AMD A4-5000 Review: The affordable ultraportable APU

By on May 24, 2013, 2:37 AM

In 2006 AMD announced 'Fusion', a project aimed to develop a system on a chip that combined a CPU and GPU on a single die. Making the dream a reality for AMD was their acquisition of graphics chipset manufacturer ATI that same year.

Fast forward to this day, AMD has taken things a step further with Kabini, the first ever quad-core x86-based SoC. AMD are releasing their first Kabini based processors today with the launch of the A6-5200 and the A4-5000. The A4-5000 that we are reviewing features four Jaguar cores clocked at 1.5GHz and a total L2 cache of 2MB, while the on-die GPU is the Radeon HD 8330. With the Kabini platform AMD is aiming to compete in the subnotebook, ultra-thin and small form factor markets.

Read the complete review.




User Comments: 34

Got something to say? Post a comment
VitalyT VitalyT said:

I don't know... with the new Atom to be announced very soon, and this new APU being sub-par with even Core i3, it can only find customers being pushed into bottom-priced mobile devices by OEM-s, if it ever happens that is.

spydercanopus spydercanopus said:

This SoC stuff is amazing. It makes me wonder how much longer until absolutely everything on a typical motherboard is fully integrated into a single die.

Skidmarksdeluxe Skidmarksdeluxe said:

This SoC stuff is amazing. It makes me wonder how much longer until absolutely everything on a typical motherboard is fully integrated into a single die.

Yeah, including the screen.

VitalyT VitalyT said:

This SoC stuff is amazing. It makes me wonder how much longer until absolutely everything on a typical motherboard is fully integrated into a single die.

Most companies that manufacture reusable components strive for universality of their solutions. Sticking all into one box works against that. It is much better/practical to pick up just the right components one needs for the system. This is why fully integral solutions fail on a wider market. People are more open to systems they can change (upgrade or downgrade).

That's why I wouldn't hold my breadth that systems as you described will make an appearance, there is no real demand.

Guest said:

At this moment in time AMD APUs are quite a lot faster though in general I wouldn't recommend gaming on an IGP. Still, should run some older games reasonably well. I don't know what performance Haswell is going to offer but at the moment AMD definitely have the edge on that front. However this review does also note that the CPU has some serious shortcomings on the processing front so it's a matter of what your needs are really

JC713 JC713 said:

Good try by AMD. I really want to see the A10 6800K benchmarked. @Steve, what driver was used? 13.4?

Guest said:

Awesome graphics, to be commended and so on. I'm not so sure about the necessity of four underperforming cores in mobile environments. Pure single-threaded power will always be important. Does Turbo really eat away power like that?

Guest said:

There was a demo running Dirt 3 on GT 640M and it was comparable to what the Haswell IGP was doing. Haswell will be a necessary step forward for Intel but will still be far behind Richland, and AMD has all the driver experience to make their stuff do actually well in games. AMD finds itself in a really nice spot all of a sudden.

1 person liked this | Guest said:

AMD focuses on multithreaded power, Win8 takes very good advantage of this for example.

Although Win7 does too after a patch/update. To a lesser extend then Win8 still though.

Software is slowly changing to multithreaded CPU's.

AMD might have less powerful chips than Intel, but they exceed well in the APU's and Multithreading sections. Where Intel just tries to shove more power down a single thread.

If you keep power usage identical and the same chipset, multithreaded 2 1.5ghz cores will always be faster then 1 single 3ghz core. Especially in a multitask environment. Even though the raw power is identical.

1 person liked this | hahahanoobs hahahanoobs said:

If you are someone that has remained unbiased, you seen this coming.

BLARG! BLARG! said:

This doesn't prove anything till we see a GDDR5 a4-5000 benchmark. That's the real pudding!

Staff
Steve Steve said:

This doesn't prove anything till we see a GDDR5 a4-5000 benchmark. That's the real pudding!

I think it proves everything. The GPU is on-die and borrows system memory, how many systems have GDDR5 memory as their system memory?

AMD focuses on multithreaded power, Win8 takes very good advantage of this for example.

Although Win7 does too after a patch/update. To a lesser extend then Win8 still though.

Software is slowly changing to multithreaded CPU's.

AMD might have less powerful chips than Intel, but they exceed well in the APU's and Multithreading sections. Where Intel just tries to shove more power down a single thread.

If you keep power usage identical and the same chipset, multithreaded 2 1.5ghz cores will always be faster then 1 single 3ghz core. Especially in a multitask environment. Even though the raw power is identical.

In no way is the issue for AMD clock speed! Its core efficiency, yes higher clocked cores using something like Turbo Boost will improve performance as we noted but ultimately its core efficiency that lets AMD down at the moment. Also not sure if you were just pointing out the Windows 8 vs. Windows 7 thing for the hell of it or not but if not we tested using Windows 8.

Guest said:

You can clearly see how intel are optimising for benchmarks.

A clear win for intel in 3Dmark, yet they don't win one of the real world gaming tests.

I forsee lots of cheering about Haswell's 3Dmark performance in the coming weeks.

hahahanoobs hahahanoobs said:

Is it worth it for AMD to continue with desktop APU's if the only thing they can improve is the GPU portion?

Staff
Steve Steve said:

You can clearly see how intel are optimising for benchmarks.

A clear win for intel in 3Dmark, yet they don't win one of the real world gaming tests.

I forsee lots of cheering about Haswell's 3Dmark performance in the coming weeks.

No one focuses on just one benchmark and Intel has not optimized for these tests. They are faster look at the encoding results, they are more significant than 3DMark.

BLARG! BLARG! said:

I think it proves everything. The GPU is on-die and borrows system memory, how many systems have GDDR5 memory as their system memory?

Actually Steve the chip was designed to support GDDR5 as system ram, where we may not get dual channel on the processor we do get insanely fast and high bandwidth controllers of GDDR5. Even the processors are optimized to utilize them better. However with standard DDR3 there is not enough channels natively constructed in the ram. Even if it's only 128 bit channels broken into 8 x 16, 4 x 32, 2 x 64. It'll allow all the cores to access it in one cycle where DDR3 fails even more in IPC with AMD chips as the cores could only access one time per cycle. Anyway the A4-5000 just using GDDR5 should gain performance nearly 100% over it's DDR3 counterpart, theoretically of course. That is within the given performance marks AMD was stating, no? Only problem with GDDR5 is it's latency, but that latency does favor a graphics card company.

Staff
Steve Steve said:

Anyway the A4-5000 just using GDDR5 should gain performance nearly 100% over it's DDR3 counterpart, theoretically of course. That is within the given performance marks AMD was stating, no? Only problem with GDDR5 is it's latency, but that latency does favor a graphics card company.

It won't, the GPU is in no way memory limited, at least that is what I believe.

ThanosPAS ThanosPAS said:

I don't get why the reviewer compares repeatedly Kabini with CPUs that are not meant to compete with it and why he pays such attention at the fact that Kabini isn't powerful with graphics. This is an entry APU meant for devices under 300$. It would be more accurate if the review included, instead of useless (for this APU) multiple game benchmarks, tests about real world productivity and how the APU handles everyday situations and how the new quad core design works with multi threaded applications. In my opinion this is not a representative review of what this APU can deliver.

Staff
Steve Steve said:

I don't get why the reviewer compares repeatedly Kabini with CPUs that are not meant to compete with it and why he pays such attention at the fact that Kabini isn't powerful with graphics. This is an entry APU meant for devices under 300$. It would be more accurate if the review included, instead of useless (for this APU) multiple game benchmarks, tests about real world productivity and how the APU handles everyday situations and how the new quad core design works with multi threaded applications. In my opinion this is not a representative review of what this APU can deliver.

AMD tell us that the A4-5000 is a direct competitor for the Pentium 2117U and we also included the slower Celeron option. So Kabini was tested against its direct competition.

There were several other tests included in the review, not just gaming but it doesn't really matter since the A4-5000 was compared to its direct competitor in every last test.

We tested with every day applications and we tested with applications that can utilize multiple threads? Did you even read the review?

ThanosPAS ThanosPAS said:

Most of tests were about game performance, why? Is it realistic to bench this APU against battlefield for example or Metro? Using FRAPS? (An outdated benchmarking technique). If not, is the methodology you used mentioned anywhere? I can find it. I 've read the review, you have an xcel test as well as zip one. Under which conditions? Why didn't you include a synthetic bench (not futuremark - or other graphic oriented bench) about these tasks (xcell+zip) under casual use - surfing the web, watching an HD movie, coping large files etc. Photoshop? Really? I use Photoshop everyday and I wouldn't even think of using Kabini on a tablet. is Intel Core i5-3570K a direct competitor? Or A10-5700? Tests could be more about what it can rather what it cannot. Have a nice day.

ThanosPAS ThanosPAS said:

"...and I wouldn't even think of using photoshop on a tablet." (correction)

Staff
Steve Steve said:

The Core i5-3570K was included mainly for the HD 4000 Graphics and the A10-5700 because it's a flagship APU, were either of them discussed in the text as competitors or even used to make our comparison? The answer is no.

