How AMD Can Beat Intel

[cnumartyr]

Get B3 Phenoms out that will OC to 3.4+ GHz on air.

Sell Black Edition everything.

Since you have to pay $1000 for an unlocked multiplier on Intel... I would DEFINATELY move to AMD if Phenom is close enough clock for clock with C2D and had an unlocked multiplier.

Not even joking, the 5000+ BE was a great idea (even though it was WAY overhyped at the time.. dark side? cmon PR you suck). Since Intel won't sell one for that cost, it would be a great idea for AMD to start selling cheap unlocked CPUs.

It might not clock as high as a QX9650.. however. It won't cost $1000+. Also.. let's be honest.. The highest mutiplier used for the Penryns will be 8.5x. To get that 4 GHz mark on retail non-extreme Penryns you would need a 470 MHz Bus. OC'd Penryns will probably end up FSB limited before 500 just like Kentsfields. If not.. then I am completely wrong.

So a B3 Phenom in the $300-400 bracket labeled as Black Edition that can clock to 3.4 GHz on air? Yea I would buy it in a second over Intel.

Otherwise, I'm getting a Penryn.

Edit: But yea.. this is really the only way I see AMD making a big come back in the enthusiast segment. Other than people buying into the entire "platform" idea.

 

[itotallybelieveyou]

3.4? Penryn does 4.0. Move to 25nm core fast?

 

[cnumartyr]

3.4? Penryn does 4.0. Move to 25nm core fast?

QX Penryns do 4 for $1000+

Haven't seen what "lower" end Penryns will do. Haven't seen what the wall will be on them just yet with a "low" multiplier.

My point was if Phenom X4 BE could be binned at 2.6 GHz, priced around $350, and be OC'd to 3.4+ GHz on air it would be a great price/performance option versus Intel on the enthusiast segment. And this is entirely possible if B3 is the final "golden" revision.

 

[itotallybelieveyou]

The flagship CPU spot seems really important to me. If AMD can make a move on it they would come back for sure.

 

[cnumartyr]

The flagship CPU spot seems really important to me. If AMD can make a move on it they would come back for sure.

Flagship vs Flagship I don't think AMD has a chance. I don't think it's a realistic goal for them to beat Intel there.

Personally I think it's great that QX9650s are hitting 4.2 GHz on air. However that doesn't matter to me. My budget for CPUs in new builds is normally $300-500 max. So perhaps I should refer to it more as "mainstream." However if Phenom was able to compete with Intel in that segment with a BE Phenom, or even a Phenom that would OC decently on the B3 revision I would be happy to switch back.

 

[caamsa]

Get B3 Phenoms out that will OC to 3.4+ GHz on air.

Sell Black Edition everything.

Since you have to pay $1000 for an unlocked multiplier on Intel... I would DEFINATELY move to AMD if Phenom is close enough clock for clock with C2D and had an unlocked multiplier.

Not even joking, the 5000+ BE was a great idea (even though it was WAY overhyped at the time.. dark side? cmon PR you suck). Since Intel won't sell one for that cost, it would be a great idea for AMD to start selling cheap unlocked CPUs.

It might not clock as high as a QX9650.. however. It won't cost $1000+. Also.. let's be honest.. The highest mutiplier used for the Penryns will be 8.5x. To get that 4 GHz mark on retail non-extreme Penryns you would need a 470 MHz Bus. OC'd Penryns will probably end up FSB limited before 500 just like Kentsfields. If not.. then I am completely wrong.

So a B3 Phenom in the $300-400 bracket labeled as Black Edition that can clock to 3.4 GHz on air? Yea I would buy it in a second over Intel.

Otherwise, I'm getting a Penryn.

Edit: But yea.. this is really the only way I see AMD making a big come back in the enthusiast segment. Other than people buying into the entire "platform" idea.

I think you have some good points.....I was kinda thinking the same thing. Gimmicks sell but these...like an unlocked multiplier etc. are cool things that might have a appeal and will help boost sales.

 

[itotallybelieveyou]

I demand unlocked multipliers on everything.

 

[caamsa]

Also you forgot all of the crippled quad cores with only three cores working......where are those? AMD said they would also sell tri-cores.

 

[cnumartyr]

Also you forgot all of the crippled quad cores with only three cores working......where are those? AMD said they would also sell tri-cores.

My question is where the heck are they going to be priced?

It'll be like.. 20 bucks more for a tri core vs dual core and another 20 bucks for a quad core.

I think AMD could gain some force in the OEM segment with a good HD refresh and the B3 stepping of the Phenom.

 

[caamsa]

My question is where the heck are they going to be priced?

It'll be like.. 20 bucks more for a tri core vs dual core and another 20 bucks for a quad core.

I think AMD could gain some force in the OEM segment with a good HD refresh and the B3 stepping of the Phenom.

