As a minor follow up to my previous review of Lenovo’s x230t tablet PC, I’ll be discussing the effects of using a dual channel memory configuration with integrated graphics. While the benefits should be apparent to some degree for all integrated GPU’s, this article will focus specifically on Intel’s newest Ivy Bridge graphics unit called the HD4000.
Here’s a video detailing showing the improvements detailed below.
Unfortunately Intel chose to encompass all of their graphics units integrated with Ivy Bridge under the HD4000 moniker.This is a problem because the performance can vary widely based on a number of things such as thermal constraints or arbitrary clock speed reduction. I would prefer a more detailed naming scheme similar to the way AMD or Nvidia name their products. I guess Intel can get away with it since they currently sell their chips based on the performance of their CPU component.
Integrated GPU’s are generally supposed to be budget oriented and therefor frequently don’t use dedicated memory. Instead they share the same memory as the rest of the system. This is usually a huge bottleneck for the GPU. By moving to a dual channel configuration, that restriction is halved causing the performance to increase greatly. Unfortunately the translation isn’t 100% efficient, but we should see gains in the range of 25%. Lets find out.
These benchmarks are just a superset of my previous article Lenovo X230T Tablet PC Review with Windows 8 RP. Please refer to that review for any system configuration or settings used. To recap, the x230t was shipped with one 4GB stick of memory. I purchased another Samsung Low Voltage 4GB stick to run the laptop in dual channel memory configuration.
Intel HD Graphics 4000 Mobile video card benchmark result – Intel Core i5-3320M Processor,LENOVO 3434CTO score: 4390 PCMarks
Intel HD Graphics 4000 Mobile video card benchmark result – Intel Core i5-3320M Processor,LENOVO 3434CTO score: 4745 PCMarks
As we can see, all area’s of the computer showed an improvement. However, the tasks pertaining to the GPU showed the largest gains as was expected. Going by the PCMark score, its an average of 7.5% increase.
Windows Experience Index
Window says the memory performance increases by 27% while the desktop and gaming graphics improve by 21% and 3% respectively.
In Civilization 5, the performance difference 23.3%. That’s a huge difference in framerate for such a small cost. Skyrim provides similar results with a 25.6% increase. That’s a 24.5% increase in gaming performance for a ~$25 extra initial investment. I don’t know about you, but I cant think of many other ways to accomplish that.
For reference, these are not hand picked games showing the most potential from this upgrade. They just happen to be the two games I enjoy playing at the moment. Civ 5 in particular is known to be CPU bound and run poorly on Intel hardware. But, with a simple and cheap upgrade we can at least bring the speed to an enjoyable level.
Someone brought it to my attention that part of the performance gain could be from the increase in memory size. I didn’t think about it too much since in my previous experiences it has usually not been the case. Its definitely a valid point and prompted me to dig further. So, here is the additional data I took to figure out if that was happening. Unfortunately I don’t have two 2gb soDIMMs @ 1600mhz to make a 100 percent definite comparison, but the data gathered should be sufficient.
First lets take a look at the memory usage of Skyrim using 8gb.
As we can see, total system memory used was only 2698MB. Lets call it 3000MB since it was creeping up. That leaves 1096MB for the GPU’s Use.
This was surprising to me, but this is what GPUz shows for memory use.
GPU memory used was only 257MB read from the largest part of the graph! I guess that’s what happens when the textures and settings are turned to low. I imagine some of the settings could be turned up without a major impact on performance.
Anyways, that means total system use was 3257MB if we use my nice inflated round system memory number. As I expected, the size increase of the memory had nothing to do with the performance gain in Skyrim and it was all from the move to dual channel.
Now lets take a look at Civilization 5 also using 8gb.
Ok, now were getting somewhere. System memory use during Civ 5 was at ~3625MB. That leaves us with 471MB for the GPU.
Well, Before the data went off the screen I was able to get a shot showing 434MB of use. I did find the highest point was 452MB, but it had already passed. Either way, its just under the 471MB that were still available for it to use.
Now Ill admit this is extremely close to the 4GB limit given in the single channel scenario, so lets take a stick of memory out and see if it reduces the memory that the GPU uses.
As we can see, total system memory usage went down to 3142MB leaving 954MB for the GPU.
At its highest point it only used 443MB this time. That pretty close to the amount that was used with 8GB of ram installed and I would call it within the margin of error. Not to mention there is still 511MB available to be used somewhere.
Making sense of it all.
Ok, from here on, its mostly going to be my opinion since I didn’t program the inner working of windows 8′s memory management nor do I have access to a Microsoft software engineer. Either way, your welcome to form your own opinions.
Lets take a look at system memory use directly AFTER closing the games. All other background programs and tasks are identical for all of these tests.
Here is Windows 8′s resource monitor using 8GB of memory.
It shows there is 1606MB of memory being used after closing the game and 278MB for the GPU (Hardware Reserved).
This is how it looked after the game using only 4GB of memory.
From here we can see that it appears Windows is being more aggressive in the conservation of its memory since there is less of it going around. The GPU still uses 278MB when nothing’s going on. The system only uses 1381MB this time which is 225MB less than when the system has 8GB.
If we add the extra 225MB windows decides to use when it has more memory to work with the 3142MB of memory used during gaming with 4GB. We are getting pretty close to the 3625MB use while gaming with the 8GB.
So, I guess my opinion is that Windows is more aggressive in reducing its memory use when there isn’t much to go around. Since we NEVER hit the 4GB mark, and there was an extra 511MB available during testing with 4GB, I’m going to call this one busted.
In this particular situation 4GB of memory was enough that it didn’t affect the benchmarks done at the beginning of this article.