You say most of the tests were about game performance, if you mean most of the tests in the gaming performance section then yes you are right. PCmark 7 is not a gaming benchmark, nor is Excel, WinRAR, Photoshop, Handbrake, Video Master Works or even the x264 Benchmark. The definition most shouldn't be used when you are talking about 4 out of 15.

We are not benchmaking this APU against Battlefield. Battlefield was one of the games used to show how the A4-5000 performs in relatively modern games and how that performance compares to the competition. The game itself doesn't really matter as long as both competing products are run under the same conditions.

Also Fraps is not an outdated benchmarking technique and we will continue to use it in the future. We show frame time performance in our higher-end GPU articles and we believe Fraps to be accurate regardless of what other sites might be going with FCAT for example.

We said in the conclusion that the A4-5000 is perfect for web surfing and watching HD 1080p movies, again you must have missed that.

No one cares what you would or wouldn't run on a tablet (not that this is a tablet only processor, we tested with a notebook :S). What is important is that we showed how the A4-5000 compares to its competitor in a number of tests, the margins were all much the same so you can safely assume for single and multi-threaded tasks that the Intel Pentium U2117 is faster. For GPU related tasks the A4-5000 is better but not by much and not enough to be useful.

Overall both are ideal for surfing the web and watching HD content as noted in the conclusion. Ohh and you can click the edit button to correct any mistakes in your posts.

ThanosPAS ThanosPAS said:

PCmark 7 is not a gaming benchmark, nor is Excel, WinRAR, Photoshop, Handbrake, Video Master Works or even the x264 Benchmark. The definition most shouldn't be used when you are talking about 4 out of 15.

I believe PCmark 7 is one of the very few tests for everyday scenarios that you used. Loading an application is an indicative example, I thought that my point was clear. Heavy workloads outside the lab result from many applications working at the same time, cache limitations, etc. Did you consider including a memory allocation test?

You say that the article isn't oriented in what usually sells a review the most? The gpu capabillities of the APU? Making a little biased?

We are not benchmaking this APU against Battlefield. Battlefield was one of the games used to show how the A4-5000 performs in relatively modern games and how that performance compares to the competition. The game itself doesn't really matter as long as both competing products are run under the same conditions.

I think you got what I meant here too. Battlefield was an example. Define competition. In general? All I am saying is that there 're levels of competition as you know, and a generalized test (using game that don't fit the power of the APU) can be misleading.

Also Fraps is not an outdated benchmarking technique and we will continue to use it in the future. We show frame time performance in our higher-end GPU articles and we believe Fraps to be accurate regardless of what other sites might be going with FCAT for example.

Time will tell...

We said in the conclusion that the A4-5000 is perfect for web surfing and watching HD 1080p movies, again you must have missed that.

[COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#111111]
COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#111111
[COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#11111
][COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#1111
1][COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#111
11][COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#11
111][COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#1
1111]Yes, I did missed that.[/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR
[/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/
OLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COL
R][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR][/COLOR]

No one cares what you would or wouldn't run on a tablet (not that this is a tablet only processor, we tested with a notebook :S).

The point was that this APU is going to be used in devices of 12-inches and less and benchmarking a rather resource-intensive program as Photoshop were people traditionally use for editing usually enough that lasts for hours, isn't the best thing you could do.

Overall both are ideal for surfing the web and watching HD content as noted in the conclusion.

Yes.

Ohh and you can click the edit button to correct any mistakes in your posts

[COLOR=#111111][COLOR=#111111][COLOR=#111111][COLOR=#111111]
hanks[/COLOR][/COLOR][/COLOR][/COLOR]

BLARG! BLARG! said:

It won't, the GPU is in no way memory limited, at least that is what I believe.

What you believe is incredibly wrong. All the APUs are heavily memory limited. Single channel is enough to halve the GPU's physical performance.

[link]

[link]

Staff
Steve Steve said:

Okay a few things that might not make me seem so incredibly wrong :S

Firstly 1333MHz memory is too slow, 1600MHz is the sweet spot and we have found this in the past when conducting far more extensive memory testing that what was shown in that thread. 1866MHz offers a slight bump in performance over 1600MHz but only very slight and we are talking only when using the high-end APU's. Beyond 1866MHz we saw next to no improvement in our tests.

Now with the A4-5000 processor which was tested with 1600MHz memory, changing the memory to 1866MHz had no impact on performance at all. The GPU side of things was simply to slow to require more bandwidth, the GPU clock needs to be increased in order to take advantage of faster clocked memory.

The A8-3510MX processor featuring the Radeon HD 6620G which was used for testing in the thread you linked is considerably more complex than that of the A4-5000 and it is backed by a much faster processor.

