Cortex A-15 v Bulldozer/Piledriver

[french toast]

They are Core Duo, not Core 2 Duo. There's quite a bit of IPC difference there. I'd also expect BD to have higher IPC than C2D. I haven't looked into it in detail but I'd guess somewhere between Penryn and Nehalem?

No im afraid as much as i love AMD BD has TERRIBLE IPC as is probably worse than even Penryn. that is why i used the comparison i did, i would not have been silly enough to try to compare Cortex A15 to Intels top chip going forward (Ivy bridge)

 

[swaaye]

AMD needs to fix some of their architectural oddities and get their 32nm process sorted out. This is assuming they are still focused on high-end x86 over there... Who knows what's going on with all of the management changes and "strategic refocusing" nonsense.

 

[pjbliverpool]

No im afraid as much as i love AMD BD has TERRIBLE IPC as is probably worse than even Penryn. that is why i used the comparison i did, i would not have been silly enough to try to compare Cortex A15 to Intels top chip going forward (Ivy bridge)

Fair enough, it seems I overestimated BD's IPC. Still, those benchmarks are Core Duo, not Core 2 Duo so you're probably still going to see ~15-20% improvement in IPC for BD over them (per core/module).

 

[french toast]

Fair enough, it seems I overestimated BD's IPC. Still, those benchmarks are Core Duo, not Core 2 Duo so you're probably still going to see ~15-20% improvement in IPC for BD over them (per core/module).

Fair enough, but that puts Cortex A15 at arounf 20-30% lower than BD...which is probably what i thought...where as i bet many including your self without thinking about it thought BD would be 2-3x the performance.

That was the purpose of this thread...would then a A15 be a better bet in a ultrabook type form than BD..takng into consideration the power savings of ARM which would be what? 3-4x over BD?.. i think 4x A15 @2.5ghz would consume about 6w underload...what do you make of BD power consumption?

 

[tunafish]

Fair enough, but that puts Cortex A15 at arounf 20-30% lower than BD.

No, it doesn't.

In realistic loads, I doubt A15 can get half the IPC of BD. (Almost all your benchmarks are heavily optimized throughput loads.)

 

[french toast]

They are Core Duo, not Core 2 Duo. There's quite a bit of IPC difference there. I'd also expect BD to have higher IPC than C2D. I haven't looked into it in detail but I'd guess somewhere between Penryn and Nehalem?

Yes they are Core duo my mistake..but that is still a 31w 65nm part clocked at 1.83ghz duel core with 2mb cache:smile: far far higher specification that the ARM chips being used to compare against it which are optimised for smartphone form factors..as opposed to laptops with the core duo.

If you look at this benchmark..you can see an AMD FX-8150 @3.6GHZ with turbo up to 3.9/4.2ghz and 8mb L2 8mb L3 8 thread.

VS.

Intel C2D E8500 @3.16GHZ with 6mb L2 cache NO L3 2 thread.

And they both get the SAME score in SYSmark 2007.... .201.
(what ever that means)
http://www.anandtech.com/bench/CPU/2

I think many are giving the IPC of bulldozer a little too much credit, for 2012 its woefull.

 

[swaaye]

What do you mean people are giving BD too much credit? It's been panned pretty extensively. AFAIK it's only selling to fanboys, crazy overclockers and to clueless big box consumers. The server version is apparently slightly appealing though for what servers get used for.

 

[hoho]

Bang per buck wise they aren't too bad but that's mostly because their high-end is only competing with Intel's relatively expensive midrange (for us here, not general public) stuff.

 

[french toast]

I mean in the context of this thread, i really don't think a Cortex A15 will be that far behind i really don't.
Seriously Core 2 duo offers comparitive IPC to BD..and thats being more than reasonable.

The general misconception (or the right one? as its not been proven yet) is that BD class would destroy ARM's best in anything other than Battery life.

If i had put that question on these forums before BD launched i would have been laughed at...it doesn't seem all that absurd now though does it?
..Which draws me to the conclusion that in an ultrabook form factor ARM beats out AMD's best...excluding graphics of course.
..Now who would have bet on that when Llano released....

 

[itsmydamnation]

What do you mean people are giving BD too much credit? It's been panned pretty extensively. AFAIK it's only selling to fanboys, crazy overclockers and to clueless big box consumers. The server version is apparently slightly appealing though for what servers get used for.

for anyone who needs lots of cores its awesome, for people like me SMT doesn't cut it because what i run on a core (qemu running some vendors (cisco/juniper/etc) hardware appliance virtually) takes 100% of that cores CPU time regardless of what it is doing. Trying to use apps to limit its CPU time just breaks other stuff.

So for a couple hundred a can build a AMD box that i would need to spend a hell of a lot more money to get a similar level of "performance*" out of an intel box.

*i understand im a tiny segment of the market unlike people like rpg.314 who make broad sweeping statements and then when getting questioned on it then move the goal posts to "for my needs" /runs!!!!

 

[swaaye]

..Which draws me to the conclusion that in an ultrabook form factor ARM beats out AMD's best...excluding graphics of course.
..Now who would have bet on that when Llano released....

Are you saying this because Llano isn't in the ultrabook form factor? Brazos could do a better ultrabook than the current ARM options. It makes more sense for Brazos to be used in small devices like that than in the Walmart specials it is used in.

I had hoped somebody would use Llano for a 13" / ultrabook design, but alas it was stuck in cheap budget books that were too big. I'm hoping Trinity gets better use.

 

[Tchock]

OP seems to be going round and round without actually realizing that once you transform from mobile to notebook, flexibility, legacy code and sheer performance beats the crap out of sheer BOM and perf/W.

