Official Improvements of Piledriver Cores.

[Xpage]

a new bobcat core, not that much improved it seems. Maybe it can play netflix if there is a good IPC increase, but the clockspeed with be only 50more mhz from what i read.


a 17w trinity, seems like an excellent HTPC CPU, however, when will those parts come out, in what quantity and for how much remains to be seen.

 

[Ajay]

actually, no...BD have 4 issue scheduller, but it need 8....
so each core only uses 2 issue...:\

ignore the half-clocked....my mistake:\

Bummer, because according to the OP images, AMD didn't increase the number of x86 decoders (still @ 4)

 

[ShadowVVL]

Is the 17w trinity for desktop or mobile?

 

[ShintaiDK]

Is the 17w trinity for desktop or mobile?

Mobile. And its most likely not gonna be cheap either..or fast.

If you look at slides it should be the lowest numbers. So 2Ghz and 464Mhz GFX.

 

[pelov]

Bummer, because according to the OP images, AMD didn't increase the number of x86 decoders (still @ 4)

AMD widened the K8 front end with Bulldozer. Each module is now able to fetch and decode up to four x86 instructions from a single thread in parallel. Each of the four decoders are equally capable. Remembering that each Bulldozer module appears as two cores, the front end can only pick 4 instructions to fetch and decode from a single thread at a time

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/2

Comparison b/w BD and Trinity

Although it's still x4, the improvements seem to be numerous.

A couple that aren't described above are the load queue going up 10%, 44 from 40, L1DTLB doubled to 64 from 32.

 

[ShintaiDK]

Looking at the issue ports for the INT core. You know why Pilediver gonna fail like Bulldozer. Not to mention the continue of the failed concept of a shared FPU.

Also its misleading at best to show 4 pipelines/issue ports for the INT core. 2 of the are FP dispatch ports.

The entire uarch of Bulldozer/Pilediver/Steamroller/etc is based on a netburst concept. Its a speedracer. But speedracers dont work. It failed for Intel, it failed for IBM and now it failed for AMD.
problem is there is no way for it to succeed. Not to mention they lack an alternative uarch to push forward. I bet AMD is racing to produce a new uarch and they will drop this uarch as soon as its engineeringly possible.

The only question is basicly, who at AMD was/is responsible for it. I bet marketing thought more gigahurts would sell better. Their APU numbering and naming aint exactly clever either. Its desperate misleading at best. And i think its the same department responsible for both.

Comparing the issue ports:

 

[ShadowVVL]

Mobile. And its most likely not gonna be cheap either..or fast. If you look at slides it should be the lowest numbers. So 2Ghz and 464Mhz GFX.

yeah i priced out an intel mobile cpu and mobo a a few weeks ago and it was close to $450.

 

[ShintaiDK]

yeah i priced out an intel mobile cpu and mobo a a few weeks ago and it was close to $450.

Why the mobile+LV/ULV CPU anyway?

For HTPC/Network storage just use desktop chip. Their idle states are around the same.

The 1W you might save on a HM77 for example vs a H77 aint worth it.

 

[Olikan]

The entire uarch of Bulldozer/Pilediver/Steamroller/etc is based on a netburst concept. Its a speedracer. But speedracers dont work. It failed for Intel, it failed for IBM and now it failed for AMD.

BD, PD or even Steamroller isn't even close to netburst concept...

....it's close to POWER 7...

Bummer, because according to the OP images, AMD didn't increase the number of x86 decoders (still @ 4)

IIRC.... steamroller will have more decoders

 

[Don Karnage]

Dear Amd,

Just copy whatever Intel is doing

 

[ShintaiDK]

BD, PD or even Steamroller isn't even close to netburst concept...

....it's close to POWER 7...


IIRC.... steamroller will have more decoders

No...

And decoders aint the problem. The uarch is simply incapable of performing compared to nehalem, sandy, haswell etc.

 

[eternalone]

So basically its gonna be a Phenom 2 with IGP??

 

[sefsefsefsef]

No...

And decoders aint the problem. The uarch is simply incapable of performing compared to nehalem, sandy, haswell etc.

So in what way is it similar to Netburst?

 

[ShintaiDK]

So basically its gonna be a Phenom 2 with IGP??

Bulldozer with IGP I assume? Llano is basicly Phenom 2 with IGP.

 

[BallaTheFeared]

More like a Athlon II that can't overclock well.

 

[ShintaiDK]

So in what way is it similar to Netburst?

Its the same speedracer concept. You bet on clockspeed to counter either a long pipeline or/and reduced core with limited issue ports to reach the desired performance. Thats also why Bulldozer/Pilediver is a 2 issue wide INT uarch with a shared FPU vs Conroe/nehalem/Sandys 4 issue wide INT/FPU.

AMD took a huge gamble and lost it. Small simple cores at high speed. They just couldnt reach the speed. I assume they had expected 5-6Ghz as stock speeds.

Another loss was the shared FPU. But that tradeoff seems to have been the first of many victims.

Netburst for example had 1 complex and 2 simple issue ports. K8 had 3 complex.

 

[ShadowVVL]

Why the mobile+LV/ULV CPU anyway? For HTPC/Network storage just use desktop chip. Their idle states are around the same. The 1W you might save on a HM77 for example vs a H77 aint worth it.

I was just checking what it would cost for mobile htpc. but just the cpu and mobo is more then a llano desktop system. if desktop triniy is any good i will get one if not i get a llano.


so is trinity coming out this month or no?

 

[dmens]

No...

And decoders aint the problem. The uarch is simply incapable of performing compared to nehalem, sandy, haswell etc.

Actually he's right, Power 7 is "speedracing" done right with a ISA that suits the design.

 

[ShintaiDK]

Actually he's right, Power 7 is "speedracing" done right with a ISA that suits the design.

Read his post again and you see the error.

 

[dmens]

Read his post again and you see the error.

There's a lot of differences between netburst and power7 and it's not just the ISA.

 

[Olikan]

Read his post again and you see the error.

my power 7 comparison is the pipeline lenght...

prescott had 31 stages
Bulldozer and Power7, around 20

 

[ShintaiDK]

my power 7 comparison is the pipeline lenght...

prescott had 31 stages
Bulldozer and Power7, around 20

And Northwood was 20 stages...

 

[nismotigerwvu]

And Northwood was 20 stages...

And Northwood was the most competitive revision of netburst and actually outperformed the Athlon XP for a short period of time, so there's your error.

 

[Olikan]

...and still diferent than netburst

EDIT:

lol, i am getting tired

 

[dmens]

Comparing BD/PD to any of the "big" cores (i.e. cores designed to win single thread perf crown) is a waste of time because it was never designed to do that to begin with, as far as I can tell.