Apple A4 vs Atom

[CU]

Has anyone seen any benchmarks of the Apple A4 vs Atom or even Nvidia Tegra?

 

[Fox5]

The Apple A4 is on a closed platform, and won't run any of the same software as atom or tegra, it's pointless.

However, it should be roughly on par, on a mhz per mhz basis, with atom and ahead of tegra. Atom runs at higher clock speeds though.

 

[Toadster]

oh but you can see it now... Atom's emulating iPAD - would that be the "HackinPAD" ?

 

[CU]

I am sure some test could be run. Just look at Anandtech's benchmarks of the iPad and Nexus One.

 

[alyarb]

anand's benchmarks? you mean the time to load a web page? can you try to understand how that provides only a dim generalization of CPU performance? loading a webpage depends on the intensity of the signal, your allotted bandwidth, the web server load/bandwidth as well as cell phone NAND performance which is typically quite low. he's really just examining overall browser performance, not CPU.

your other alternative is that java benchmark, but java doesn't account for a lot of netbook/nettop basic productivity. which is what the atom is for. yes it's faster than A4, but it doesn't do the same job. The iPad is numerically a slower computer than atom but it's still faster than the iphone and can run for ten hours, 99% of the time running an app that it's fast enough for.

if someone knows how many amp-hours you get in an ipad battery, we can try and figure out if a pinetrail slate/tablet would be better/faster per clock-watt.

 

[CU]

It shouldn't be that hard to write an app for the iPad like Super Pi or some other cpu benchmark? Then you should be able to compare it to anything.

 

[alyarb]

true, but your super pi program would have to be so small that it fits inside the A4's cache. outside of the A4 CPU you have to deal with a slow and meager amount of system RAM. the numerical comparison isn't fair because of pinetrail's memory performance (after all, it has to deal with vista/7. totally different kind of computing).

if you want to look holistically from machine to machine, then yes, all atom computers are in general faster, more powerful computers than ipad/iphone type devices.

 

[IntelUser2000]

anand's benchmarks? you mean the time to load a web page? can you try to understand how that provides only a dim generalization of CPU performance? loading a webpage depends on the intensity of the signal, your allotted bandwidth, the web server load/bandwidth as well as cell phone NAND performance which is typically quite low. he's really just examining overall browser performance, not CPU.

*snip*

if someone knows how many amp-hours you get in an ipad battery, we can try and figure out if a pinetrail slate/tablet would be better/faster per clock-watt.

Right there. He's talking about performance, not battery life or power consumption. I'd assume Atom is anywhere from 50-100% faster in single thread than the Cortex A8 based devices at the same clock. The more computation intensive benchmarks that shows MFLOPs and such will have 50% advantage, and when its memory bound it probably reaches even greater difference.

There's a reason the ARM cores use such little power and have 1-2mm2 die size. Considering how Anandtech benched it against a Netbook which uses absymally slow 4200RPM drives, I think the comparison is fine. Plus, the Windows OS is much more complex than whatever the iPad has.

Just cause the Atom is 2-issue in-order doesn't make it slow, I'd assume it might be almost on par with the coppermine Pentium III's which features a much complex 3-issue out-of-order core.

 

[Fox5]

Just cause the Atom is 2-issue in-order doesn't make it slow, I'd assume it might be almost on par with the coppermine Pentium III's which features a much complex 3-issue out-of-order core.

The Atom is roughly on par with the pentium 4 in clock for clock performance, which is pretty incredible for an in order design.

 

[Bateluer]

The Atom is roughly on par with the pentium 4 in clock for clock performance, which is pretty incredible for an in order design.

Err, I don't think that's true. Having owned an N270 Netbook, currently own an Atom 330 nettop, and having owned a few different P4s, the P4s are definitely faster, even at the same clock speed.

 

[aggressor]

anand's benchmarks? you mean the time to load a web page? can you try to understand how that provides only a dim generalization of CPU performance? loading a webpage depends on the intensity of the signal, your allotted bandwidth, the web server load/bandwidth as well as cell phone NAND performance which is typically quite low. he's really just examining overall browser performance, not CPU.

your other alternative is that java benchmark, but java doesn't account for a lot of netbook/nettop basic productivity. which is what the atom is for. yes it's faster than A4, but it doesn't do the same job. The iPad is numerically a slower computer than atom but it's still faster than the iphone and can run for ten hours, 99% of the time running an app that it's fast enough for.

if someone knows how many amp-hours you get in an ipad battery, we can try and figure out if a pinetrail slate/tablet would be better/faster per clock-watt.

3\4ths of the things you use to slight Anand's benchmarks are invalid. He used his own local webserver to carry out the tests:

I loaded all pages off of a server on my own network. This is as close to a pure CPU and OS/browser benchmark as we’ll get.