And (un)fortunately, once you get into ultrabook territory, the display uses enough power (and the battery gets to be big enough) to make perf/w on the SOC a much lesser issue.


That, and assuming that A15's going to scale really gracefully in IPC on a fixed set of workloads while using the competitor's design meant to clock high in the first place = not sure if serious or...

 

[french toast]

OP seems to be going round and round without actually realizing that once you transform from mobile to notebook, flexibility, legacy code and sheer performance beats the crap out of sheer BOM and perf/W.

And (un)fortunately, once you get into ultrabook territory, the display uses enough power (and the battery gets to be big enough) to make perf/w on the SOC a much lesser issue.


That, and assuming that A15's going to scale really gracefully in IPC on a fixed set of workloads while using the competitor's design meant to clock high in the first place = not sure if serious or...

You seem to be missing the point, leaving software 'legacy' code and the measurements of the screen power draw to one side for a moment.
The discussion is about the actual performance and power consumption of the said CPU's.

Of course the powerdraw is not trivial, if you have 4 bd cores that maxx out @ say 18w (for example i have no idea of the actual figures) and the A15's the same at 6w...you don't have to be a genius to work out thats a whole lot (12W) of either extra power budget for graphics..extra battery life...smaller battery..or a mixture of all three.

The Question is..''could you make a better ultrabook using Cortex A15's rather than Bulldozer?''...

 

[french toast]

Are you saying this because Llano isn't in the ultrabook form factor? Brazos could do a better ultrabook than the current ARM options. It makes more sense for Brazos to be used in small devices like that than in the Walmart specials it is used in.

I had hoped somebody would use Llano for a 13" / ultrabook design, but alas it was stuck in cheap budget books that were too big. I'm hoping Trinity gets better use.

Perhaps your right, But that is for a seperate thread.
As Trinity comes in a 17w 'ultrabook' state..and Bulldozer is AMD's top line, highest performance CPU going forward, i thought it would be interesting to compare the two high end parts form AMD and ARM for that form factor, just to see how far ARM have come along peformance wise, and/or whether AMD (and therefor x86) have got some competition right where many wouldn't have thought possible a couple of years ago.

 

[swaaye]

I think it makes more sense to put ARM against the x86 players in the netbook space instead. And of course Intel and AMD are trying to be involved in tablets and those tablets are not so great right now due to battery life and heat issues (fans!). The 22nm Atom will be interesting though......

 

[denev2004]

Well that is a very dramatic point, BUT i have read somewhere that those type of tests are more heavilly optimised for x86..and also are done across different operating systems... so is that a fair comparison??.

For the record i expect bulldozer to win in outright performance..but obviously A15 to win by far on power consumption...the thing im trying to get to is - will ARM's best chip be a better solution for anything up to a ultrabook than AMD's and maybe Intels all things considered. (including batterylife).

I think it could be quite close, alot closer than people think...and that may mark the first time ARM can encroach into x86 Territory when all sides are using their best troops...

Although it is not really "fair", its a useful reference.
Other factors that might affect performance such has OS and Memory is not that important compared to Processor itself.
Not by saying actually iOS and Android actually use less resources than Win7 does.

But there is one thing to be considered, ARM said A15 will have significant performance improvement. And it seems it will by looking at its architecture. 3-issue shall be powerful

 

[french toast]

I think it makes more sense to put ARM against the x86 players in the netbook space instead. And of course Intel and AMD are trying to be involved in tablets and those tablets are not so great right now due to battery life and heat issues (fans!). The 22nm Atom will be interesting though......

Yea i agree with what you are saying, maybe ill start another thread after this one for that very subject..but this thread serves a specific purpose which i outlined above.

 

[Tahir2]

Right french toast...

So specifically Bulldozer/Piledriver vs ARM A15 - (forget for a moment there will be multiple iterations of the A15 architecture built by various companies).

Answer:
Bulldozer/Piledriver Bulldozer/Piledriver would rip the ARM A15 a new pipeline. Totally.

Further clarification needed?

 

[french toast]

Right french toast...

So specifically Bulldozer/Piledriver vs ARM A15 - (forget for a moment there will be multiple iterations of the A15 architecture built by various companies).

Answer:
Bulldozer/Piledriver Bulldozer/Piledriver would rip the ARM A15 a new pipeline. Totally.

Further clarification needed?

Lol, blunt..but i like it!

Maybe, but at the same time would the bulldozer/piledriver burn you a new battery?

All things considered Tahir2..all things considered.

 

[tunafish]

BUT i have read somewhere that those type of tests are more heavilly optimised for x86.

I would actually contest that. I think it's more that x86 has been (much) more heavily optimized for that kinds of loads. Either way, it's irrelevant -- what's relevant is that x86 + the software stack that will be available to it will be much faster at desktop type loads than ARM + the software stack it has will be.

The issue isn't that people overestimate BD ipc, it's that you vastly overestimate what ARM will get. As I said earlier, the next ARM will probably reach or exceed x86 ipc on pure scalar numbercrunching. It's just that that isn't a realistic load for a desktop/laptop pc. For tasks that the desktop user cares about, like spell-checking and grammar checking text while layouting it, what matters is rapidly traversing complex and large data structures. And ARM really really falls behind there. Even if they did awesome evolutionary improvements over their last generation, they are not going to get to half of BD ipc when running at 2.5GHz.

And I'm not an ARM hater. I actually like the arch quite a bit, and the future ARMv8 will probably be the cleanest and best mass-produced isa ever. It's just that I have programmed on a lot of arm systems, and I have programmed on a lot of x86 systems, and I think your expectations are really unreasonable.