CPU/IGP tests

[bgt]

Using Winrar Benchmark for CPU and Furmark(1980x1024) for IGP testing":

1. Furmark
2. Winrar
3. Winrar (+Furmark)

i5-2500K= 145,5000, 3900
i3-3225= 304, 3800, 2100 HToff= 302,2300, 1300
A4-5300=203, 2000, 1600
A10-5700=425, 3800, 3000
i3-2100T=75, 3200, 3000

Why does the 3225 has such a bad Winrar benchmark when the IGP is loaded? Throtteling because CPU is overloaded perhaps? Can anyone test another 4000 IGP in an i5 or i7 please and put the results here.

 

[ShintaiDK]

Because Furmark uses the CPU as well. And the 3225 is a dualcore with HT?

How does the test look if you disable the HT on the 3225.

Your numbers also seems abit high compared to this:

 

[SPBHM]

Because Furmark uses the CPU as well. And the 3225 is a dualcore with HT?

How does the test look if you disable the HT on the 3225.

Your numbers also seems abit high compared to this:

winrar 4.20 beta does 3800KB/s on my i3 2100,

the A4 is a dual core without HT (and shared resources), so the loss should be worse if this is more limited by the CPU!?

 

[ShintaiDK]

Another reason could be memory bandwidth, since winrar benchmark is very sensitive to that. And the A4 is simply too slow CPU wise to really make a big impact there.

I get 4700Kb/sec. Even tho mine CPU wise is around 10% faster than the 2500K that gets 5000. In this case I simply assume the OP got lower latency on his memory.

And to the OP, try test with your A10-5700,

 

[bgt]

OP????what does it mean?
Updated my benchmark and you can see that under heavy strain HT is not very good and that explains why the A4-5300 feels faster/more fluid.
The 3225 cpu was not throtteling down as I could see in CPU-Z. It remained at 3.2Ghz when using both benches.

 

[ShintaiDK]

OP????what does it mean?

Means you in this case

Can you try with your A10-5700?

 

[ehume]

OP = Original Poster or Original Post, depending on context.

 

[bgt]

Means you in this case

Can you try with your A10-5700?

I did, look at the beginning of this thread!!

 

[bgt]

OP = Original Poster or Original Post, depending on context.

:thumbsup:

 

[ShintaiDK]

I did, look at the beginning of this thread!!

I see your numbers without HT plans out much more regular. In other words disregarding the IGP as an issue.

Seems Furmark just got a high execution unit demand compared to Winrar, if I was to throw a quick guess.

 

[bgt]

winrar 4.20 beta does 3800KB/s on my i3 2100,

the A4 is a dual core without HT (and shared resources), so the loss should be worse if this is more limited by the CPU!?

An A4-5300 is 1 real core + 1 extra integer core(=without FPU and some shared cache).
The A10-5700 is a double A4-5300(hence the freqs are a bit different)=like an i3-3225 only much better(only my opinion! since I use them both and can compare). Power consumption is a bit different but to me irrelevant in daily use.

 

[sm625]

All IGP's suffer from multitasking problems when they both use the gpu. Furmark plus youtube results in dropped frames. (Even youtube 360p). I've tested an i7 ivy bridge and a llano A4. Same goes with a simple game like league of legends plus a youtube. That is why I strongly recommend agaisnt ALL IGPs, no matter how fast they are, if you plan on running more than one of anything media/gaming oriented at one time. And in the year 2013, who doesnt do that?

From my previous review thread:

Furmark running has virtually no effect on my general usage experience on my machine that has the discrete gpu. But on these integrated graphics systems, the effect is horrific. I noticed the same thing on my llano system. If you load the gpu it inexplicably slows down the whole system to a crawl. This does not happen the same way with discrete cards, even a supposedly far less capable discrete card! To my knowledge, not one tech review site has revealed this problem. And it definitely is a problem.

The point still stands. Anyone even close to the power-user end of the spectrum is going to be severely aggravated by the performance of even the best IGPs. And it still is not being covered by the tech media in the way I think it should be.

 

[imported_paulc871]

Those are some good scores

 

[tynopik]

That is why I strongly recommend agaisnt ALL IGPs, no matter how fast they are, if you plan on running more than one of anything media/gaming oriented at one time.

can you come up with a benchmark that shows this that doesn't involve furmark?

 

[SPBHM]

An A4-5300 is 1 real core + 1 extra integer core(=without FPU and some shared cache).
The A10-5700 is a double A4-5300(hence the freqs are a bit different)=like an i3-3225 only much better(only my opinion! since I use them both and can compare). Power consumption is a bit different but to me irrelevant in daily use.

as far as I know AMD have been selling this CPU (A4 trinity) as Dual Core, 2 cores with shared resources (a single module), to be honest I wouldn't recommend the 3225 most of the time, if you need to use the IGP for something Trinity is clearly faster on desktops, if you don't, the HD 2500 should be enough, but... as for the CPUs overall, it's clear that both are close in most applications



yes, maybe it related to how it handles memory and the L3 cache with the IGP!?

