PassMark Rankings of i3 and AMD A8-3850

[grahamk]

These two processors:
AMD A8-3850 APU with Radeon HD Graphics 6075

and:
Intel Core i3-2100T @ 2.50GHz 3245

According to these scores the AMD flat out kills the i3-2100 - almost 2x the performance. But then I find this:

http://www.anandtech.com/show/4476/amd-a83850-review/3

Which shows the Intel outperforming the AMD on a regular basis, and not too far behind when it loses. I've seen similar comaprisons which show the same results.

I'm a bit curious about what might be causing this discrepancy.

Also - I am building a linux system to run Virtual Box. Does the passmark CPU performance test lend itself well to this type of application?

Thanks!
Graham

[David (PassMark)]

You aren't comparing apples to apples.

The AnandTech review was looking at the i3 2105. The link in our charts you posted was for the i3 2100T. So they are not the same CPU.

Also our charts show the multithreaded performance when all cores are used. On a per core basis (or single threaded testing) the i3 looks much much better. But for applications that can fully use 4 cores the A8 does better (as the i3 only has 2 cores).

We are not aware of any problems with VirtualBox. But we use VmWare and VirtualPC in house.

[David (PassMark)]

We had a couple of other people ask about this, so it though I would post this additional detail.

In some of the individual CPU tests the A8 does really well (like finding prime numbers). In other CPU tests it does very badly like the SSE based matrix calculations.

It is not uncommon that some CPUs are better suited to particular tasks. But A8 shows an extreme variation larger than any other we have looked at.
We can't explain this at the moment.

This page highlights the differences in the individual tests. http://www.xtremehardware.it/eng-rev...01106305400/8/

We aren't the only benchmark showing the A8 are being OK,
http://www.bit-tech.net/hardware/cpu...-3850-review/4
and the Crypto benchmark here,
http://www.legitreviews.com/article/1649/4/

We think some of the other published benchmarks show the AMD chip as being a bit slower because,
1) They aren't running on all cores
2) They are running 32bit code on 64bit hardware. Where as the AMD chips can be up to 6 times faster running 64bit native code in simple cases.

We also can't see any thing obviously wrong with the results (e.g. signs of manipulation for example). But we'll keep an eye on it. If it becomes more of an issue we'll get an A8-3850 and do some detailed investigation.

[David (PassMark)]

This post goes into some additional detail.
http://www.passmark.com/forum/showth...1395#post11395

It looks like these AMD chips are significantly better with 64bit native applications compared to 32bit applications.

Update Oct 2011:
See this post for a lot more details what what is going on.
http://www.passmark.com/forum/showthread.php?t=3348