i3 6100 vs i5 4460

Coolnquiet_1

Commendable
Dec 11, 2016
2
0
1,510
I have a problem boys and i'm sure you can help me. I3 6100,ddr4 2133,gtx 1060 6gb/rx480 4gb or i5 4460,ddr3 1600, gtx 1060 3gb/rx470 4gb
An old platform like 4460 with ddr3 support or a new platform with only two cores which i think won't be enough in 2 years.I don't want to do an upgrade too quickly.
 
The new i5s aren't really any better than the Haswell models. If you don't want to upgrade soon, but going Skylake means you need to go down to an i3, than go for the Haswell i5-4460. DDR4 doesn't matter and the platform didn't change much otherwise.

If you can wait for a higher budget, then you'll always be able to get something better. If you can't wait, then go for the i5 and don't worry about anything. While Hyperthreading does offset the i5s advantage somewhat, the i5's advantage is still undeniably there and future upgrades sake doesn't matter since nothing supported by the LGA 1150 platform is much better than what the LGA 1151 platform has to offer. Even the high end i7s are almost identical in performance.

Even if you wanted the i3 for its clock speed, Haswell has i3s with the same performance that still save you money.
 
There isn't much performance difference between Haswell and Skylake. Intel uses a tick-tock update cycle, where they redesign the CPU one generation, then shrink its lithography the next. So even though it looks like they're two generations apart, from a design and construction standpoint they're just one generation apart. Consequently, the performance difference is slight, with Haswell actually beating Skylake in certain functions.

https://www.pugetsystems.com/labs/articles/Haswell-vs-Skylake-S-i7-4790K-vs-i7-6700K-641/

The obvious reasons for picking Skylake are lower power consumption and temperatures (mostly irrelevant on a desktop), a better integrated GPU (irrelevant since you're adding a dedicated GPU), USB-C support, and DDR4 memory support. The last one doesn't really make any significant performance difference in real-world tasks at this time. I just include it because it means you can take memory you buy for a system today, and potentially re-use it in a system you buy 5 years in the future.

The vast majority of VRAM is taken up by textures. You're probably not going to be doing 4k gaming on a 1060, so 3GB will probably be just fine since you can skip the 4k textures. That said, the price difference between the 3GB and 6GB version is probably trivial, so I would second the above advice to just save up a little more and get the 6GB version. A 1080p or 1440p game probably won't need more than 3GB (using 2k textures), but if you add a ton of texture mods I could see you exceeding 3GB.

The 2 core vs 4 core thing doesn't matter for some games, makes a difference for others. It depends on how well the game is coded to take advantage of parallel processing. Hyperthreading isn't really that useful in most real-world use cases. Hyperthreading lets you use unused parts of the core as if were a separate core. But here's the thing - if you need extra cores, it's probably because the cores you have are already being maxed out doing a particular operation. And if the extra virtual cores are trying to do the same operation in parallel, well you can't hyperthread it because those parts of the physical cores are already in use. It takes a very eclectic and parallelizable computing task to really take advantage of hyperthreading. Video rendering and certain compression algorithms are about the only real-world tasks where you'll see a big benefit from hyperthreading. So in the games where more cores help, the i5 will be a lot better off than the i3.
 
Actually, gaming takes great advantage of Hyper-Threading. The thing is that while a command is waiting on memory data, another command can execute and things like that. You can see as much as a 20% to 30% performance boost in raw performance and even more in some circumstances since many games stall with only two cores without Hyper-Threading because of the aforementioned reason. Check out Tom's several articles on the topic.

Regardless, as much as around 30% improvement can't touch a nearly 100% improvement from doubling the core count. While many games don't scale perfectly well, you can get a good percentage of that in most intensive modern games, especially modern FPS games and other largely multiplayer games. These tend to scale very well with more cores/threads. Prime examples are BF3 and BF4 with large maps and 32+ players.

Otherwise, I mostly agree. I'd argue that regardless of tick-tock, Intel hasn't made much progress since Sandy Bridge, but with that said you're still correct about Skylake compared to Haswell.