i5 4200U vs Pentium G3220


Recommended Posts

Hey all. I was looking at the benchmarks for the i5 4200U and the Penitum G3220 and their scores are very close. Could I assume that the G3220 will perform about the same as the i5 4200U? I ask because I have a Yoga 2 Pro which is really fast and I'd like to build the desktop equivalent.

 

Intel Core i5-4200U @ 1.60GHz

3,378

http://www.cpubenchmark.net/high_end_cpus.html

 

Intel Pentium G3220 @ 3.00GHz   

3,217
 

http://www.cpubenchmark.net/mid_range_cpus.html

Link to comment
Share on other sites

In terms of performance, the only interesting thing that the i5 has going for it is that it is hyperthreaded (HT). They are both 4th generation i7 based processors. They are both dual core processors. They both have the same amount of caches. The i5 is significantly lower clocked and has a worse on-die GPU. And the Pentium is going to get much better single thread performance based on the clock speeds numbers.

 

The downside of the Pentium is that it uses much more energy based on the TDP numbers.

 

See:

http://ark.intel.com/products/77773/

http://ark.intel.com/products/75459/

http://cpuboss.com/cpus/Intel-Pentium-G3220-vs-Intel-Core-i5-4200U

 

EDIT: Suppose, I should clarify that I'd say they would perform about the same or even that the Pentium is going to perform a bit better in practice because of the single thread performance differences. The HT is what is making the i5 edge out on a bit on the passmark results that you are seeing, but that is unlikely to translate very well to real-world workloads. You'd only really see benefits in cpu intensive workloads when both of the cores in the processor are pegged (read as: not all that likely under most consumer workloads).  

Link to comment
Share on other sites

The i5 also has QuickSync, a much better iGPU and AVX 2 support.

The desktop would just be for email,web, youtube and Office. Would the igpu or avx2 make a difference?

Link to comment
Share on other sites

In terms of performance, the only interesting thing that the i5 has going for it is that it is hyperthreaded (HT). They are both 4th generation i7 based processors. They are both dual core processors. They both have the same amount of caches. The i5 is significantly lower clocked and has a worse on-die GPU. And the Pentium is going to get much better single thread performance based on the clock speeds numbers.

 

The downside of the Pentium is that it uses much more energy based on the TDP numbers.

 

See:

http://ark.intel.com/products/77773/

http://ark.intel.com/products/75459/

http://cpuboss.com/cpus/Intel-Pentium-G3220-vs-Intel-Core-i5-4200U

 

EDIT: Suppose, I should clarify that I'd say they would perform about the same or even that the Pentium is going to perform a bit better in practice because of the single thread performance differences. The HT is what is making the i5 edge out on a bit on the passmark results that you are seeing, but that is unlikely to translate very well to real-world workloads. You'd only really see benefits in cpu intensive workloads when both of the cores in the processor are pegged (read as: not all that likely under most consumer workloads).  

Thanks for the info

Link to comment
Share on other sites

The i5 also has QuickSync, a much better iGPU and AVX 2 support.

 

Do you see anything that shows the GPU is better other than quicksync being built in? I'm asking because the GPU is actually clocked significantly lower than the Pentium

Link to comment
Share on other sites

The desktop would just be for email,web, youtube and Office. Would the igpu or avx2 make a difference?

Not for those things. AVX is for floating point operations which are rare in general. Hell, you'd be hard pressed to find a compiler that would even be able to use the new AVX2 instructions effectively even if you did want to use them given their age.

Link to comment
Share on other sites

Do you see anything that shows the GPU is other than quicksync being built in? I'm asking because the GPU is actually clocked significantly lower than the Pentium

Ugh, it's past 2AM here, I shouldn't be trying to answer technical questions.

  • Like 1
Link to comment
Share on other sites

Ugh, it's past 2AM here, I shouldn't be trying to answer technical questions.

 

I'm on UTC-5 time also!

Link to comment
Share on other sites

Good question

 

I'm a bit confused by this question because I think Andre and I both concluded the Pentium is probably better given the higher clock. Andre pointed out that it doesn't have built in Quicksync though which is used for hardware acceleration of video playback.  

Link to comment
Share on other sites

I thought i5's were quad core?

 

Depends on the model, look at the ARK Intel links I linked for this particular model. This one is 2 core - HT (so seen as 4 logical processors). pentium, celeron, i3, i5, and i7 are all marketing jargon and meaningless in terms of feature set. But, generally, they are in a particular order which I believe now is: pentium < celeron < i3 < i5 < i7.

 

EDIT: well they aren't _completely_ meaningless in terms of feature set (e.g. i7s are probably all 4 core at the very least and I think anything below an i5 is a dual-core non-HT or lower)*, but they aren't really consistent.