Yea and who knows how they will perform. That could make for a big mess in the market....I checked for Phenoms and they are all sold out on the sites that I checked.

I still think the AM2 is the way to go if you are going to buy AMD until they get out a better version of the phenom.

If you go intell the best is their Q6000 chip. If I was going to build a new system I would have built a system around that.

 

[cnumartyr]

Yea and who knows how they will perform. That could make for a big mess in the market....I checked for Phenoms and they are all sold out on the sites that I checked.

I still think the AM2 is the way to go if you are going to buy AMD until they get out a better version of the phenom.

If you go intell the best is their Q6000 chip. If I was going to build a new system I would have built a system around that.

I build my last XP rig around that.

I plan on building a Vista 64 rig in Feb/March. I'd like to see Phenom BEs because I do like the idea of it and give Intel some competition.

 

[sailer]

How can AMD beat Intel? With a stick, while Intel was sleeping.

Ok, on a more serious note, if AMD can get out some B3 revisions with unlocked multipiers at a low enough cost, then maybe AMD has a good chance in the mid-range market. No flagship leadership here, but a solid mid-range performer. After that, AMD needs to learn how to stop lying about its products. Hearing the used-car saleman pitch of how great the product is and time after time seeing results that are bad turns people off. Hear enough lies and only the die-hard types will continue to buy.

Another thing is to make sure that products are available. So Phenom is offically released. See any advertised at Newegg? For that matter, where are all the motherboards to support it? When will they arrive; two weeks, four weeks, longer? AMD needs a change in management desperately, the sooner the better. Lee Iacoca helped save Chrysler back in 1980. AMD needs someone like him now.

 

[masteryoda34]

AMD said there would be Phenoms available toward's the end of the week. And to be fair, Intel started showing off the 3.2 Penryn today also, but you can't buy it. (Not that I'd spend $1000 on a CPU)

I think AMD should unlock all their multipliers until they can get more competitive.

 

[caamsa]

How can AMD beat Intel? With a stick, while Intel was sleeping.

Ok, on a more serious note, if AMD can get out some B3 revisions with unlocked multipiers at a low enough cost, then maybe AMD has a good chance in the mid-range market. No flagship leadership here, but a solid mid-range performer. After that, AMD needs to learn how to stop lying about its products. Hearing the used-car saleman pitch of how great the product is and time after time seeing results that are bad turns people off. Hear enough lies and only the die-hard types will continue to buy.

Another thing is to make sure that products are available. So Phenom is offically released. See any advertised at Newegg? For that matter, where are all the motherboards to support it? When will they arrive; two weeks, four weeks, longer? AMD needs a change in management desperately, the sooner the better. Lee Iacoca helped save Chrysler back in 1980. AMD needs someone like him now.

Ahhh with Chrysler I do remember there being a little old 1.2 billion dollar loan from the Govt. Chrysler has been in the position of bankruptcy about 3 times. How many times has AMD been in that position?

 

[sailer]

The 1.2 billion to Chrysler was more of a guarentee on the loans that it already had than to new borrowing. Since AMD isn't quite that bad off yet, think of how much AMD could be helped by someone with equal leadership talent?

 

[caamsa]

The 1.2 billion to Chrysler was more of a guarentee on the loans that it already had than to new borrowing. Since AMD isn't quite that bad off yet, think of how much AMD could be helped by someone with equal leadership talent?

I believe that you are 100% correct about AMD needing better leadership.

 

[JuiceJones]

This still would not be competitive... Q6600s hit 3.4Ghz+ on air easily for 280 bucks, who cares about locked multiplier.

Maybe if Tolliman tri-cores are sub $150 and surpass 3ghz they can edge out the Pentium Dual-cores and E4X00 series for the entry-level. That's the only AMD option I could even possibly see worth buying, provided you don't already have an AM2 system.

 

[epsilon84]

3.4GHz from a Phenom is enough to beat Intel? You're joking right? Current G0 Q6600s are doing 3.6GHz on air, and at a higher IPC too. Then there is the 45nm chips...

Let's see,
3.4 x 0.9 = 3.06GHz C2Q @ 65nm
3.4 x 0.85 = 2.89GHz C2Q @ 45nm

Dude, that's hardly competitive, Black Edition or not.

AMD wants an Intel killer? Make a Phenom overclock to 4GHz+, with unlocked multis, at $250. And that's against a 65nm C2Q. Against 45nm, make that 4.5GHz+. No kidding.

 

[cnumartyr]

3.4GHz from a Phenom is enough to beat Intel? You're joking right? Current G0 Q6600s are doing 3.6GHz on air, and at a higher IPC too. Then there is the 45nm chips...

Let's see,
3.4 x 0.9 = 3.06GHz C2Q @ 65nm
3.4 x 0.85 = 2.89GHz C2Q @ 45nm

Dude, that's hardly competitive, Black Edition or not.