Finally don't just take my word for it, here are Legit Reviews results from way back...

http://www.legitreviews.com/article/1652/6/

Note how there is a big jump from 1333MHz to 1600MHz and then nothing from 1600MHz to 1866MHz.

BLARG! BLARG! said:

Okay a few things that might not make me seem so incredibly wrong :S

Firstly 1333MHz memory is too slow, 1600MHz is the sweet spot and we have found this in the past when conducting far more extensive memory testing that what was shown in that thread. 1866MHz offers a slight bump in performance over 1600MHz but only very slight and we are talking only when using the high-end APU's. Beyond 1866MHz we saw next to no improvement in our tests.

Now with the A4-5000 processor which was tested with 1600MHz memory, changing the memory to 1866MHz had no impact on performance at all. The GPU side of things was simply to slow to require more bandwidth, the GPU clock needs to be increased in order to take advantage of faster clocked memory.

The A8-3510MX processor featuring the Radeon HD 6620G which was used for testing in the thread you linked is considerably more complex than that of the A4-5000 and it is backed by a much faster processor.

Finally don't just take my word for it, here are Legit Reviews results from way back...

http://www.legitreviews.com/article/1652/6/

Note how there is a big jump from 1333MHz to 1600MHz and then nothing from 1600MHz to 1866MHz.

Yeah that's because the processor still only runs as a single channel. You pair it with GDDR5 which has it's multiple channels tethered into the ram system and the fps will climb! However we won't see the GDDR5 laptops till a manufacturer approves to build one. Llano is a horrible example, but it's still valid. Trinity does even better and that can only leave future ones to do better, usually.

JC713 JC713 said:

I think this should be retested with the 13.6 beta driver.

Staff
Steve Steve said:

I think this should be retested with the 13.6 beta driver.

Its not a driver issue, the drivers are well matured. Its a complete lack of GPU power issue.

GhostRyder GhostRyder said:

With APU's, ive really only seen in real life (IE hands on, I don't mean internet benches) the Mobile Versions (I bought my sister as a present for example, a laptop with an A10 5800m Processor laptop which is a very nice laptop).

Personally after viewing this, it seems like the A10 for example is actually not bad (I was actually quite surprised it was right behind the 3570k in many of the tests).

This A4 though, looked like it was lagging behind but at the same time I kinda like it for the idea of a decent affordable mobile option because of its low power usage and decent integrated graphics. I think the APU's will really start to shine in the years to come once they can start getting even higher powered GPU's on the chip and giving the CPU more power. I like all the different comparisons in terms of benches games and performance in general because it really shows and highlights the potentials of each chip and where they lie.

JC713 JC713 said:

Its not a driver issue, the drivers are well matured. Its a complete lack of GPU power issue.

True. Are we going to see a A10-6800K benchmark when it is released?

Staff
Steve Steve said:

True. Are we going to see a A10-6800K benchmark when it is released?

You will

Guest said:

The power of the AMD Jaguar APU lies in the multiprocessing capabilities and this can be

seen in the OpenCL acceleration tests. The future of computing is OpenCL and many

more apps from Adobe to Web browsers GL to image/video editing are taking advantage of the power in CPU plus GPU computing.

Please take a look at the TechReport review of the A4-5000 where they use OpenCL accel

and the Kabini outperforms Intel Ivy Bridge i3 and even i5 either outright or in terms of performance per watt and certainly in terms of price performance.

techreport.com/review/24856/amd-a4-5000-kabini-apu-reviewed

Guest said:

Techspot - you should've done a performance per watt comparison, because the A4-5000 uses about half the watts at idle and notably less than half at load than the celeron or pentium do. It wins in performance/watt like a boss. That's a big deal in today's market.

Guest said:

I have been using an HP Pavilion DV4, which has an Intel Pentium Dual-Core T4200 @ 2GHz, for the past 2 years as my main PC.

I also use Photoshop CS6 and CC with no problems.

Of course.. I have it tweaked out for max performance.

I just got one of these AMD A4 chips for christmas,

and have been researching the specs.

Im saddened to discover the AMD A4 only slightly out-paces my 7-year old Intel :(

- that said.. I hope the A4 atleast supports Virtualization (Intel's VT-x)?

@ the reviewer

I've been reading reviews, benchmarks all morning.

This is the most thorough review I've seen.

Excellent job.

Load all comments...

Add New Comment

TechSpot Members
Login or sign up for free,
it takes about 30 seconds.
You may also...
Get complete access to the TechSpot community. Join thousands of technology enthusiasts that contribute and share knowledge in our forum. Get a private inbox, upload your own photo gallery and more.