Should AMD have focused on a Steamroller high TDP product line instead of Vishera?

[TerryMathews]

This is just AMDs version of the P4 Emergency Edition.

 

[Erenhardt]

this feeble attempt to compete against a juggernaut in full rampage mode needs to stop for the sake of AMD's shareholders and AMD's future.

What? They are not competing with ARM. They know they have 0 chances with them.
So they targeted lazy intel which is running backwards with its latest hassfail release.

 

[Torn Mind]

Interesting poll results, guess these 220W Visheras and Intel's Pentium 4 EEs are a decent stopgap while being the under performer of the x86 duopoly.

Rather than have them be a standard Vishera FX-8350 that nets them perhaps $170 in revenue, they can charge perhaps $300 for these chips and pocket $250 in revenue. These chips are already coming out of the fab, so they might as well try to make as much money off of them as possible.

 

[NTMBK]

I actually wish that they had brought out a Richland version like this, not the FX. 4 cores cranking at 5GHz would give pretty solid performance, and if the IGP was disabled they could keep the TDP at least reasonable. (130W?) Not to mention it has a much more modern motherboard platform.

 

[Idontcare]

This is just AMDs version of the P4 Emergency Edition.

That was exactly what I thought too.

AMD's experiment with Netburst continues to come full-circle.

It won't finish coming full-circle until the day AMD scraps netburst and puts into market a version of their cat line (Jaguar at the moment) that is lower-clocked but higher IPC and completely puts to shame the entire 'dozer lineage in terms of absolute performance as well as performance/watt.

Until then it will continue to be history repeating itself, complete with the fanboys who can't bend over backwards far enough to argue how awesome everything about Intel's Netburst AMD's 'dozer CMT must be

 

[Torn Mind]

What? They are not competing with ARM. They know they have 0 chances with them.
So they targeted lazy intel which is running backwards with its latest hassfail release.

I doubt Intel is their primary concern these days, but rather their mere survival and thus accumulating as much cash as possible. Since there will always be buyers who spends a $100 more for just a few more Mhz, this is AMD way to reel in the brand loyalists and e-peen strokers that are still around and open to buying an AMD product.

 

[TerryMathews]

That was exactly what I thought too.

AMD's experiment with Netburst continues to come full-circle.

It won't finish coming full-circle until the day AMD scraps netburst and puts into market a version of their cat line (Jaguar at the moment) that is lower-clocked but higher IPC and completely puts to shame the entire 'dozer lineage in terms of absolute performance as well as performance/watt.

Until then it will continue to be history repeating itself, complete with the fanboys who can't bend over backwards far enough to argue how awesome everything about Intel's Netburst AMD's 'dozer CMT must be

Well, at the time I ran some NetBurst P4s for the same reason I'm running a FX now. There are some good values available when you overclock.

At the time I ran a 2.4c at 3.2. Later I ran a pair of 1.6 Xeons at 3.2. Unbeatable performance for the price.

Now I've got a FX4100 @ 4.4GHz. I paid $60 for it.

This line for AMD is a failure. That doesn't mean there aren't good buys within the line.

 

[JimmiG]

I think the current FX 9370 and 9590 are actually a bad idea and if they were determined to create a ~200W TDP product line it should be Steamroller based. Kaveri is supposedly on track for release this year, I think a manufacturer certified high clocked Kaveri would have been the way to introduce this marketing idea. But I'm not sure it's a good marketing idea in general, too much like the Pentium 4 EE.

I don't think it had any negative effects on Steamroller. My understanding is that AMD found that a small portion of their chips were capable of 5 GHz, so instead of selling them as 4 GHz chips, they figured they could sell them as 5 GHz chips and make a bit of a profit and also get a bit of PR by having the world's first 5 GHz chip.

 

[Idontcare]

...and also get a bit of PR by having the world's first 5 GHz chip.

IBM has been shipping 5.2GHz chips since 2010.

 

[ShintaiDK]

I don't think it had any negative effects on Steamroller. My understanding is that AMD found that a small portion of their chips were capable of 5 GHz, so instead of selling them as 4 GHz chips, they figured they could sell them as 5 GHz chips and make a bit of a profit and also get a bit of PR by having the world's first 5 GHz chip.

Its up to 5Ghz. It would be fraud to market them as 5Ghz since they are 4.7Ghz chips. Just like my 4670 is not a 3.8Ghz chip.

 

[JimmiG]

IBM has been shipping 5.2GHz chips since 2010.

World's first x86 5 GHz chip.

Its up to 5Ghz. It would be fraud to market them as 5Ghz since they are 4.7Ghz chips. Just like my 4670 is not a 3.8Ghz chip.

They could easily sneak in the "up to" in fine print somewhere people wouldn't notice it.

 

[Erenhardt]

5.0 GHz in turbo is guaranteed, so...

 

[ShintaiDK]

They could easily sneak in the "up to" in fine print somewhere people wouldn't notice it.