AMD wants an Intel killer? Make a Phenom overclock to 4GHz+, with unlocked multis, at $250. And that's against a 65nm C2Q. Against 45nm, make that 4.5GHz+. No kidding.

I'm assuming the errata thing actually deals the 10% performance knock that people are claiming.

Also.. do you really think an 8.5x multi Penryn will clock 4.5 GHz and break 500 MHz FSB on a quad? If you do, I hope you are right.. it'd be fun.

I'm not talking about an Intel killing, I'm talking about a mid-range alternative that can OC well.

 

[NMDante]

I'm assuming the errata thing actually deals the 10% performance knock that people are claiming.

Also.. do you really think an 8.5x multi Penryn will clock 4.5 GHz and break 500 MHz FSB on a quad? If you do, I hope you are right.. it'd be fun.

I'm not talking about an Intel killing, I'm talking about a mid-range alternative that can OC well.

From my understanding, the errata has nothing to do with performance gains, but with the system freezing at certain times.

The bug causes the system to freeze when a certain combination of instructions coincides with extraordinarily high traffic.

This bug can only be reproduced in the lab but does not occur under normal, real-world conditions. It is still present in the 2.20 GHz and 2.30 GHz versions of the Phenom (9500 and 9600).

So, I don't see how this errata will add 10% more performance, when it only happens when a "certain combination of instructions coincides with extraordinarily high traffic".

 

[cnumartyr]

From my understanding, the errata has nothing to do with performance gains, but with the system freezing at certain times.

The bug causes the system to freeze when a certain combination of instructions coincides with extraordinarily high traffic.

This bug can only be reproduced in the lab but does not occur under normal, real-world conditions. It is still present in the 2.20 GHz and 2.30 GHz versions of the Phenom (9500 and 9600).

So, I don't see how this errata will add 10% more performance, when it only happens when a "certain combination of instructions coincides with extraordinarily high traffic".

I'm trying to give them the benefit of the doubt one last time. I'd really like to see AMD compete again and I just don't think the B2 Phenoms are the answer. Now I'm just hoping ATI keeps them afloat.

I guess I'm trying to make some excuses for them because I don't want to see the market dominated by Intel, even though that's where we're headed.

 

[homerdog]

At this point I really don't see how AMD can come back. By the time they get up to 3GHz with their B3 steppings Penryn will be in full swing. Also, Intel is clocking their 65nm processors very conservatively; they don't really even need 45nm to beat Phenom. It's just another nail in the coffin. And then Nehalem

*this message brought to you by my S939 Athlon X2 3800+*

 

[cnumartyr]

Here's what I was looking for.

http://www.heise.de/newsticker/meldung/99177

10%

 

[epsilon84]

I'm assuming the errata thing actually deals the 10% performance knock that people are claiming.

Also.. do you really think an 8.5x multi Penryn will clock 4.5 GHz and break 500 MHz FSB on a quad? If you do, I hope you are right.. it'd be fun.

I'm not talking about an Intel killing, I'm talking about a mid-range alternative that can OC well.

Do you even know what the errata does? It causes random instability under high workloads at 2.4GHz or higher, that in itself does not make Phenom lose 10% IPC. There is a workaround for this errata, that disables the L3 cache altogether (I think), and this incurs a 10% performance hit. In practical terms, it's hardly a fix since a 2.3GHz without the 'fix' will be much faster than a 2.4GHz with the 'fix' enabled.

No, I don't expect the lower end Yorkfields to break 500FSB. Having said that, I do expect current P35 boards to do 450FSB minimum, around 470 - 480FSB on 'good' mobos, with X38/X48 hopefully getting close to 500FSB with more mature BIOSes (they do about 480FSB currently).

Oh, and it's an 8x multi for the Q9450, not 8.5x. Yes, I'm not too happy about it too, since it will most likely be FSB limited, not architecturally. However, it's the only 45nm I can afford, at ~$320 it's already $50 more than a Q6600.

So realistically, we are looking at around 3.6GHz - 4.0GHz from a low end Q9450, which is equivalent to a 4.14 - 4.6GHz Phenom. Which is why I said if AMD is serious about beating Penryn in overclocking, they had better find a stepping that does 4.5GHz+ on air.

 

[cnumartyr]

I was refering to the article that according to AMD says they take a 10% performance hit from it. However if the case is as you stated it, then no I'd say there isn't hope there. I understand what a bug is, but I was under the impression (again from the article and others) that the cause of it caused the 10% performance hit, not the disable of the L3 cache as a work around.

However for the $500ish Intel Q9550 will have the 8.5x Multiplier. I don't think we will see 500 FSB from the Penryn, but I may be entirely wrong.

I do however STILL believe this is the best way for AMD to catch up and try to get some of the market back.

Btw do you have a link to some Penryn quads doing 480 MHz Bus on the X38?