• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Wii U CPU Benchmarks (highly unscientific)

W

wsippel

Banned
Thanks to our good friends at Audiokinetic, we just got a couple of Wii U CPU performance figures. Well, two. Still, it's better than nothing, and the first actual benchmark I know of. The only problem is that they didn't release any figures to compare them to, which is why I tried to roll my own, running the same tests (I think) on the only Windows machine I have access to. So while this test won't tell us how the Wii U CPU performs compared to Xenon or Cell, it should give us an idea how well it holds up compared to PC CPUs. Or not.

The system I used for comparison purposes is my HTPC, as the free version of Wwise is Windows only and all my other machines are running OSX or Linux. I used the same version of Wwise (2012.2) and tested the effects mentioned in the release notes, with mono samples as specified in the release notes, not touching any settings as I assume the benchmarks are for the default settings.

Test setup:
AMD A6-3500 @ 2.4GHz
4GB DDR3, 1.6GHz CL6
Wwise 2012.2 64bit
Windows 7 x64

The baseline on this system is 0.2785% as reported by the Wwise profiler, all figures are an average over 60 samples. And these are the results:

Code: 
Wwise 2012.2 CPU Load (lower is better)

               |   Wii U | A6-3500 |  i7-920
------------------------------------------------
Peak Limiter   |   0.20% |   0.08% |   0.05%
Delay          |   0.07% |   0.04% |   0.01%

It seems the Wii U CPU is about half as fast as a mid range PC CPU from 2011. No surprises here, considering the A6 alone draws at least as much power as the whole Wii U system, and is a 32nm part - not 45nm like the Wii U CPU.

If anyone wants to replicate my test, get the sample I've use here and Wwise here. The Wwise release notes with the benchmark results are available here.


EDIT: i7 performance figures by Durante.
 
Oppo

Oppo

Member
While the effort is appreciated, I'm not actually sure this is better than nothing. It's just too vague to be considered useful in practical terms, I would think. This will just fuel more speculation. (Which is ok, just... y'know.)
 
W

wsippel

Banned
Nerfgun said:
While the effort is appreciated, I'm not actually sure this is better than nothing. It's just too vague to be considered useful in practical terms, I would think. This will just fuel more speculation. (Which is ok, just... y'know.)
Click to expand...
It's actually pretty solid info, and might very well be the most precise performance info we'll ever get. ;)
 
A

AzaK

Member
Thanks wsippel.

One thing we need to find out is the actual test configuration used on Wii U so we can see if your tests are comparative. Things like Threads/Cores used, audio file format used and things like that. I imagine you don't have the Wii U test setup information by any chance?
 
W

wsippel

Banned
AzaK said:
Thanks wsippel.

One thing we need to find out is the actual test configuration used on Wii U so we can see if your tests are comparative. Things like Threads/Cores used, audio file format used and things like that. I imagine you don't have the Wii U test setup information by any chance?
Click to expand...
I used the Wwise profiler info. I think Wwise is single threaded, but the A6 has three cores just like the Wii U, so the comparison would be fair either way. The sample is 48kHz 16bit mono uncompressed, but that doesn't matter as I profiled the effect stage only, not the decoding or mixing stage.
 
1

1-D_FTW

Member
wsippel said:
Thanks to our good friends at Audiokinetic, we just got a couple of Wii U CPU performance figures. Well, two. Still, it's better than nothing, and the first actual benchmark I know of. The only problem is that they didn't release any figures to compare them to, which is why I tried to roll my own, running the same tests (I think) on the only Windows machine I have access to. So while this test won't tell us how the Wii U CPU performs compared to Xenon or Cell, it should give us an idea how well it holds up compared to PC CPUs. Or not.

The system I used for comparison purposes is my HTPC, as the free version of Wwise is Windows only and all my other machines are running OSX or Linux. I used the same version of Wwise (2012.2) and tested the effects mentioned in the release notes, with mono samples as specified in the release notes, not touching any settings as I assume the benchmarks are for the default settings.

