[Nintendo Maniac 64]

We've all seen the disappointing performance of Bulldozer, and how Thuban keeps up with it most of the time.

Which begs the question: why all the R&D on Bulldozer in the first place if a die-shrink of Thuban would have worked just as well if not better? And that's without incorporating the IPC improvements from Llano to make a theoretical Phenom III or something...

It just doesn't make any sense!

 

[Dmac73]

Quote:

Originally Posted by Nintendo Maniac 64;15282195
We've all seen the disappointing performance of Bulldozer, and how Thuban keeps up with it most of the time.

Which begs the question: why all the R&D on Bulldozer in the first place if a die-shrink of Thuban would have worked just as well if not better? And that's without incorporating the IPC improvements from Llano...

It just doesn't make any sense!

Moar cores/higher clockspeeds = Better marketing in server and the extreme high percentage of unknowing client and desktop users.

Also, you couldn't add more cores AND clockspeed to Phenom 2, among other issues with that. Straight from *chew.

Besides, BD was years into development by then, and the "simulations" they had at the time depicting BD performance were OBVIOUSLY way off. Not even AMD knew actual performance till the very first set of samples which im assuming were late 09 or early '10.

 

[noak]

Last time I checked, thubans were 45nm

Llano were the only 32nm chips they have made so far.

I don't think you understand how this stuff works. You don't just decide "Oh, lets make a 32nm chip today.

The factories that make these chips at different NM processes are made 5+ years in advance.

You just can't slap extra cores to a CPU and call it game.

You need to improve the architecture of the cores and CPU.

Sadly, Intel is the only CPU maker actually pushing the limits

 

[wedge]

Here's the thing to remember about Bulldozer: It's the first product from a whole new architecture. It's just a baby, and it's going to grow up.

Deneb/Thuban processors are the last products from an architecture more than 10 years old. People are asking why AMD didn't just die shrink and add on 2 more cores... Don't you think they thought of that? It wasn't going to work. If they could snap their fingers and make Phenom II's faster, they would have done it by now. Those processor's have hit a wall, and they can't go any faster.

Bulldozer is picking up where Phenom II left off. It obviously has way more performance potential than is currently seen (just look at the transistor count). If they can follow their roadmap for the next few years, then good things will come from this.

 

[Nintendo Maniac 64]

Quote:

Originally Posted by wedge;15282439
Deneb/Thuban processors are the last products from an architecture more than 10 years old.

By that logic, Sandy Bridge is over a decade old.
Quote:

People are asking why AMD didn't just die shrink and add on 2 more cores... Don't you think they thought of that? It wasn't going to work. If they could snap their fingers and make Phenom II's faster, they would have done it by now. Those processor's have hit a wall, and they can't go any faster.

Bulldozer is picking up where Phenom II left off. It obviously has way more performance potential than is currently seen (just look at the transistor count). If they can follow their roadmap for the next few years, then good things will come from this.

Here's the problem with that...
Quote:

People are asking why Intel didn't just die shrink and make it dual-core... Don't you think they thought of that? It wasn't going to work. If they could snap their fingers and make Pentium III's faster, they would have done it by now. Those processor's have hit a wall, and they can't go any faster.

Netburst is picking up where Pentium III left off. It obviously has way more performance potential than is currently seen (just look at the transistor count). If they can follow their roadmap for the next few years, then good things will come from this.

And remember, Intel had a WAY higher R&D budget, and yet they still couldn't make it work long-term...

Sometimes a whole new architecture isn't the best way to go, and instead it's better to continuously refine what works, AKA the Pentium 3 to Pentium M to Core Duo and beyond.

You don't always need a revolution if you can just keep evolving.

Quote:

Originally Posted by Dmac73;15282236
Also, you couldn't add more cores AND clockspeed to Phenom 2, among other issues with that. Straight from *chew.

Quote:

Originally Posted by noak;15282260
You just can't slap extra cores to a CPU and call it game.

I think you guys need to check Thuban's current IPC against Bulldozer again, not to mention the IPC improvements used in Llano. Who said anything about clockspeed?

 

[Stefy]

Quote:

Originally Posted by wedge;15282439
Here's the thing to remember about Bulldozer: It's the first product from a whole new architecture. It's just a baby, and it's going to grow up.

Deneb/Thuban processors are the last products from an architecture more than 10 years old. People are asking why AMD didn't just die shrink and add on 2 more cores... Don't you think they thought of that? It wasn't going to work. If they could snap their fingers and make Phenom II's faster, they would have done it by now. Those processor's have hit a wall, and they can't go any faster.

Bulldozer is picking up where Phenom II left off. It obviously has way more performance potential than is currently seen (just look at the transistor count). If they can follow their roadmap for the next few years, then good things will come from this.

Pointless statement as there's absolutely nothing supporting it.

Bulldozer failed, and will continue to fail until the opposite is proven.