AMD A10-4655M vs Intel i7-3517U

Hi all, I'm considering to buy a "sleekbook" 6z-1000 from HP.
I tried to search the web for comparisons between AMD A10-4655M Quad core processor and the Intel i7-3517U Dual core (the last is common for ultrabooks as Samsung Series 9).
Considering virtual machines, I think the quad should perform better than the dual.
Also the graphic power of the apu makes me groove for OpenCL...
What do you think about? Maybe someone can post his detailed windows experience index (showing the cpu value)?
many thanks

 

The AMD A10-4655M level with the old A8-3550MX 45W APU, therefore superior of I7-3517U Dual Core. Here you can read a proper review of 6z-1000 with A10-4655M.

Some comparsion:
Intel Core i7-3517U: around 630 at 3DMark11
A10-4655M: 895 at 3DMark11

About CPU performance: 4 cores will handle better more threads than 2+HT. Do not listen Chinebench or similar syntetic craps, means nothing in real situations .

 

The i7-3517U will give you better CPU performance. Don't be fooled by the "quad" labeling -- it does not have four cores in the same sense that previous processors did (there are four integer cores, but only two FPUs and only two of various other resources usually associated with cores). As a result, the A10 "quad-cores" tend to be slower than Ivy dual-cores with hyperthreading. I can't find any benchmarks for the 4655M, but here is a comparison for some applications from notebookcheck for the A10-4600M:

i7-3517U vs. A10-4600M

Cinebench R11.5 64Bit: 2.8 vs. 2.0
WinRAR 4.0: 2376 vs. 1891.5
TrueCrypt AES: 1.4 vs. 1.2

The only difference between the 4600M and the 4655M is that the 4600 is clocked at 2.3Ghz whereas the 4655M is only 2.0GHz. Thus, the 4655M will be even slower.

On the other hand, unless Intel is much more aggressive about Turbo, the 4655M should have a better GPU.

 

good to know, Althernai
Atom_Anti, I know A10 is awesome in 3D, but I don't need it (except some apps with opencl support).

many thanks!

 

Althernai; Chinebench roll the dice, check a real CPU situation:



+he Considering virtual machines, where fore integer cores rather working as quad core as 2+HT.

 

That graph doesn't actually have any Ivy CPUs on it. The difference between floating point and integer workloads is that in the former, the A10 will lose by a hilariously large margin whereas in the latter, it will merely lose. Here are a bunch of real-life applications comparing the i5-3427U to the A10-4600M. Note that the i5-3427U is clocked at 1.8GHz with 2.8/2.6 single/dual max turbo whereas the i7-3517U is clocked at 1.9GHz with 3.0/2.8 single/dual max turbo. Thus, the 3517U will be faster than the 3427U and, as I already mentioned, the 4655M will be slower than the 4600M... but the 4600M still loses in practically everything to the 3427U.

I do not know about virtualization -- I think manzamanna is the first person I have heard of to try doing it with these power-optimized CPUs. However, Ivy is faster than Trinity in practically every CPU test I can think of so I'm pretty sure the 3517U will be faster there as well. And while the A10's do have four integer cores, there are a lot of shared elements between the pairs in a module (L2 cache, instruction fetching and decoding, etc.) so Intel's nearly independent cores are not directly comparable to AMD's.

 

That benchmark has neither of the CPU's that OP listed.

Atom_Anti, first you say that synthetic benchmarks don't matter and then list 3dmark scores. And the 3200 point 3dmark06 score you listed for the i7-3517 is the CPU only score, not the overall score. The A10-4655m from the other link you posted has a CPU score of 2000... And then you claim that the the A10-4655m is faster. Right.

OP, both of those CPUs are going to be mediocre for "serious" VM usage. If you really want good performance with VMs you should get a full powered i7.

 

I sad Chinebench or similar syntetic craps do not matter. I'll delete 3dMark06 scores, seems unfair for comparsion, you are right at that point.
And of course I'm claiming A10-4655M faster than I7-3517, since these are APUs which means lot more than just a CPU. Therefore we have to consider overall CPU+GPU performance. But let's back to only CPU tasks and reviewing Althernai link; where I see minimal difference in Winrar contest, plus we are already know newer version of Winrar support OpenCL acceleration which gives big boost to AMD. I do not know what is TrueCrypt, but supports acceleration via Intel's AES-NI instructions, probably not favor for AMD.

Finally he need multithreading, so try something similar on I7-3517:

The real power!

I tell I7 will fail to do this.

 

That doesn't make any sense -- they are not interchangeable, at least not in a way that is meaningful to most people. There are very few workloads that are performed by both -- more now than before, but still very few. If your workload demands a fast CPU, go with Intel. If you need a fast GPU, go with AMD. If you want both, you need Intel + discreet graphics (or wait for Haswell). manzamanna's workload is mostly CPU-heavy therefore Intel is the better choice.

TrueCrypt is encryption/decryption software. If you look just below the AES graph, there's also Twofish which is without any special instructions -- the i5 still wins.

You can see the OpenCL applications on the next page of the same review. As the reviewer says, "the A10-4600M doesn't really distance itself from the 17W Intel CPUs—not even with its IGP chipping in via OpenCL." It wins one and loses one, but the difference is small -- the Ivy IGP is not that bad and Amdahl's Law does the rest.

And yes, the difference between the 3427U and 4600M is not much for many things... but keep in mind what is being compared. The 4600M is the fastest mobile product AMD is offering right now -- it is optimized for speed and has a TDP of 35W. The 3427U is a 17W processor intended for ultrabooks and it's not even the fastest of the 17W processors. TDPs from Intel and AMD are not directly comparable, but 17W is definitely much less than 35W... and yet the 3427U wins practically across the board. The difference between the i7-3517U and the A10-4655M is going to be much larger.

 

I think you may have answered your own question here. The A10 will have the upper hand in 3D graphics but you say you do not need it. If better CPU performance is what you need, then the choice is obvious, go with the i7. But in my opinion, in real life usage the difference might not be as dramatic as benchmarks make it out to be. Both of the processors will handle daily tasks with ease. The biggest difference will come down to price and other factors such as battery life, construction, keyboard and display quality to name a few.

I second this. Unless you're purposely after a LV/ULV chip or the ultra-thin form factor, (performance-wise) you would be better off with a standard voltage CPU like a quad i7. What is your budget and or purpose/goals for this computer? More details would be of help.

 

Not sure if you understand the importance of APUs.
What if someone want both, but there is no option for discrete graphics? or simply he just do not want extra power consumption and heat due to 2nd engine? Maybe need longer battery time?
Than he need to decide which one is the better APU. While AMD APUs are capable for everything; decently can run CPU tasks and games, laptops with Intel APU will feel like a crippled machine, because cannot run games in decent quality.

 

Not sure you understood his statement?

If workload demands a fast cpu, Intel... if workload demands both (cpu/gpu) then Intel + discreet graphics...

Seems very logical to me.

What you're saying is that if you don't care about fast (cpu or gpu...) then go with AMD (and they barely make sense (cpu performance-wise) unless you're just gaming...).

 

Passmark scores the 4655M as slightly faster:

3517U: 3999
4655M: 4162

PassMark - AMD A10-4655M APU - Price performance comparison

PassMark - Intel Core i7-3517U @ 1.90GHz - Price performance comparison

 

Perhaps you are talking about big Alienwares category or desktops, but take a look to the article's title and try to understand my one before post.

 

No, I know which thread I'm posting in - you're too focused on 'AMD rules'... without considering other posters (not mine) viewpoints.

 

OP wants CPU power for VMs. Benchmarks have been posted showing that the i7 will be faster, including benchmarks from yourself.

Then you go on to talk about how the A10 has a stronger iGPU while downplaying the weaker CPU... After OP has stated that he doesn't need GPU power.

I think you need to follow your own advice.

 

We can't have both -- not for another 7-10 months, anyway. Every user needs to decide what is more important for their typical loads. If the question was about a thin-and-light gaming laptop, then my answer would have been the A10-4655M, but since the purpose was virtualization, the i7-3517U is definitely the better of the two.

 

OP, you mentioned WEI scores. Here's some reviews and general impressions from recent owners of the HP ENVY Sleekbook with A10-4655M and A6-4455M.

Proper Envy 6z review
Envy 6z Initial Impressions
HP ENVY Sleekbook 6z Review: Is Being Thin Enough? Discussion
HP ENVY Sleekbook 6-1010us Quick Review

 

I'm still here because considering other posters and cannot understand why so important more and more CPU power. All today's CPU is decently enough for every commonly used tasks .

We already can have it, since Llano! and when Haswell APU appears, than might can get close to Trinity, but Kaveri APU is coming too...

 

Do Passmark scores mean anything? If they do, the 4655M is the clear winner.

Sent from my HTC VLE_U using Tapatalk 2

 

Passmark is garbage. According to Passmark, the i7-3615qm is faster than the desktop i5-3570k.

 

Yeah, we've had this discussion before -- Passmark is basically a random number generator. There's very little rhyme or reason to it. If you want more examples of its gibberish, the i7-3517U and A10-4655M are both somehow faster than the Phenom II X4 955, a desktop processor with 4 full-fledged cores clocked at 3.2GHz and with a full complement of L3 cache. Passmark has no relation whatsoever to any other benchmark or real world application. I'm kind of curious what they're doing to achieve these results (counting instruction sets and multiplying by the number of cores and clock speed?).

 

..um. Why is synthetic single core performance even mentioned when looking at low-end mobile cpus..?

And, I mean, the i7 is faster when using some of the unique instruction sets it supports. On the other hand, OpenCL performance on the apu setup can be expected to be very high. There are a few common situations with hardware video decode as well where the apu setup is going to perform well.

Besides that, what we're really interested in is the power-consumption, no? Specially while running all the cores, including the apu. But also while idling. Since the Llano setups already were fairly impressive when it came to that, compared to the similarly rated intel setups..

I mean, you're not buying a 13 inch sleekbook for synthetic performance, or single-core server performance..? Right?

 

Don't forget the Intel's i7 is superior in more ways than one. Not only is it built on a smaller more efficient (17W TDP vs 25W TDP) architecture (22nm vs 32nm) it is also able to use 1600 RAM vs the A10's 1333 RAM. It's clock frequencies are higher, it also utilizes L3 cache, and has a Turbo feature that works a lot better than what AMD has developed. The technology behind it is more advanced, no question about it, compared to what AMD has come out with but YOU WILL pay more for that privilege. What AMD has done with their strengths and weaknesses shouldn't be overlooked either. They provide a solution that is able to compete on some levels (IGPU, battery life, good enough CPU) but it does costs less. It'll work for some, and it won't work for others. Analyze your needs and just choose accordingly.

 

This statement has solved my quad vs dual core doubts. (thanks)
It is clear that A10 is better for gaming and apps with massive opencl support (for example Photoshop).
I don't think we should consider benchmarks because they test a large variety of usage.
for example:
accelerated video encoding/decoding > a10 wins
3D software or games > a10 wins
software with opencl support > a10 wins
software without multi-core optimization > cpu with higher clock wins
near all other cases > i7 wins

As I told, I was looking for this:
software optimized for multi-core > better A10 quad or i7 dual?
The Althernai's reply hit the mark