First of all, it was not easy to test BF4. The Origin client software seems to be buggy, at least under Windows, and with SLI it seems to be even worse, see more details here.
Anyway, this post will overview the performance of the Y510p with I5-4200M and GT 750M SLI, compared to the Toshiba Qosmio X70 with an I7-4700MQ and GTX 770M in Battlefield 4. I will try to compared the performance of these laptops in single player and multiplayer modes.
True, these laptops have different GPU. The Y510p has the GT 750M SLI setup and the X70 has the GTX 770M, so what the hell do I think to myself when I compare them? Well, the answer is that I mainly try to figure out how the I5-4200M fairs in the new BF4 game, compared to the I7. We know that the BF4 engine could use more than two cores (source 1 and source 2), just like Crysis 3, but whether it comes into play in practice in such laptops and for what degree, remains unknown.
I would also recommend checking this list of benchmarks of BF4 by people on overclock.net, though it has only desktop parts currently and only rather high end ones.
|Lenovo Y510p||Toshiba Qosmio X70|
|CPU||I5-4200M, 2 / 4 (cores / threads), 2.4 - 3.4GHZ||I7-4700MQ, 4 / 8 (cores / threads), 2.4 - 3.4 GHZ|
|GPU||GT 750M SLI. core@967-1058MHZ (Turbo Boost), GDDR5@1250MHZ||GTX 770M|
|Drivers||Nvidia WHQL 331.65||Nvidia WHQL 331.65|
Battlefield 4, v184.108.40.206 was in use.
Battlefield 4 single player benchmark
|Model||Y510p, GT 750M GDDR5, I5-4200M||Toshiba X70, GTX 770M, I7-4700MQ|
|Ultra@1080p||18-34, AVG: 26||17-28, AVG: 25|
|Ultra, MSAAx2@1080p||22-37, AVG: 30||22-33, AVG: 27|
|Ultra, MSAAx2, PP=medium@1080p||24-38, AVG: 32||20-40, AVG: 28|
|High, MSAAx2@1080p||26-44, AVG: 33||27-44, AVG: 33|
No throttling took place in either of the laptops. The numbers are more or less on the same level and more importantly, we can’t say that the Y510p has any disadvantage compared to the X70 with its I7.
The game itself felt better with the X70, but it is because the y510p had problems of stuttering and artifacts, probably due to the SLI.
Lets look at CPU/GPU usage:
|Model||Y510p, GT 750M GDDR5, I5-4200M|
AVG total CPU usage / GPU usage
|Toshiba X70, GTX 770M, I7-4700MQ, AVG total CPU usage / GPU usage|
|Ultra@1080p||76% / 99%||34% / 99%|
|Ultra, MSAA off@1080p||76% / 99%||34% / 99%|
|High||79% / 99%||40% / 99%|
We see an interesting thing here. When I lower the settings, the CPU usage actually goes up. I don’t know if it some wrong testing method of mine that made it looks so, but I repeated this test several times and it was the same. These numbers might be the result of the CPU waiting less for the GPU – but why does it waits at all? maybe this is the Frostbite 3 3D engine implementation to avoid unnecessary cpu load or whatever. GPU is maxed out a 99%, at least according to HWInfo.
Second thing we see is that the Y510p is far more loaded than the X70. It is about as twice as loaded – inline with the fact the I7 has twice the cores and given I set the clocks to be the same (and no throttling occurred).
It would seem that an I7 is not necessary here, even for ‘ultra’ settings. However, I’m pretty sure that with a fastest GPU – maybe a GTX 680M – the I7 advantages will start show up. The cpu benchmarks I’ve seen (again – source 1 and source 2, for example) are all made with very fast GPU setups – Nvidia Titan and a Radeon R9 290X, both are way faster even than the GTX 780M with easy advantage of 80%-100% in performance, at least. So, this is not our case.
Battlefield 4 multiplayer benchmark
Multiplayer games sometimes pose much more stress on the CPU due to having a lot of players which need to be rendered and since BF4 single player is quite rubbish and it is widely known that the most enjoyable thing in the world is to frag your best friend, this is an important benchmark.
Unfortunately, I could barely test the Y510p in multiplayer mode. The issues described above were even worse and it took me a day just to get several readings. The only reliable results I have are for ultra settings, but they are spread over several maps:
|Model||Y510p, GT 750M GDDR5, I5-4200M||Toshiba X70, GTX 770M, I7-4700MQ|
|Ultra@1080p||AVG: ~30||AVG: ~30|
I’ve already canceled my order and got refunded – might test it when it will be less buggy. Anyway, the performance was quite similar. There might be some variations in the performance or differences for other settings, but I simply couldn’t make myself try and test it again after like 12 hours – not only it is buggy, but with Fraps it is even more so. I can assure you that I’ve played without Fraps and HwInfo and got these results over and over – some maps got even higher FPSs.
Lets look at CPU/GPU usage, several maps, mainly “Siege of Shanghai”
|Model||Y510p, GT 750M GDDR5, I5-4200M, AVG total CPU usage / GPU usage||Toshiba X70, GTX 770M, I7-4700MQ, AVG total CPU usage / GPU usage|
|Ultra@1080p||85% / 73%||55%-61% / 99%|
The results are strange. I guess that the Y510p CPU/GPU load tests were with some other map or a different players number – this specific data is messed up. However, we see something interesting in the load department – the CPU loads were considerably higher, especially in the Toshiba X70, under multiplayer, suggesting that indeed the multiplayer scenario is more CPU intensive, as we already knew it would probably be.
Anyway, I haven’t seen lower FPSs than the ones on the multiplayer, though the reason is unclear to me.
Analysis & conclusion
First of all, BF4 is quite playable on very high settings even with $900-$1000 gaming laptop (like these).
As for I7 vs I5. It is obvious that for these levels of gaming power, the I5 laptops have nothing to worry about and I would suggest getting an I5 laptop if the price is right (like the Y510p that is selling for $850 as I write these lines). It doesn’t mean that there is no advantage to the I7 CPU. newer 3D game engines llike Frostbite 3 can make a good use of at least one extra core. but it will come to importance with more powerful GPUs probably, or even more CPU-intensive scenes, for example, if all 64 players are concentrated in one place. These are rather rare cases, though.
The games are still more GPU-bounded in most cases and the core count, although obviously important, will not have a significant impact if you have at least 4 threads.
If we look at other benchmarks over the web (linked above) also support this claim and show that for the same core clocks, even an I3-3220 will be almost as fast as an I7-3960X with six cores (12 threads) or the much higher clocked FX-8350 with 8 cores. So really, the advantage of a mobile I7 over a mobile I5 won’t be that big anyway.
About being future proof. Buying a laptop is buying a package, less so buying a desktop or a desktop part. If you’ll pay another 15%-20% (or $100-$200) for the I7 version of the Y510p in question, you’ll get only minor advantage in rare cases while this money, while if you buy the I5 version for cheap, you’ll be able to sell it a year or two after for some reasonable price and get one of the new models, with much higher gaming performance, probably better battery performance, most surely much better screen and lower weight – you’d want to do it and it will cost you virtually the same like getting an I7 version now.
Other laptops, like the Clevo W230ST have a cheaper i7 upgrade ($75 at XoticPC) so you might as well get it if you really want.
Last point and final words. The Mantle and NVAPI APIs, if utilized (DICE said they will utilize Mantle, for example), will take considerable load off the CPU due to considerably lower complexity than going through DirectX or OpenGL. In those cases/games, an I5 CPU will be even more reasonable and it will get more so as more support for Mantle/NVAPI take place. That’s more future proof.
Finally, I’d suggest the I5 version of Y510p if the price difference is too big, like now – around $200. If you can get the I7 version of the Y510p or other gaming laptop you like, for a reasonable price, then go with it, since it has its advantages here and there. Remember that I’m talking about games only – if you need a laptop to run several VMs, then an I7 is probably the better choice.