Test setup:
AMD A6-3500 @ 2.4GHz
4GB DDR3, 1.6GHz CL6
Wwise 2012.2 64bit
Windows 7 x64

The baseline on this system is 0.2785% as reported by the Wwise profiler, all figures are an average over 60 samples. And these are the results:

Code: 
Wwise 2012.2 CPU Load (lower is better)

               |   Wii U | A6-3500 |
------------------------------------
Peak Limiter   |   0.20% |   0.08% |
Delay          |   0.07% |   0.04% |

It seems the Wii U CPU is about half as fast as a mid range PC CPU from 2011. No surprises here, considering the A6 alone draws at least as much power as the whole Wii U system, and is a 32nm part - not 45nm like the Wii U CPU.

If anyone wants to replicate my test, get the sample I've use here and Wwise here. The Wwise release notes with the benchmark results are available here.
Click to expand...

I'd hardly call the A6-3500 a good mid-range PC from 2011. Maybe a decent laptop.
 
I

Iceblade

Member
EDIT: How can you be sure this is comparative to the Wii U? Have you actually seen its guts, or are we just going off rumoured specs?
 
W

wsippel

Banned
Iceblade said:
EDIT: How can you be sure this is comparative to the Wii U? Have you actually seen its guts, or are we just going off rumoured specs?
Click to expand...
Huh? The Wii U numbers are official performance figures released by a licensed middleware provider.
 
S

specialguy

Banned
Iceblade said:
EDIT: How can you be sure this is comparative to the Wii U? Have you actually seen its guts, or are we just going off rumoured specs?
Click to expand...

Of course not. He got the benchmark info from the manufacturers documentation in some way.

I'm inclined to dismiss the benches as overrating the CPU. However A6-3500 is 3 Llano cores @2.1 ghz (2.4 turbo). Wii U CPU is 57% as fast in one bench 40% as fast in the other. Average ~50%.

So maybe Wii U basically features the equivalent of a 1.0-1.2 ghz-ish tri core phenom.

I still suggest that's vastly overrating it. But maybe, the raw clock deficit is enough.
 
W

wsippel

Banned
specialguy said:
Of course not. He got the benchmark info from the manufacturers documentation in some way.

I'm inclined to dismiss the benches as overrating the CPU. However A6-3500 is 3 Llano cores @2.1 ghz (2.4 turbo).

So maybe Wii U basically features the equivalent of a 1.0-1.2 ghz-ish tri core phenom.

I still suggest that's vastly overrating it. But maybe, the raw clock deficit is enough.
Click to expand...
The A6 I benchmarked is overclocked - as stated in the OP.
 
S

specialguy

Banned
So that kicks Wii U CPU up to ~1.2 ghz equivalent (2.4 ghz?)? No big. Does ocing push the turbo up as well?

3 k10 cores at 1.2ghz would indeed be fairly puny. Still, it would run say a laptop fine, and there's no way I expect Wii U anywhere close really. 3 K10 would also be a lot of transistors which, I am sure Wii U CPU is not.

I cant find rough tri core llano CPU-only transistor count (given it has an integrated GPU...). No telling really. Phenom II X3 is 758m (but I think thats counting 4 cores even though one is deactivated) but again no telling exactly what 3 llano cores is at (or at least, I dont feel like trying to find out right now), except probably a lot more than Wii U CPU.
 
O

onQ123

Member
So where does this fit with the rumor that I heard about the Wii U CPU having about the same power as the Vita CPU?
 
L

LeleSocho

Banned
Weaker than a not so good mobile CPU.... that's some weak shit we have here.

Maybe the best thing is to compare it to a PowerMac G5 to see how it roughly compare to a Xenon
 
P

Plinko

Wildcard berths that can't beat teams without a winning record should have homefield advantage
Wasn't the rumor that the CPU was weak and the GPU was relatively more complex in comparison?
 
H

Hot Coldman

Banned
A Human Becoming said:
I'll ask that the question on everyone's mind: how does it compare to the Gamecube?
Click to expand...

