Greetings again! Not long ago i did a Core i7 3770 HT test. According to that previous test Intel Ivy Bridge Core i7 processor offers no hyper-threading (HT) performance in single player gaming. On the contrary, HT turned on on the Core i7 3770 did hurt the performance slightly when compared to just the 4 physical cores of Core i7 3770. This time i've put the Intel Skylake Core i7 6700K to the shooting wall for "execution". This test can not be compared to any other tests i've made before, based on my profile, due to new NVIDIA drivers, updated game versions, different PC, some different game settings, some different testing methods. I've tested 21 game on 1920x1080 resolution with all graphical settings set on maximum, turned on, except: no anti-aliasing was used. Some exclusive NVIDIA features in games like Far Cry 4 and Witcher 3 were turned off. Physics effects were turned on, except for NVIDIA's exclusively based Physx effects, which were turned off. Some games have their own build-in benchmarks, while for others i used Fraps custom 15 seconds benchmarks. TEST SETUP Intel Core i7 6700K 4 - 4.2 GHz Asus Maximus 8 Ranger Kingston Hyperx Fury 2X8 GB DDR4 2133 MHz C14 Patriot Pyro 120 GB sata3 Windows drive WD Red 2 TB sata3 game drive Gigabyte GeForce GTX780 Ti GHz Edition 3 GB Windows 7 Pro 64 bit NVIDIA Forceware 361.43 VIDEO PRESENTATION
Personally, online games is where HT really offers the big difference. I did test my i7 950 a stock 3ghz with HT on vs my i5 4460 3.2 with turbo boost turned off on booth. In most single player games the cpu usuage was higher on the i5 vs the i7. Games still felt smoother but this was on games from 2013 and up. Older games, the HT does affect performance. But I have seen many newer games balancing the HT loads much better then they used too. This is all my personal experience and opion, But I much would rather have a i7 first gen then any i5 that isnt not a K model. And the ipc on a 4460 is much better then an i7 950 ipc for single core. But even a 950 @ 3.6 with HT beats on a 4460 quite nicely.
Thanks for testing this all, its really informative. You gotta switch to imgur though, photobucket is so awful.
Thanks for testing this, really interesting. Wouldn't have thought HT to give much bonus on most games, but honestly I expected it to not hurt performance by how long it's been around now.
I like your review! But tomb raider 2013 benchmark tool - your performance seems be not right, sorry... ComputerĀ“s specs are for my brother: - i7 4790k oced to 4.4GHz HT on - not 4.2 for all cores - corsair 2x8gb ddr3 2400 - gtx 780 ti sc - all stock speeds - asus z97-ar - win 10 x64 He got 82 fps min, 142 max and 108.2 average (camera moving not bug) at fullhd with ultimate preset and lastest update of 2013. Could be OSD enabled or video recording to affect your performance.
Really nice job, thank you for all this information!. I play both, single and MMO games so im not sure if i should try HT off sometimes, but i think i will.
We'll see. Then i can decide what CPU to get in my next upgrade. For now, i5 6600 probably beats any past gen i7's when it comes to gaming.
Thread update - i've included a video presentation. I really do not care! Perhaps my friend's Tomb Raider was not patched to the latest version - there is no other explanation other that that.. I am not some stupid idiot to record or do some background things while bench-marking, and am disappointed that you could even think so..:3eyes:
Thanks for these very useful benchmarks. My brother recently got himself an i7 5930K and HT was enabled by default. I showed him this thread and advised him to disable HT, which he did. He often plays 'Insurgency', a game based on the source engine. After disabling HT, the game has higher minimums. It's kinda hard to consistently test the same scenario, but there's a part with bots and it's pretty easy getting the game to dip below 50 fps (counter ticks as low as 46 fps) with HT enabled. With HT disabled, it's basically impossible to get that part to drop below 50 fps, even when really trying. I always thought HT was supposed to increase performance, not the other way around.
Well imo none of those games are really cpu multicore intense, except Crysis3, Hitman and WatchDogs. Skyrim apparently has a nice boost too by avg fps http://www.hardwarecanucks.com/foru...ntel-haswell-i7-4770k-i5-4670k-review-18.html http://ht4u.net/reviews/2013/intel_core_i7_4770_4670_haswell_cpus_test/index38.php But Idk about Watch Dogs by you, could be driver bug or something.. Look here 6600K falls behind 20fps vs 6700K http://www.hardware.fr/articles/940-17/cpu-jeux-3d-watch-dogs-total-war-rome-2.html http://www.purepc.pl/procesory/test_procesora_intel_core_i76700k_skylake_premiera_lga_1151?page=0,17 I also tested Hardreset a lot and it has a bit inconsistent lowest fps, one time min 82 the other 70 or so, it happens for a brief second at the beginning by car explosions. Im keeping HT on, well I always have been since I bought this cpu 2.5years ago and didn't see any bad fps drops yet.. Well nothing out of context to make me go and disable it.:nerd: EDIT: but then again maybe its also a OS thing, win7 is not super friendly to newer cpu architecture/multi threading, win8.1 or even win10 would be a better choice to test all those games.