That wouldnt be illegal in some countries. It would have to be quite clear if you advertise with something that is not obtainable in all situations.

ISPs already got punished quite hard in the same cases.

 

[ShadowVVL]

I can only see $250 tops for fx 9000 if power consumption is through the roof. If it is over $250 haswell i5 will be a better buy.

 

[galego]

IBM has been shipping 5.2GHz chips since 2010.

None of them was x86/x64... and none was a 8-cores chip. AMD has just broken another world-record.

Its up to 5Ghz. It would be fraud to market them as 5Ghz since they are 4.7Ghz chips.

The point is that are not 4.7GHz because the clock is variable and a 220W TDP rating would allow more sustained turbo than the series 8000. Also it is ridiculous to talk about fraud by a mere 0.3 GHz.

Some examples of fraud

http://www.undeadly.org/cgi?action=article&sid=20060930232710&mode=expanded

http://ethics.csc.ncsu.edu/risks/reliability/pentium/study.php

http://sharikou.blogspot.com.es/2006/05/intel-woodcrest-performance-claim.html

If you find something similar regarding the FX-9000 series let us know. Thanks.

 

[ShintaiDK]

The point is that are not 4.7GHz because the clock is variable and a 220W TDP rating would allow more sustained turbo than the series 8000. Also it is ridiculous to talk about fraud by a mere 0.3 GHz.

That principle was settled long ago. And AMD is not gonna be any exception. Its both scamming the consumer and breaking the marketing act.

Nice touch tho that you are willing to allow AMD to cheat people for ~6%. In my country tho, that will be a case by the consumer protection handled to the police for further charges.

 

[Vesku]

"Runs at up to 5GHz out of the box" - should be a useable advertising line internationally.

 

[sushiwarrior]

I think it's ridiculous to suggest AMD should instead work on a core which isn't even out yet. What do they have available? Vishera. So they did what they could with Vishera. Steamroller won't be ready for another few months, and even then they'll need proper yields to get a high TDP chip.

 

[Torn Mind]

"Runs at up to 5GHz out of the box" - should be a useable advertising line internationally.

Yeah, many stores say "Up to so-and-so percent off" when they have a promo running and yet many of the discounts are nowhere near the maximum.

 

[Vesku]

I think it's ridiculous to suggest AMD should instead work on a core which isn't even out yet. What do they have available? Vishera. So they did what they could with Vishera. Steamroller won't be ready for another few months, and even then they'll need proper yields to get a high TDP chip.

If they are having yield issues after delaying 28nm, a half node, for a full year (that's not counting delay bumps from 32nm hiccups)... well AMD is going to have a baaaaad time.

 

[Ancalagon44]

That was exactly what I thought too.

AMD's experiment with Netburst continues to come full-circle.

It won't finish coming full-circle until the day AMD scraps netburst and puts into market a version of their cat line (Jaguar at the moment) that is lower-clocked but higher IPC and completely puts to shame the entire 'dozer lineage in terms of absolute performance as well as performance/watt.

Until then it will continue to be history repeating itself, complete with the fanboys who can't bend over backwards far enough to argue how awesome everything about Intel's Netburst AMD's 'dozer CMT must be

Interesting thought - Jaguar is already faster than K8, clock for clock, despite not being as wide an architecture.

Jaguar with a 4 wide execution path instead of 2?

 

[Idontcare]

None of them was x86/x64... and none was a 8-cores chip. AMD has just broken another world-record.

Is that how AMD is advertising them?

Is AMD advertising this as the "world's first x86/x64 5GHz 8-core chip?"

...or are they being intentionally misleading by not specifying under what limited caveats their 5GHz is a world's first?

 

[Arachnotronic]

What? They are not competing with ARM. They know they have 0 chances with them.
So they targeted lazy intel which is running backwards with its latest hassfail release.

I'm sorry, but do you even read what you type? Haswell is anything but lazy.

 

[sushiwarrior]

If they are having yield issues after delaying 28nm, a half node, for a full year (that's not counting delay bumps from 32nm hiccups)... well AMD is going to have a baaaaad time.

By yield I mean it took Vishera a while to get a sufficient supply of chips binned to hit 4.7 at low volts. It is extremely uncommon to see a brand new chip on a brand new process (and we all know how lovely GloF is) launch with a "high binned" model right off the bat. It takes a while for the process to mature and for enough cream of the crop to be made.

 

[thepath]

Good idea,

5GHz, 8 cores, will be faster than Core i7 Haswell (Default) in a lot of apps, it will look good in the graphs if you know what i mean

1- It is not 5GHz CPU. It is 4.7GHz base clock with 5GHz max turbo clock

2- just keep in mind that i7 4770 beats FX-8350 in heavy multi-threaded benchmarks. And FX-9590 has only 17.5% higher clock speed when comparing the base clock speed. So, FX-9590 is going to be just slightly faster than i7 4770K in heavy multi-threaded benchmarks (maybe just ~ 10% faster). i7 4770 is still going to rape it in single and lightly threaded benchmarks.