Here's the prototype:

8o149.jpg
 
W

wsippel

Banned
LeleSocho said:
Yep you are right, sorry but i'm not used to think an APU as a desktop CPU so my mind immediately thought of a mobile one.
Click to expand...
As I wrote, I use that particular chip in my HTPC - which is also my "Steambox". As cheap and low power as it might be, it easily runs circles around PS3 and 360.


specialguy said:
long shot but, do you have any data for x360/ps3 benches on this prog wsippel?
Click to expand...
Nope. I'd be very interested in those figures as well.


onQ123 said:
is it really?
Click to expand...
It is. Really. The Wii U CPU might be weak, but it's still an order of magnitude faster. And by "order of magnitude", I don't mean "a few" - more like "a few dozen".
 
P

Plinko

Wildcard berths that can't beat teams without a winning record should have homefield advantage
onQ123 said:
So where does this fit with the rumor that I heard about the Wii U CPU having about the same power as the Vita CPU?
Click to expand...

LOL...where did you read that "rumor?"
 
C

chertipros

Member
onQ123 said:
So where does this fit with the rumor that I heard about the Wii U CPU having about the same power as the Vita CPU?
Click to expand...

Lol what? That was a thing? I wouldn't be surprised if it was though
EDIT: yeah i would. Isn't the psvita cpu clocked low as hell making it pretty subpar?
 
C

Caelus

Member
If the Wii U's CPU is on par with the Vita, it's already been said, how come HD multiplatform games are running fine on the Wii U CPU, and the Vita is supposed to be less clocked than the HD twins?
 
D

Durante

Member
wsippel said:
It is. Really. The Wii U CPU might be weak, but it's still an order of magnitude faster. And by "order of magnitude", I don't mean "a few" - more like "a few dozen".
Click to expand...
A few dozen what? Orders of magnitude? That's ridiculous. Even a "factor of a few dozen" would be greatly overstating it given these benchmarks.
 
T

teh_pwn

"Saturated fat causes heart disease as much as Brawndo is what plants crave."
wsippel said:
I'm not sure you realize that Xeons and Opterons have no place in this discussion...
Click to expand...

Xeons and Opterons have far worse performance per core than consumer level processors because:
1. Intel i5/i7 core designs come out first before Xeon derivatives of that design do.
2. Consumer level processors have fewer cores per area, allowing better cooling and faster clock speeds.

Note that the i7 3770K scores 10384 with 4 cores. The top Xeons have 8 cores and aren't cracking 20k. Why? Lower clock speeds (3.2 GHz vs 3.5 GHz), less cache per core.

You're also ignoring the i5 I posted, which is the mid ranged PC CPU.
 
W

wsippel

Banned
Durante said:
A few dozen what? Orders of magnitude? That's ridiculous. Even a "factor of a few dozen" would be greatly overstating it given these benchmarks.
Click to expand...
Few dozen times. And no, it's not greatly overstating anything. ARM CPUs simply aren't as powerful as most people tend to believe.
 
D

Durante

Member
wsippel said:
Few dozen times. And no, it's not greatly overstating anything. ARM CPUs simply aren't as powerful as most people tend to believe.
Click to expand...
A few dozen times? That's just completely wrong. Sorry.
It's not like the Vita has an ARM11. A Cortex A9 isn't nearly as abysmal in terms of IPC as it would have to be for a quad core to be "a few dozen" times weaker.
 
You must log in or register to reply here.

Similar threads

Tom Warren - Sony’s PS5 Pro is real and Sony is asking developers to get games ready over the summer, with a push for ray-tracing support
News Hardware Hype 2 3 4
186
13K
DanielG165 replied
Reducing the CPU bound game lows on 2.9GHz/4.3GHz boost Core i5-10400
Analysis Hardware
13
2K
PaintTinJr replied
Switch benchmarked running PC games Natively - GoW, DMC5, GTAV, Genshin, PSV, PS2. Cinebench, 3DMark, Geekbench scores vs Snapdragon - Full breakdown
Analysis Hardware Platform 2
54
6K
justiceiro replied
  • Locked
"I Need a New PC!" 2023. 6-24 Cores, Frame Generation, Enhanced Ray Tracing & Direct Storage.
36 37 38 39 40
2K
100K
DGrayson replied
Intel Core i9-13900 ‘Raptor Lake’ Early CPU Sample Tested, Up To 50% Faster Than Core i9-12900 ‘Alder Lake’ at Just 3.7 GHz Clock Speeds
7
2K
kikkis replied
Top Bottom