Remember that with hyperthreading you are still using 4 physical cores. If the games already saturate the cores hypterthreading will do little to help because the core is already at full utilisation. However, you avoid switching between different threads so much because threads can run in parallel (even if each thread runs slower) which could be beneficial for performance. This could benefit certain tasks like online gaming if the thread priority is higher for the network side of things. The reason why it may be detrimental to performance is because as I said above, each thread runs slower and you have to remember thread prioritisation. If the right threads got prioritisation in this instance, you would probably find the hypterthreading performance not be a hinderance and potentially fractionally faster. This is just my view on it, it may not be correct . Having 8 physical cores will be different, especially if they have equivalent core performance to having 4 cores, since you aren't dividing 4 cores into 8 parallel threads limiting performance of each thread, the 8 threads can run on the full core. This won't mean twice the performance, it just means potentially twice the performance potential. This would get a little more difficult to quantify with more than 8 threads running since you still need to switch between the threads on that 8 core if more than 8 need to be run. So, if you are running 8 threads on a 4 core machine, each thread may only be useful at operating at say, 90 percent of the core potential. If running them in hyperthreading mode, you can better utilise the core performance up to 100 percent, and thus you benefit (as well avoiding any latencies, if present in switching between the two threads).
I don't have any driver bug or something. The same HT performance drop happened with Core i7 3770. Watch Dogs version 1.03.471. So there are 3 possible factors: Windows 7 HT under-performance or my Watch Dogs version 1.03.471 does not benefit HT performance, or is it a level dependent scenario where HT counts.. I hope you have not linked me a benchmark of online game, cause that's a different scenario... Well, Hard reset benchmark needs to be tested many times in a row to get the best minimal frame rate performance out of many instances - that is what i do with this game every time.
Min/max and avg fps only tell one part of the story. There is a reason so few reviews rely on only fps graphs anymore.
I don't think those comparisons are good enough. The differences between the 6600K and the 6700K are far more than just HT. The 6700K has a higher core clock and more L3 cache, which could also explain some of the gaps. Not in the case of Watch Dogs though, that difference seems too huge to be attributed to clock speed and L3 cache. Weird. Artas compares the same processor. EDIT: I'm wondering, is there some kind of profiler tool that allows us to enable/disable HT on the fly? It's a bummer that my bro has to reset his PC to enable HT when he plays a game that might actually benefit from it, and then disable HT again when he wants to play another game. Hassle.
No, but I believe there is a way to get around it . The first method would be to set the affinity in task manager details tab by right clicking on the game executable name and then selecting 'Set affinity'. By selecting CPU 0, 2, 4, 6 it should be in affect, non-HT. Alternatively, make a batch file with the following: @echo off start "game" /affinity 55 "game_executable" The game is the title, the game_executable is the name of the game executable . You can forego the game title (you can also set it to whatever you like), but if you delete it you can't use quotes around the program name as it will then think you are setting the window title. When run, this will start the game with the affinity already set. Setting the affinity to 55 means you are using cores 0, 2, 4, 6. The reason why it is 55 is because the affinity is set as hexadecimal. Basically you set the affinity as binary starting with the highest core (so the core order is 7, 6, 5, 4, 3, 2, 1, 0, remembering core 0 is the first core). So for cores 0, 2, 4, 6, the setting in binary is 01010101. Assuming you have Windows 10 etc, put this into Windows Calculator, set it to programmer mode, then click on 'bin' on the left. Type in 1010101 (you can't type the first 0, but this is fine), and then click hex. You will see it is 55. Now, if you start the game with this and check the affinity in task manager, you will see it says only cores 0, 2, 4, 6 are selected. Not sure if this actually works like this in reality in terms of giving non-HT performance in affected games, but I guess it's worth a try. There is a remote possibility that it could be fractionally faster, at least in theory. This is because non-game threads can still run parallel to the game threads, and maximise core use if the cores aren't maxed out.
Thanks for the suggestion! I'll ask my bro to try it out. Indeed, I'm afraid this method won't give the same results as having HT disabled, but we'll see.