Please explain why 4 modules should not be used in a 2x4 and 2x2 configuration. As long as the both dual channels are paired, I see no reason why the second dual channel should have the same size modules as the first dual channel. It is the frequency that needs to be matched across all modules not the size. The size only needs to match the module being paired to.
Indeed. I still have a 2500K overclocked to 4.8Ghz since day one and have the money to upgrade but I dont think it´s really worth it. I have since upgraded the rest: 8Gb of RAM, a 256Gb 960 Evo SSD and a Nvidia 1070. But the CPU is still holding on strong for everything, to be honest, and some articles and videos out there compare it to current CPUs and the difference is not that great, especially when OCed.
You lost gaming performance then, countless articles showed that in games, from time to time, having way to much RAM actually performed worse then machines with 8GB-16GB. It wasn't much of a loss, 3-10 FPS, give or take.
There are still good DDR3 systems that do support 2GB modules. I do understand where you are coming from now, but that is only true if you disregard DDR3.
I mean method 1 showed allocation and method 2 showed the performance impact. It's pretty clear for the latest and greatest titles you'll be right on the edge with 12GB's, certainly better than 8 but for some headroom you'll want 16GB's.
Ohh I wasn't questioning the article, it was the comment that had me confused. You cleared that up when you said DDR3 was off the table. I probably should have realized it was in reference to the article's DDR4 only but I didn't. Actually I didn't know DDR4 didn't come with 2GB modules either, so that was part of the confusion.I'm disregarding DDR3
If you find this article useless, especially based on the information provided in your own comment, then it’s your inability to comprehend the information in front of you that has been the point of fail.
Your son is 100% correct. Most modern multiplayer games such as Battlefield 1 stutter quite a lot with 8GB’s of memory and this is amplified on older systems such as his Haswell quad-core due to the lower memory speed. If he were to update to 16GB’s the experience would be much better.
You were hoping to find an article that would support your argument, unfortunately your wrong so you didn’t find what you were after.
MSI Afterburner with Riva Tuner and yes that was me playing with those memory sticks
What ZackL04 said.
R is don't tell someone they cant comprehend your article when its utterly awful and provides next to nothing with statistics and is poorly articulated.
There is multiple articles already stating MHZ of memory fail to yield anything more then a 1-5% increase in frame rate aside from ryzen due to the newer architecture.
If your going to rip on someone for speaking their mind on a comment then you should at least have the capacity to have a proper article with relevant information and not this nonsense posted. The article benchmarks show VRAM comparisons, not SYSTEM ram comparisons using dynamic VRAM testing samples instead of a static card and altering system memory limits which would correlate with the title. The author, I assume you also fails to have the understanding of how games allocate resources, the test themselves show useless results as the different VRAM buffers as shown in Call of Duty WWII show the 1080ti and 1060 have nearly identical results, well ****, where did the other 5GB of information go? you know all those textures that the TI had cached in VRAM. But you state "simular usage"
Hint, whenever your trying to quantify some "theory" you need analytical data that's not dynamic, that means not playing another map and being like whatever its close enough, I saw the numbers be similar. Yet what you posted is nothing close being a 5GB difference.
So lets break this down for you and hopefully you take the time to re-run the testing methodology if you care at all and would like to provide relevant information for your viewers.
1. Get a static setup of either VRAM or SYSTEM ram, a static benchmark then run the tests and invert the VRAM and SYSTEM ram configuration making the other the static module.
2. Realize that unless Shaders are forcefully stored in memory then your test is skewed
3. The only accurate piece of information is that when VRAM allocation is exhausted then system RAM and page file/hdd is utilized in that order for resources. I'm not going to bother speculating on specifics, given the system is running 32GB the testing methodology also needs to show specific resource utilization for the game being tested for page file as I highly doubt the game is using that over system ram. I'm not even sure why the that value is included unless the system is running 16gb or less on a card with 4gb of vram or less.
The TL
First off, you forgot to mention the brand name/model of your computer chair (hint hint). Secondly, haven't you heard that SLI is on its way out? A lot of new games don't and won't support it going forward.
Honestly I have 64GB of ram in my workstation having a 40 million poly model in 3ds max and zbrush 64 and photoshop open and premiere and I haven't run out of memory once ... 128gb could be good only if you are in scientific research or AI and you use huge datasets like calculating fluids , or the weather ...
