Review Suggestion

[Guest]

Hello.

 

I've registered to the forums because I wanted to suggest making a review. So... straight to the point.

 

In regards to AMD's APUs and Intel's Intergrated graphics solutions (especially current gen), memory has become increasingly important in builds where the iGPU is either the only, or the main graphics subsystem. After lots of reasearch I've concluded that there don't seem to be any reviews making apple to apples performance comparisons of the various available memory modules. For example, AMD APU owners will be very interested in knowing which memory module is faster at 2400MHz. So a direct comparison of various memory modules (ex. Kingston Hyper X, Teamgroup Vulcan, G.Skill etc.) on a 7850k tested in a few games should yield reliable results based on which one can judge which memory module is faster.

 

Right now, you cannot really know what memory is faster until you buy it. Memory Latency and frequency set the bar but then again, the various memory modules have significant differences amongst them due to the usage of different chips and controllers.

 

On the other hand, Intel processors are starting to become an interesting choice in regards to the integrated graphics. They now pack slightly more SPUs and high frequencies even on lower end models like the i3 (although only mobile CPUs for now). That processor's desktop counterpart might make up for an interesting choice for ultra compact and silent systems. Hence the importance of memory performance in regards to the iGPU.

 

So the bottomline is that a test should be made, on the same system config (one for Intel and one for AMD), with various dual channel, dual DIMM memory modules of various manufacturers. For AMD I would suggesta test on the 2400MHz modules while for Intel, either 1600MHz or 1866MHz.

 

Thank you!

[VioDuskar]

who is this guy and why does he sound like he has a valid idea?

[Blebekblebek]

thing this kind of test has been done repetitively since SDRAM vs RAMBUS vs DDR
and it doesn't effective in real world application.

 

as for IGP performance, it still not very cost effective to get better RAM to OC instead save it and get add on GPU later

in most AAA games, it doesn't even matter when you can only play with 20fps or lower, 3-7fps boost isn't going to help you.

[Guest]

thing this kind of test has been done repetitively since SDRAM vs RAMBUS vs DDR

and it doesn't effective in real world application.

 

as for IGP performance, it still not very cost effective to get better RAM to OC instead save it and get add on GPU later

in most AAA games, it doesn't even matter when you can only play with 20fps or lower, 3-7fps boost isn't going to help you.

 

Ofcourse. I mostly refer to budget builds, HTPCs, compact builds (supermarket/commercial centers PCs running videos on large monitors, maybe 4k capabilities and some 3D is useful there) and semi-portable power efficient builds (car setups, camper setups and any other pimpy application).

 

Also, honestly, I'm very into budget SoC builds. I aknowledge that although my 7850k has inferior CPU performance, the main bottleneck is the memory itself. I would love to see a direct comparison which will probably prove very useful in the upcoming AMD AND Intel models. Don't forget that many users are not into AAA gaming and simply require a graphics solution for HD media playback and maybe some entry level gaming. iGPUs are generally sufficient for those tasks.

 

So the idea is that for example, one memory module is 5% more effective than another one and only costs like 15$ more while a dGPU solution comes at least 60$. If the iGPU is enough, one might want to invest is a little bit more juice. After some mild OC on my system I tell you, 50MHz of extra actual frequency on my modules yielded something like a ~5% avg fps improvement so I trully believe that when it comes down to memory bottlenecks, slight differences make for tangible differences in the user's experience.  

[Blebekblebek]

I understand what you talking about

I also fan of SoC my self, I've been trying to max out IGP for every system that I own (even my self playing games mostly on intel HD just because it can)

but again, the solution that you mention isn't really effective for AAA games

take a look at 7850k price, including motherboard I can get the cheapest G1820 + H81 Board + HD 7750/R7 250 for at least same price and crush performance 7850k (for gaming)

 

for HTPC, scrap the GPU and you still can enjoy HD playback without hitch.

my HTPC running on G1610, and my gaming machine only G2020, I want to build Kaveri system but I don't like the power consumption.

 

anyway, testing various brand/type/speed for 5% boost IGP performance isn't really attractive to most people, especially gamers when they only care about 60fps which is close to impossible to achieve with AAA games.

[Guest]

I understand what you talking about

I also fan of SoC my self, I've been trying to max out IGP for every system that I own (even my self playing games mostly on intel HD just because it can)

but again, the solution that you mention isn't really effective for AAA games

take a look at 7850k price, including motherboard I can get the cheapest G1820 + H81 Board + HD 7750/R7 250 for at least same price and crush performance 7850k (for gaming)

 

for HTPC, scrap the GPU and you still can enjoy HD playback without hitch.

my HTPC running on G1610, and my gaming machine only G2020, I want to build Kaveri system but I don't like the power consumption.

 

anyway, testing various brand/type/speed for 5% boost IGP performance isn't really attractive to most people, especially gamers when they only care about 60fps which is close to impossible to achieve with AAA games.

 

 

Yes, but I'm suggesting this with an eye to future developments. These currently small differences will become much more relevant in future DDR4 setups due to the increased bandwidth per pin of the newest memory. I believe that this whole test would be useful for some but most would appreciate its  technological aspect. 

 

Anyway, this is my opinion and "thirst" for optimization.