I meant was that Agena will use one socket and 4 RAM slots so it won't be subject TO ANY LATENCY PENALTIES. WIth the "assumed" increased perf we can say that Agena will be to K8 what C2Q is to Core Duo.

Let's hope Agena does better than that, otherwise it will suck relative to C2Q.... actually, that would be just as funny as the 4x4 launch if they could not beat C2Q, maybe it would put to rest this mis-informed and misguided 'glued' together mentality.

I have never had any qualms with Intel's choice to place two cores on a die. I am talking about single threaded perf. Kuma/Bulldozer has the goods.

 

[turpit]

What I meant was that Agena will use one socket and 4 RAM slots so it won't be subject TO ANY LATENCY PENALTIES. WIth the "assumed" increased perf we can say that Agena will be to K8 what C2Q is to Core Duo.

That falls squarely into the catagory of we are going to have to wait and see. But frankly, I dont see that AMD needed a 4x4 platform to validate latency.

I want 2 sockets without ECC and so do a lot of people. Thsi is a good step. If MAD keeps it as a staple platform then there will be new chipsets and mobos ( the first ones ALWAYS incurs R&D costs) that will be cheaper. Intel thought that Itanium would be the desktop 64bit OS but it turned out to be AMD64.

I think the high-end will follow this trend for 2008 since BullDozer will at least match Core 2 perf in most areas based on different analyses.

It is a step, but it is only a step down a path. Whether or not the path is a safe one or not remains to be seen. Right now, after the first step it ceratinly does not appear to be so. However, it is only one step, and should AMD continue to walk this path, the further they go, the outcome may change. In the meantime, having taken the step and seen the initial results, AMD has to decide whether to keep walking or not. Again, I dont see that they have much choice in order to save face.

AMD does not need to be looking to catch Core2 in 2008. Trying to catch up 1 1/2 years down the road is already a lost battle. They need to be looking to trump in spades. I do not believe Intel will remain stagnent with Core2 as they did with heatburst, and even if they do, if they make the transition to the 45nm node succesfully in 08, the point is moot. The events of this year indicate Intel has learned its lesson. Events also seem to indicate that AMD became complacent with Intels Heatburst ineptitude and wrongly expected it to continue unabated.

Had Intel stayed pat with C2D, the tune people are singing in regards to 4x4 right now would be different. Intel hasnt stayed pat and 4x4 falls shorts as a solution to Intels current C2Q, both in terms of performance and expense. Future optimizations will improve 4x4s perfromance, and costs will decrease, but the same will most likely hold true for Intels products. Again, AMD needs devote all its resources to K8L and 65nm instead of pursuing improvements or trickery for a surpassed Uarch. Developing a platform now, that will be implementable under K8L is not totally unsound, but again, it is sucking valuable resources from K8L, and the concept of multi socket for the home DTPC is hardly proven. AMD did not take the crown back with 4x4 and they probably wont for the next six months if ever. With that loss, they lost the valuable PR that crown holds when it comes to sales. Agena is still a big huge if. I wont assume its going to perfrom well until its on the streets, just as I wouldnt make presumptions with C2D. 4x4 is also not likely to be profitable for AMD any time in the near future if ever, so in the end, all they've accomplished with 4x4 to date is to suck resourses way from K8L. 4x4 is going to look like a bad choice for some time to come.

Finally, be carefull about presuming that single core is limited because of slots or busspeed. It will be easy enough for the chipset/mobo vendors to present an 8 slot non EEC ram/20 SATA mobo single socket mobo that it will support an 8 core CPU if they see that as what the market wants. Much easier than multi socket.

 

[elbert]

Given right now the performance of a single FX isnt the same as the C2Q does but by summer the Agena FX should be about an equal to the C2Q. That said Intel should have an answer for 4X4 sometime so yes I see all but the cheapest motherboards being 2 socket. The server side will move to 4 sockets as some point. I agree as with "waiting vs investing now" as I have stated a 4X4 is a bad short term investment. The 65nm may help atleast in price and power requirements but even then a 4X4 on 1207+ is the best choice.

Ahh, well, there is the root of the discussion. Your are of the opinion that multisocket is the next step. My opinion is otherwise, at least for the next few years, so for us to continue debating that portion of the converstaion is circular.

However, as far as upgradability, I dont see single socket mobos being any less "upgradable" than multi socket. Yes, you can put 8 sticks of 2GB Ram in the 4x4 vs 4 sticks in the single, but if you are upgrading incrementally over time as your post seems to be saying, that seems to presume that mobo/chipset developement will stagnate. While Ive seen the "socketed" chipset concept batted about, I dont really see that as a viable option. So you seem to be of the mind that buying a mobo and expecting to be able to upgrade the hardware over a 4-6 year period based on capacity is more cost effective that upgrading a lower capacity system every 2-3 years. I would disagree with that based on the current rate of tech advance. Additionally, I see no reason why in the future, either near or far, mobo manufacturers cant or wont simply ad more I/O interfaces to support more system hardware. As such a single socket system designed for octocore should be just as incrementaly expadable as a dual socket system supporting quadcores.

Since I dont have any info as to what route Intel or AMD are going to take, the above argument also falls into the catagory of circular.

We are just going to have to wait and see which way the manufacturers go.
Currently you see proof due to while Intel losses 600MHz on the C2Q over the C2D's max OC. This is at the same nm which means alot as to which motherboards new CPU's are released. In trying to double the cores this will always occur on a single CPU and the Q6700 losses MHz to the top released C2D. AMD on the 4X4 gains 200MHz in refinements on their FX and 90nm by part using the 2 sockets. If AMD tryed 4 cores on the 90nm it would run about 1.4~1.6GHz. The 2 sockets is the only reason FX-7X's beat the C2D.

Your point is valid but only if you stay on the bleading tech edge as the second CPU and upgrade of RAM on a 2 socket motherboard would be much cheaper for about 2 times the performance.

No, The same time frame because waiting longer to upgrade the 2 socketed system would cost more as prices rise once you get near the end of CPU or memorys useful life. The 2 socket motherboard would be upgrade about 18 to 24 months for the best savings on the upgrades and stay in use for about 72 months.

True the AM2 can go to quad core but at the same time the 4X4 can go up to 2X quad cores.

 

[1Tanker]

So how does it feel to have a wife and kid and

NO LIFE?

You mean, how does it feel to get laid, without having to pay for it, or blow-up your partner(blow-up doll) ?

PS. It's nice, and if your personality ever changes for the better(unlikely), you might find out.