With HT enabled, it will be approx 3-5fps slower in most games while it can help a lot when you are doing other tasks (like compressing or converting things for example).

The OC-ed 2500K is enough for everything if it comes to gaming (including streaming while playing), so an i7 in a gamer only rig is almost useless atm (perhaps it's better to spend the money on a stronger video card). But since you have it already , don't worry, HT is really great and it will be even a greater in win8 (you will probably feel it in games as well).
Be happy about your super CPU and use every single feature it has.

tl&dr: do not disable

i7 is definitely a waste of money when it comes to games. No games will take advantage of the hyper-threading of eight cores. i5 2500K is what people should buy :)

depends which games, Sup Com is better with i7.

oh definitely, but for the most part it's not really helping MOST games. Yes, some use it :)

*4 cores, 8 threads.

If you use your computer for actual important purposes (i.e. almost anything other than video games), then leave it on. But as you said, you play games 90% of the time -- but you should probably still leave it on regardless. The FPS loss is negligible if your graphics card isn't 10 years old. The only game in my recent memory that had a slight issue with hyper-threading is BF3, and that problem was quickly resolved for most people in a patch. I never turned off HT on my own i7 and never had a problem in BF3 either.

Be proud of the fact that you can buy one of the best processors and don't gimp it unless you absolutely must

Some benchmarks:

3DMark Vantage (HT on: 26072 | HT off: 23191 | % diff: +12.42%)

Street Fighter VI (HT on: 109.17 | HT off: 109.34 | % diff: -0.16%)

Resident Evil 5 (HT on: 140.7 | HT off: 136.6 | %diff: + 3.00%)

Borderlands (HT on: 77.4 | HT off: 76.88 | % diff: + 0.67%

Dirt2 (HT on: 115.63 | HT off: 116.63 | % diff: -0.86%)

Far Cry 2 (HT on: 130.97 | HT off: 123.49 | % diff: +06.05%)

Hawx (HT on: 208 | HT off: 192 | % diff: +08.33%)

Dawn of War 2 (HT on: 79.39 | HT off: 78.59 | % diff: +01.02%)

World In Conflict (HT on: 100.67 | HT off: 97 | % diff: +02.56%)

Crysis (HT on: 40 | HT off: 39 | % diff: +02.56%)

Batman Arkham Asylum (HT on: 196.67 | HT off: 200.67 | % diff: -00.99%)

Left 4 Dead 2 (HT on: 255.67 | HT off: 255.33 | %diff: +00.13%)

Overall performance gain with hyper-threading on: 2.9%

Test System:
DR3 1600 Mhz C9
Core i7 950@ 3.85 Ghz
Radeon 5870 Crossfire (overclocked - 900Mhz / 1250Mhz)
Windows 7 ultimate
All Tests were run at least three times with turbo mode off, and the results were averaged.
Resolution: 1920x1200 / All settings very high (4x aa)
Source: here

Comparing things with the "inferior" Bloomfield, when it's about HT efficiency with the Sandy Bridge architecture is not much of a help imo.

ps.: I didn't mean that those results are wrong ofc, just saying it's not about 2500k vs 2600K:)

What about the i3-2100 with its hyper threading and only 2 cores? What should be done with it for gaming ?

better buy a C2Q then.

could you provide a link please?

http://www.rue-montgallet.com/prix/comparer,p...2,1108=eq4

No, I meant a test where the C2Q ownz the SB-i3 in games, I was thinking about something to back up this:

"better buy a C2Q then"

(and let's just ignore the fact the the C2Q needs twice as much power and you can't replace it with an i5 later)

SB-i3 is a different story and HT helps a lot when you only have 2 cores.
This is because how the programmers design and compile the multithreaded code and how the threads are distributed at runtimes. Most games only use 2 or 3 cores (adding the fourth core usually means nothing). When you only have two cores while you would need a third one, HT comes really handy to get closer to a "3 cores" like performance.
That's why the i3 is hell of a good gamer CPU because it provides 2+ cores Sandy Bridge power for a low price.

edit: note that the thread distribution and management is completely redesigned in Win8, and people with HT enabled CPU will see improvements in multithreaded applications compared to win7 (iirc some games gained as much as 10%)

if you can find a game that is unplayable with an i7.... Then your video card sucks

or the game sux ;)

HT is good for multitasking. Video rendering etc.
When it comes to games it uses fewer cores but they work fine.

IF you just want your pc for games you don't need HT.
But if you do lots of things except for gaming and such as rendering and you want it finish everything quicker. I would recommend getting HT enabled processor.

Does hyper-threading really make much difference for rendering? Rendering is an easy task to completely occupy a processor with, with one heavy-weight thread per processor. Hyper-threading is only an advantage when you have significant processor stall so you can process other things while the processor waits for new data for the stalled thread.

you should re-read what HT is because stalled threads are never good. They actually run synchronous so as long as 2 threads don't have anything to do with eachother you can run them simultaneously. To re-synchronize them one of the processes will (most likely) have to be stalled though.

So in the worst case you can just put all windows processes on one core and all gaming related stuff on the other (if they are not threaded at all).

>you should re-read what HT is because stalled threads are never good.

I know what hyper-threading is. Some processes inevitably cause stalling and hyper-threading allows more efficient use of the processor when that happens. If a processor is already at 100% then hyper-threading will slow it down as you gain no resources while adding additional thread over-head.