 

[IntelUser2000]

Err, I don't think that's true. Having owned an N270 Netbook, currently own an Atom 330 nettop, and having owned a few different P4s, the P4s are definitely faster, even at the same clock speed.

You sure about that? Maybe you want to check it again? Going from slower to faster isn't always noticeable, but if you go the other way it is. It's kinda like when I got my X25-M, it didn't feel much faster, but when I went to a slower HDD based system, suddenly I felt like a spoilt child because it was hard to go back to it.

The Pentium M can hold its own with Pentium 4's with 50% higher clock speed and 1.6GHz Atom is approximately on par with 900MHz Pentium M in single-threading. I think P4's might be slightly faster than the Atom, but its close enough.

 

[Drenlin]

It may even depend on which P4 you're talking about. Prescott's architecture is different enough from Northwood's that they could have called it the Pentium 5.

 

[dguy6789]

Err, I don't think that's true. Having owned an N270 Netbook, currently own an Atom 330 nettop, and having owned a few different P4s, the P4s are definitely faster, even at the same clock speed.

The 1.6Ghz Atom is roughly as fast as a 1Ghz Pentium 3. That puts it in the same area as P4 performance from a clock for clock perspective.

 

[jaydee]

Err, I don't think that's true. Having owned an N270 Netbook, currently own an Atom 330 nettop, and having owned a few different P4s, the P4s are definitely faster, even at the same clock speed.

I distinctly remember when the P4's first came out (Prescott right?), the PIII's clock for clock were a decent amount faster. Since Atom 1.6Ghz is typically categorized as PIII 1.0-1.2Ghz speed, I would think Atom and P4 are at least in the same ballpark.

 

[CKTurbo128]

I distinctly remember when the P4's first came out (Prescott right?), the PIII's clock for clock were a decent amount faster. Since Atom 1.6Ghz is typically categorized as PIII 1.0-1.2Ghz speed, I would think Atom and P4 are at least in the same ballpark.

Nope, the first P4s used the Willamette core. But otherwise, you are correct, the first P4s were somewhat slower than PIIIs clock-for-clock.

 

[Fox5]

Nope, the first P4s used the Willamette core. But otherwise, you are correct, the first P4s were somewhat slower than PIIIs clock-for-clock.

By the end, I think the P4s were getting close to the p3 in clock for clock performance. Each variant of northwood really helped performance (bandwidth/latency hungry?) and a ton more fpu power. Still, considering how many less transistors atom uses than p4, it's kind of impressive it can even match the early p4s clock for clock.

 

[CU]

http://www.anandtech.com/show/3640/apples-ipad-the-anandtech-review/17

Cool he must have saw my thread.

 

[Mr. Pedantic]

Can such a comparison even be made? The two don't even use the same instruction sets, not to mention the fact that the A4 runs on a completely different, completely closed OS.

 

[IntelUser2000]

The 1.6Ghz Atom is roughly as fast as a 1Ghz Pentium 3. That puts it in the same area as P4 performance from a clock for clock perspective.

The 1.6GHz Atoms are on par in single thread with a 900MHz Pentium M, which is far better than 1GHz Pentium 3. I think it should do around 1.3-1.4GHz Pentium III(coppermine core).

The Pentium 4's looked bad initially, but later on they weren't far off from the Pentium III: http://anandtech.com/show/757/9

Notice how even the 1.3GHz Pentium 4 can beat 1GHz Pentium IIIs. Part probably had to do with SSE2 optimizations and Pentium 4 specific code, but anyways.

Atom 1.6GHz
=1.6GHz Pentium 4
=1.4GHz Pentium III "Coppermine"
=1.3GHz Pentium III "Tualatin-256"
=1.2GHz Pentium III "Tualatin-512"
=900MHz Pentium M

 

[jaydee]

Nope, the first P4s used the Willamette core. But otherwise, you are correct, the first P4s were somewhat slower than PIIIs clock-for-clock.

That's right. Good memory (you didn't cheat by googling did you?)

The 1.6GHz Atoms are on par in single thread with a 900MHz Pentium M, which is far better than 1GHz Pentium 3. I think it should do around 1.3-1.4GHz Pentium III(coppermine core).

The Pentium 4's looked bad initially, but later on they weren't far off from the Pentium III: http://anandtech.com/show/757/9

Notice how even the 1.3GHz Pentium 4 can beat 1GHz Pentium IIIs. Part probably had to do with SSE2 optimizations and Pentium 4 specific code, but anyways.

Atom 1.6GHz
=1.6GHz Pentium 4
=1.4GHz Pentium III "Coppermine"
=1.3GHz Pentium III "Tualatin-256"
=1.2GHz Pentium III "Tualatin-512"
=900MHz Pentium M

RDRAM... Now there's a blast from the past!