 

* Disclaimer: I didn't check if my examples were strictly true.

Link to comment
Share on other sites

Not all of them, some are just dual core with hyperthreading.

NVM posted too late :p

 

I think you did add a bit of a relevant information though, they are all either 2 core HT or 4 core -- so seen as 4 logical processors. There's 4 in the mix somewhere with i5s ;-)

Link to comment
Share on other sites

about the GPU

4200U has 20 execution units, at base 200Mhz = 64GFLOPS, and it turbos up to 1Ghz = 320GFLOPS

G3220 has 10 execution units, at base 350Mhz = 56GFLOPS, and it turbos up to 1.1Ghz = 176GFLOPS

edit:got confused,4200U is a 4400 with 20EU not 16EU,updated numbers

  • Like 1
Link to comment
Share on other sites

about the GPU

4200U has 16 execution units, at base 200Mhz = 52GFLOPS, and it turbos up to 1Ghz = 256GFLOPS

G3220 has 10 execution units, at base 350Mhz = 56GFLOPS, and it turbos up to 1.1Ghz = 176GFLOPS

For Youtube and hi def videos, which would be better? Is 256 that much higher than 176 in real world performance? Would The Pentium likely turbo during these things?

Link to comment
Share on other sites

For Youtube and hi def videos, which would be better? Is 256 that much higher than 176 in real world performance? Would The Pentium likely turbo during these things?

for youtube and videos, you will notice zero difference, its will mostly affect gaming and 3d applications. and yes they will both turbo so you will be a lot of times above base frequency.

Link to comment
Share on other sites

about the GPU

4200U has 20 execution units, at base 200Mhz = 64GFLOPS, and it turbos up to 1Ghz = 320GFLOPS

G3220 has 10 execution units, at base 350Mhz = 56GFLOPS, and it turbos up to 1.1Ghz = 176GFLOPS

 

Very nice catch! Where did you find the execution unit counts, I had just assumed they were the same given the generation? Those are curiously absent on the ARK pages. I also think you are right to be looking at the turbos for this; the base probably doesn't really matter so much.

Link to comment
Share on other sites

Very nice catch! Where did you find the execution unit counts, I had just assumed they were the same given the generation? Those are curiously absent on the ARK pages. I also think you are right to be looking at the turbos for this; the base probably doesn't really matter so much.

you can find the numbers of EU anywhere,anandtech, Wikipedia

http://en.wikipedia.org/wiki/Comparison_of_Intel_graphics_processing_units

also, each EU has 2, 4 vector SIMDs, so 8 MADs per EU for a total of 16 FP ops/clock per EU.

and the base is basically the minimum,where it probably goes when idle. you will be running high when doing 3d stuff.

Link to comment
Share on other sites

What about the IGP performance?

I believe all Haswell CPUs with IGP have the same features, the Pentium has higher clock speeds so it might be faster but there's no indication of exactly what "model" it is (i.e. HD 4000, 4400 etc) although these might just be marketing terms. In any case both will be satisfactory for your needs and not what you want if you want to run current-generation 3d games. For that you should get a dedicated video card.

  • Like 1
Link to comment
Share on other sites

you can find the numbers of EU anywhere,anandtech, Wikipedia

http://en.wikipedia.org/wiki/Comparison_of_Intel_graphics_processing_units

also, each EU has 2, 4 vector SIMDs, so 8 MADs per EU for a total of 16 FP ops/clock per EU.

and the base is basically the minimum,where it probably goes when idle. you will be running high when doing 3d stuff.

 

After checking, it looks like Intel purposely left them out of the architecture specifications: http://www.anandtech.com/show/6355/intels-haswell-architecture/12. Go figure. Looks like the actual numbers were discovered after the releases.

 

I do follow the FMA talk, but I'm always wary of doubling FLOP/Cycle since people seem to see it as an average as opposed to theoretical maximum. Mainly because it assumes every floating point operation you need is a multiply-add and that you can fill the vector registers perfectly per clock. 

Link to comment
Share on other sites

After checking, it looks like Intel purposely left them out of the architecture specifications: http://www.anandtech.com/show/6355/intels-haswell-architecture/12. Go figure. Looks like the actual numbers were discovered after the releases.

 

I do follow the FMA talk, but I'm always wary of doubling FLOP/Cycle since people seem to see it as an average as opposed to theoretical maximum. Mainly because it assumes every floating point operation you need is a multiply-add and that you can fill the vector registers perfectly per clock.

yeah flops are not an accurate measure of performance, but I guess there has to be a way to do a numbers comparison to have an idea of where the chip stands. most flop measurements double the mad ops,like anandtech and others. its just like bandwidth comparisons,where its all theoretical.

Link to comment
Share on other sites

This topic is now closed to further replies.