"When execution resources would not be used by the current task in a processor without hyper-threading, and especially when the processor is stalled, a hyper-threading equipped processor can use those execution resources to execute another scheduled task. (The processor may stall due to a cache miss, branch misprediction, or data dependency.)"

http://en.wikipedia.org/wiki/Hyper-threading

kapca: "When it's done well, the HT optimized code will run faster on 8 logical cores compared to 4 physical cores only, because the duplicated architectural state of the CPU means more available resources."

That depends on the task. If it is a stall-heavy task then yes, hyper-threading will help. If it isn't then it will slow it down. Hyper-threading doesn't magically allow the processor to crunch more instructions per unit time. The duplicated state does not mean more available resources, only potentially more efficient use of what's already there.

dude you're making less and less sense... and I'm not even drunk.

That processes stall has very little to do with hyper threading!
You should read the first half of the 1st line on the wiki page and click on the link.

"Hyper-threading (officially Hyper-Threading Technology, and abbreviated HT Technology, HTT or HT) is Intel's term for its simultaneous multithreading....."

The link is.... simultaneous multithreading.

What you do is. You take a processor-core, clone it a bunch of times (not per definition a perfect clone but I think atm all are but I haven't read up on that) and slap a few of those on a single processor (usualy it's done in values of 2^x where is = 0 - 3 but some have done it differently). Assign a thread to each core. This way you can have multiple threads (or processes) execute tasks at the same time.

So for instance... If I write a program that would generate all prime numbers while in a different thread (of the same program) I'd write a piece of code that would solve the traveling salesman problem. This would run (aproximately) 2ce as slow on a single core processesor as it would on a multi-core processor.
Also if I would manage that I'd get 2 Fields medals in one year haha :)

You're confused, hyper-threading is not the same as a processor being physically multi-core. The OS sees a hyper-threaded core as being two physical cores but it isn't, it's still one. You can happily use multi-threaded code without needing hyper-threading and plenty of multi-core chips, such as a dual-core ARM A9 that you'd find in a Samsung Galaxy SII, don't have hyper-threading.

Another example for you: the Intel Atom processor 230 only have one processor core but they do have hyper-threading.

http://ark.intel.com/products/35635/Intel-Atom-Processor-230-(512K-Cache-1_60-GHz-533-MHz-FSB)

Cores: 1 Threads: 2

Again Hyper-threading is not just basic multi-core, multi-threading, it's a way of running two threads on a single core potentially more efficiently.

http://www.theregister.co.uk/2002/06/18/what_...threading/

"Hyper-Threading is a clever way of making a single chip operate like two separate devices without implementing two cores on one die."

k the slapping more cores on a cpu is a wrong description of it ;)
(I only write code so I don't actually have to deal with the hardware stuff).

But the rest of the post still remains valid ;).
Simultaneous multithreading has little to do with processes stalling.

"Simultaneous multithreading has little to do with processes stalling."

It does when you're talking about how you get more out of a processor with hyper-threading. It's fairly apparent that you have no idea what you're talking about.

nope, the thing HT does is that it allows the processor to be more fully used.
Aka when a logical comparison is done on 'one side of the processor' you can do a floating point calculation on 'the other side' (there are no real sides btw, but there are parts specifically designed for either function).

The duplicated architectural state (it's even in that shitty wiki btw) literally means that there are more stuff there (e.g.: duplicated registers) and not just something what's only emulated.

HT requires SMT support. When a code uses the CPU, it sets the Program Counter (PC) register where next sets of instructions are, this set is what defines the "thread", and the distribution, the order and the size of these threads are all factors what you can optimize for in your code.
HT and it's support by the hardwares and the softwares are changed a lot since the Pentium4. SB is not only alternate the execution stages between the two logical cores, but does a lot of other things to make ILP and TLP coexist more efficiently.

Today, you have the tools to design and optimize your code to detect and avoid thread performance problems (load imbalance, idle times, etc), and you also have the ways to optimize for HT. Most of the performance problems with the i7 are overheads btw, caused by inefficient taskswappings, because the code is divided into to many threads and chunks, and even the creation of those threads are not worth it sometimes, but that's something what you can fix or avoid with proper optimization.(this is why it's possible that game developers release a patch sometimes what fixes the slowdowns with i7 CPUs after they optimize the code for HT)

The bottom line is that (with Sandy Bridge), if you optimize your code properly for HT, it won't hurt the performance, but it's very important that you need to optimize for it, because what I said in #17.

"Rendering is an easy task to completely occupy a processor with"

When it comes to HT, it's not about how much load the CPU gets, but it's more about if the code is optimized for HT or not.
In the case of Sandy Bridge CPU (or with earlier implementations), some internal structures of the core are shared among the two logical threads (queues, buffers, scheduler, etc), and sometimes it's useless that you have duplicates of the registers and other parts, because if you have a buffer with 10 available "slots" and you turn HT on, each logical core will only get 5 of those, even if it's the running code is only single-threaded.
So you have to make sure that the two threads running on the same physical core are never stall because of waiting for each other. When it's done well, the HT optimized code will run faster on 8 logical cores compared to 4 physical cores only, because the duplicated architectural state of the CPU means more available resources.

ps1.: In the case of Bulldozer, AMD made dedicated integer pipelines for each logical cores, so when you run integer heavy calculations, Bulldozer will act more like how two physical cores would. (still useless against Sandy Bridge:D)

ps2.: Ivy bridge can dinamically allocate the resources according to the running code, so - for the first time since we have HT in CPUs - single threaded applications will fully use all the available resources in the core (would get all the 10 entries of the buffer in my example above).