Video Card Memory Analysis: 256MB vs. 512MB Review
How many forums have you visited in the last month with threads discussing whether 512MB video cards are really necessary? Well, hopefully, this article will discuss Video Card Memory Analysis 256MB vs. 512MB Reviews in detail.
Simply put, the data that cannot be stored locally in VRAM is transferred to system memory, where you will incur large latency penalties across the board. Technically, the only penalty incurred should be fps losses. But a typical symptom you’ll probably encounter is stuttering caused by texture swapping from the system RAM to the VRAM. This is generally known as cache thrashing.
If you are interested in a graphic card Foxconn 8600 GTS Graphic Card will be a great option.
Overlook
This takes place because textures that need to be displayed on the screen obtain absolute priority. So if a texture needing to go on screen is currently being held in system RAM it will need to be moved to the video card ram before it can be displayed. As I said, technically, this should not happen. But inefficient memory management in many game engines guarantees the stuttering issue is a common one.
With AGP, there is an option within your motherboard’s bios that allows you to select how much system memory can be used by the GPU. This is known as the AGP aperture size. You can choose various options (64, 128, 256, etc.). But most people recommend you use one-quarter the capacity of your system memory.
With PCIe there is no such configuration left to the end user, it’s all handled automatically by the video card driver. From what I can see, its decision is sensitive to the amount of system memory you have available and will at maximum only ever double the amount of available memory.
I’m not fully sure this is the case as I don’t have 2x1GB of memory here to see what happens. I’m confident that a 256MB card will still be doubled to 512MB regardless and a 512MB card will have 1GB available. I’ll be able to confirm this in the next few days and will update this article with my findings.
Test setup and notes
The cards tested are a standard 256MB 7800 GT and Gainward’s 512MB 7800 GT. A typical 7800 GT straight from Nvidia with no bios modifications made by the AIB partner will contain a delta of 40 MHz.
If you have no idea what I just said, visit this page. Gainward’s card does not have any delta, if the core is clocked at 400 MHz, the geometry, vertex, and ROP clocks will all be 400 MHz, too. To solve this inconsistency I modified Gainward’s bios using NiBiTor, added a 40 MHz delta, and flashed the new bios to the card.
Obviously, Gainward’s card is also clocked much higher, so I brought the clocks down to standard 7800GT speeds. Both cards were clocked a 400/500 with a delta of 40 MHz and the memory timings are identical. With these modifications, there are no distinguishable differences between the two cards other than memory capacity.
The tests consist of the usual frame rate comparisons at various resolutions and image quality settings. But they are more geared towards evaluating what happens when vast amounts of memory are requested rather than looking at playable gaming settings.
I also decided to use an excellent little app called Video Memory Watcher which allowed me to watch how much texture memory is used in both the VRAM and system RAM during testing.
I’m not entirely sure how reliable this data is. But from my limited testing, the results obtained are repeatable and, in my opinion, reliable enough. Still, the memory usage results should be read with a pinch of salt, albeit an interesting pinch of salt.
To keep things consistent, Video Memory Watcher was not running when the frame rate benchmarks were run, just in case it caused a slowdown.
Half-Life 2: Lost Coast Results
Now we just max out the settings by adding HDR into the mix. Most people will want to run this game with these settings for maximum image quality. Why else would you spend £250 or more on a GPU?
You can clearly see that the only resolution that can run within a 256MB frame buffer is 1024×768. Anything higher and your going to be storing data in system RAM but does this matter? Looking at the results above. We get the first indication that a little texture swapping does very little, but a lot can cripple performance.
At 2048×1536 it’s using 434.2MB, and the 256MB card needs to put 143.54MB of texture data into the system RAM. Not only this, but the totals do not add up. VMW reports the 256Mb card is using a total frame buffer size of 388.23MB whereas the 512Mb card shows 434.2MB. Where has 45.97MB gone? Could it be that we’re seeing signs of cache thrashing?
The performance numbers would suggest this, and so would my subjective look at the time demo used. At specific times in the time demo, the 256MB card would crawl along, exhibiting typical stuttering caused by cache thrashing. I cannot confirm this is the case, but it looks this way to me.
Summarising all the tests, it looks as if 512MB clearly is necessary for this game but only really from a technical viewpoint. But we like those. 2048×1535 with everything on is unplayable on both cards. So I’d have to say that though 512MB is definitely faster it adds no real-world gains to this game.
Initial findings show that a little texture swapping has no real impact on performance and is nothing worth worrying about. Let’s see if the others games react differently, starting with Quake 4.
F.E.A.R. Results
I used the F.E.A.R benchmark test for these results. We see no changes between a 256Mb and 512Mb card with no AA/AF. The final test dips into system memory on the 256MB card but the amount is so little it should make no difference.
Adding 4xAA and 16x AF into the mix, we see minimal gains, though they are there and cannot be ignored. I was expecting F.E.A.R. to show the biggest difference out of all games on the test. At this point something didn’t feel right, so I decided to move to FRAP and played through part of a level.
The results are very different from what we gathered in the benchmark. To be honest, using the F.E.A.R benchmark for testing video cards is pointless and misleading. It does not reflect the stress caused by the game at all. Don’t forget that as I was just playing through a level you cannot do frame-by-frame analysis.
That’s a 90.7% increase! F.E.A.R. has shown itself to be the most demanding game for memory requirements out of all the games here. Add to this the intense use of pixel shader operations and you can easily see why it brings even the most high-end machines to their knees.
Overall Results
Overall these results give us indications that we are no more than at the beginning of the transition from relying heavily on texture operations to arithmetic operations. We can’t all breathe a sigh of relief that ALU operations will mean we never need to see a 1GB GPU, we will.
But it does show that memory requirements might slow down and only increase when they are really needed, thanks to pixel shaders. The future for video memory seems to be one of efficiency, but for the next few years I think we’ll see memory requirements continue as they always have done.
Conclusion
In conclusion, the difference between a video card with 256 MB of memory and one with 512 MB can affect performance, especially for more demanding games and applications. A video card with 512MB of memory generally provides better performance, allowing for higher-quality graphics settings, better frame rates, and smoother overall performance. However, you should note that the overall effect of video card memory depends on various factors, including the specific card and the system used. Therefore, it is essential to consider specific needs and usage scenarios before making a decision.
Also Read:
