What is holding AMD back from increasing instructions per clock?

[thestain]

I have read that no one in their right mind will buy a Phenom 9500 or 9600 due to problems and that fix will be in for Phenom 9900 when it comes out, but..

Why can't AMD increase the instructions per clock from old 3 to something similar to what Intel is getting in the 4 plus range?

Wouldn't this make AMD more competitive performance-wise?

and.. why the ego stroking name.. nothing phenomenal about Phenom yet..

Pride before the fall.. final fall of AMD?

 

[jones377]

It was probably cheaper to make K10 based on K8 rather than making a completely new 4-issue microarchitecture. Nevertheless, it's just one factor why Phenom has a lower IPC than Core, one out of several. The poor caches are probably the biggest reason. Too small, too slow both in bandwidth and latency..

 

[myocardia]

Originally posted by: thestain
Why can't AMD increase the instructions per clock from old 3 to something similar to what Intel is getting in the 4 plus range?

Because instead of spending all of that money they made during the Athlon 64's reign on designing a faster processor, they decided to make Hector Ruiz one of the highest paid CEO's on the planet, and let him buy out a video card company for ~3 times what it was worth. Those two mistakes took all of their R&D money, so what they're trying to sell the public now is a very slightly improved (clock for clock, that is) quad-core Athlon 64.

Pride before the fall.. final fall of AMD?

It's definitely possible, with the way they're bleeding money. Don't hold your breath, though.

 

[boatillo]

Originally posted by: myocardia

Originally posted by: thestain
Why can't AMD increase the instructions per clock from old 3 to something similar to what Intel is getting in the 4 plus range?

Because instead of spending all of that money they made during the Athlon 64's reign on designing a faster processor, they decided to make Hector Ruiz one of the highest paid CEO's on the planet, and let him buy out a video card company for ~3 times what it was worth. Those two mistakes took all of their R&D money, so what they're trying to sell the public now is a very slightly improved (clock for clock, that is) quad-core Athlon 64.

Pride before the fall.. final fall of AMD?

It's definitely possible, with the way they're bleeding money. Don't hold your breath, though.

I like to think of this a little differently. Ok Athlon64 had AMD leading in performance benchmarks for once and they were making some OK gains in some of their markets because of their solid chips. I like to think that what AMD did was look at the long-term picture and what they could do to get there. They were making some excess profit, their stock price was looking very good, so they leveraged their current...puesdo-tie with Intel to give them an opportunity to REALLY give it to Intel somewhere down the road. Maybe they will fail with their longterm plan, maybe they will put a serious hurtin' to Intel with all this integrated graphics they are trying for. Either way, they are still just in the organizational part of this plan and fruits probably won't come for several years.

 

[Nemesis 1]

OK . I don't know a lot about amd other than in a 2 horse race that requires endurance I pick intel .

Is it true that before AMD 64 came out that AMD bought a company and used that tech in AMD 64. Also is it true that Befor ATI released the 9700pr0 that they bought a company that used that tech. It is posiable AMDs enginners just aren't up to the task. I mean it took 2 outside companies to give them the teck that allowed their moment in the sun to shine.

 

[thestain]

I think that this really shows that when you need guerrilla warfare to prosper and survive against a much larger foe, small incrmental change can and will be a disaster.

AMD needs to innovate, to change faster than larger Intel and currently they are moving... sloooowwwerrrr.

The Stain

 

[zsdersw]

Originally posted by: boatillo
I like to think that what AMD did was look at the long-term picture and what they could do to get there. They were making some excess profit, their stock price was looking very good, so they leveraged their current...puesdo-tie with Intel to give them an opportunity to REALLY give it to Intel somewhere down the road. Maybe they will fail with their longterm plan, maybe they will put a serious hurtin' to Intel with all this integrated graphics they are trying for. Either way, they are still just in the organizational part of this plan and fruits probably won't come for several years.

Let's say for the sake of discussion that your assessment of the situation is accurate; that's what AMD was/is shooting for. What all of that fails to take into account is the fact that Intel isn't sitting still either. Does or did AMD really think Intel wouldn't (or wouldn't be able to) pursue and bring to market its own GPU/CPU combo?

 

[Nemesis 1]

WRONG thread sorry

 

[myocardia]

Originally posted by: Nemesis 1
Is it true that before AMD 64 came out that AMD bought a company and used that tech in AMD 64. Also is it true that Befor ATI released the 9700pr0 that they bought a company that used that tech. It is posiable AMDs enginners just aren't up to the task. I mean it took 2 outside companies to give them the teck that allowed their moment in the sun to shine.

Only one of those is true. ATI didn't design the 9700 Pro, it was just a variation of a FireGL card, made by the company they bought, FireGL. AMD designed the Athlon 64 themselves, with some (I'm not sure how much, though) help from IBM. AMD did buy out a company, and based their original Slot A Athlon on the other company's processor, though.

 

[BitByBit]

Originally posted by: thestain
Why can't AMD increase the instructions per clock from old 3 to something similar to what Intel is getting in the 4 plus range?

In reality, neither AMD nor Intel get anything close to 4 IPC, or even 3. Core 2, like K8/K10, is capable of retiring 3 instructions per clock, but achieves nowhere near this in real world applications.

On paper, Core 2 and K10 look equally matched; both can execute 3 Int/Float instructions per clock, and 2 x 128-bit SSE per clock. Both feature load reordering, dedicated stacks, improved branch prediction and prefetch over previous generations. Cache bus width is the same in both, with the exception of K10's wider instruction fetch. But despite implementing all these key features that Intel instroduced with Core 2, K10 hasn't benefitted nearly as much. It could be Intel's instruction fusion abilities, or K10's narrower store bus giving Core 2 the edge, but until we get explanations from the engineers that designed it, we can only speculate.

 

[rogue1979]

Originally posted by: myocardia

Originally posted by: Nemesis 1
Is it true that before AMD 64 came out that AMD bought a company and used that tech in AMD 64. Also is it true that Befor ATI released the 9700pr0 that they bought a company that used that tech. It is posiable AMDs enginners just aren't up to the task. I mean it took 2 outside companies to give them the teck that allowed their moment in the sun to shine.

Only one of those is true. ATI didn't design the 9700 Pro, it was just a variation of a FireGL card, made by the company they bought, FireGL. AMD designed the Athlon 64 themselves, with some (I'm not sure how much, though) help from IBM. AMD did buy out a company, and based their original Slot A Athlon on the other company's processor, though.

ATI had FireGL during the Radeon 8500 era, before 9700 was even a rumor.

The on die memory controller was a great innovation that put AMD on top for a good while.

AMD's engineers are just fine, however their leadership and financial situation may not be the best.