 

[zebrax2]

Maybe IGP is saturating the bandwidth leaving the CPU bandwidth starved?

 

[bgt]

... as for the CPUs overall, it's clear that both are close in most applications yes, maybe it related to how it handles memory and the L3 cache with the IGP!?

For me the A4-5300 is on a par with the 3225. A10-5700 is absolutely faster in daily use over the 3225. Its more equal to the 2500K. Just a small detail.

 

[SPBHM]

For me the A4-5300 is on a par with the 3225. A10-5700 is absolutely faster in daily use over the 3225. Its more equal to the 2500K. Just a small detail.

really? on what?
the winrar benchmark you posted is telling a different story,

the memory score of your i3 seems... wrong
http://forums.anandtech.com/showpost.php?p=34497173&postcount=255

even with single channel my 2100 did 7.4, or the IGP really affects the memory score on WEI that much?

single module trinity on par with the ivy bridge i3?
http://www.tomshardware.com/reviews/a10-5800k-a8-5600k-trinity-apu,3241-4.html

something is wrong if your testing or parts.

 

[bgt]

Will do a test on the i3-3225 again tonight and check the settings again.

BTW a bench test with Adobe does not tell enough about how good a certain pc is.
You need a bit more real life tests to tell how good it is. The A4-5300 is just awsome in both fluidity when using it and graphic qualitiy.

 

[sm625]

can you come up with a benchmark that shows this that doesn't involve furmark?

I wish I had numbers, but I have no way of obtaining a count of dropped frames on youtube. But it can be easily replicated without using furmark. If you have even a modest (2D) DX9 game like LoL running in one window, while youtube 480p is playing in another window, the frame rate on the youtube clip will be low and inconsistent. An i7-3770 with HD4000 will stutter. You can see it quite clearly. It is a very simple real world test.

This same test performed on an E6600 with Nvidia ION pciE x1 discrete card shows much better performance, even though the nvidia ION (9400) card scores lower in 3dmark, passmark, and several other gaming benchmarks on its own. The problem occurs even with multiple flash apps running. The window not in focus will see a serious performance penalty.

Most/all tech review sites do not cover performance of the content in the window not in focus, but nonetheless real people in the real world do want good performance from their background apps. Especially if your system supposedly has more than enough horsepower to run these apps on their own, in focus.

 

[bgt]

Ok, tested the 3225 again and found the difference to be the memory being 4Gb. So now its also 8Gb. Will update all the figures.

 

[bgt]

Using Winrar Benchmark for CPU and Furmark(1980x1024) for IGP testing":

1. Furmark
2. Winrar
3. Winrar (+Furmark)

i5-2500K= 145,5000, 3900
i3-3225= 304, 3800, 2100 HToff= 302,2300, 1300
A4-5300=203, 2000, 1600
A10-5700=425, 3800, 3000
i3-2100T=75, 3200, 3000
i7-3770K= 321, 8400, 4100

It shows the 3770K is heavily influenced by the IGP. The CPU calculating power drops significantly. So for a laptop thats dramatic because base clocks are also a lot lower.

 

[Idontcare]

Using Winrar Benchmark for CPU and Furmark(1980x1024) for IGP testing":

1. Furmark
2. Winrar
3. Winrar (+Furmark)

i5-2500K= 145,5000, 3900
i3-3225= 304, 3800, 2100 HToff= 302,2300, 1300
A4-5300=203, 2000, 1600
A10-5700=425, 3800, 3000
i3-2100T=75, 3200, 3000
i7-3770K= 321, 8400, 4100

It shows the 3770K is heavily influenced by the IGP. The CPU calculating power drops significantly. So for a laptop thats dramatic because base clocks are also a lot lower.

You dataset is missing the 4th value, the furmark value when running test condition #3. You note the performance hit to winrar but you did not capture the performance hit to furmark from the same test conditions.

And are you really capturing changes in "CPU calculating power" by using the Winrar benchmark as your CPU performance bench? Aren't you more likely to be assessing ram/IMC bandwidth congestion with this benchmark combo?

If I were to attempt to determine "CPU calculating power" while the IGP was doing work too then I think I'd run something like LinX w/1GB ram where you know the CPU is busy and then run either furmark or some other graphics benchmark that is known for not being dependent on the CPU.

 

[bgt]

the 4th value is nearly the same as the 1st so not really a change. Winrar just suffers when doing them both. Will try it with linx.

 

[ShintaiDK]

I am sure your numbers would also improve with Winrar, if you increased that 3770K memory to 1866 or higher as your 5700 system.

But again, winrar is a terrible way to measure it. Specially if you wish to test for a computational/throttle limit.

WinRAR is one of the benchmarks in our testing suite that benefits immensely from having faster memory. Moving from DDR3-1333 to DDR3-2400 speeds the process up by 20%, with the biggest gain moving from 1333 to 1600, and noticeable gains all the way up to 2133 C9.