• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Hyperthreading

Take a look at this bacon, a guy here has posted figures for BF3 HT on and off on his 2700k. Check the minimum's out...

http://www.overclock.net/t/1196856/...0-7970-7990-owners-thread/19620#post_18983056

If you're really interested then I'll do some benchmarks on mine later as well. i7 920 @ 3.8Ghz + single GTX 680.

While many games don't use it atm, there is a growing list that do. So if you can afford an i7 then imo having HT available certainly won't do you any harm.
 
Last edited:
Thank you for that Wucked :)

@ Martini sure his core for core performance is lower so I understand the HT is getting used then, but whats the difference between that and a modern i7's threads being used in modern games when paired with high end GPU's.
 
In those type of situations, it'll make a difference, look at the level and player count.
You'd probably be maxing out the physical cores almost.
 
We've seen times where the i5 has become "bottlenecked" because of its 4 core limit. An i7 I would hazard a guess to, provided the games do support the multi-threads like Crysis, BF3 etc, would do better in.

Would you disagree with that?
 
Bacon? said:
Thank you for that Wucked :)

@ Martini sure his core for core performance is lower so I understand the HT is getting used then, but whats the difference between that and a modern i7's threads being used in modern games when paired with high end GPU's.
Click to expand...

If the physical cores of a newer CPU are pushing that GPU to the maximum, and coasting along (Not being maxed out) why would the HT then make the performance any better? It's just redundant extra performance.

Bacon? said:
We've seen times where the i5 has become "bottlenecked" because of its 4 core limit. An i7 I would hazard a guess to, provided the games do support the multi-threads like Crysis, BF3 etc, would do better in.

Would you disagree with that?
Click to expand...

No, but I haven't said anything to the contrary, that's what I've been saying.
The i7 would perform better as the physical cores are maxed, and the logical cores can help as extra performance.

If you've got the spare GPU grunt that needs the extra CPU push, of course it'll show an improvement, as that's what I've been so called "speculating" about.
But say if you had a low end GPU set up, the i7 wouldn't make a difference, hence why I've been saying it depends entirely on the situation, such as GPU/CPU speed, etc.

In Crysis, you see the PD stuff being really good, it's rumoured to be down to the FMA instructions.
That would increase core for core performance, so on Haswell, even though Crysis can use the extra cores, with the IPC improvements and instruction set speed up, it could negate the i7's HT performance boost.
 
Last edited:
Ahhhh ok. Apologies I thought you were getting accross the i7 would be no benefit. And I know it would be no benefit while the i5 was not hitting 100%. But I see now you mean the same thing I mean.

Although thats interesting stuff on Crysis, I thought it did well on PD because it could max out all its cores, though I was surprised the i7 did not do better.
 
Martini1991 said:
No, but I haven't said anything to the contrary, that's what I've been saying.
The i7 would perform better as the physical cores are maxed, and the logical cores can help as extra performance.
Click to expand...

This is valid. If a game is decently threaded and pushing your physical cores then having HT will likely give you better performance. Otherwise it probably wont.

But its also reasonable that HT helps when there is additional load created by applications outside of a game. For example its not unusual for me to have BF3 running with teamspeak, some music or video, various web pages, livestream, MS Office, etc all running minimized, or on my second screen. These all create their own threads which need servicing. Which may explain why even on smaller maps and player numbers I experience better performance with HT on than off.
 
Like Wucked, i ran an i7 920, albeit at 4.2ghz. In bf3, compared to my current i5 3570k. Minimum frame rates were slightly better. Same gtx 670 gpu used on both rigs. The difference was pretty slight though, certainly not game breaking. User on here, (illuz) was running a 3570k with an almost identical setup to mine, when he went to sli 670's, the non ht 3570k bottlenecked the cards a bit, switching to a 3770k eliminated the bottleneck.
 
My minimums got a nice boost in battlefield 3 by switching from an i5 to an i7. I sold the i5 to pay for the i7. I ended paying about £85 for the i7 after the i5 sale.
 
LtMatt said:
My minimums got a nice boost in battlefield 3 by switching from an i5 to an i7. I sold the i5 to pay for the i7. I ended paying about £85 for the i7 after the i5 sale.
Click to expand...
Thats my plan as well, though ill probably end up going the secondhand route for a delidded 3770k. Ill be going sli 670 at some point so an i7 will help with that.
 
Warning: I am writing this response instead of doing actual work so it'll probably be overly-complicated and boring!

Like any multi-threaded programming approach, the performance benefits for a game (or any piece of software) are complicated to calculate.

I think it's important to start by getting what HyperThreading is straight in your mind.

As a simple example, if we say that a single core equates to 1 unit of performance, then a quad core CPU equates to 4 units of performance (this is VERY simplistic). HyperThreading is effectively a clever system of ensuring unused CPU capacity is used, currently by allowing each CPU core to operate on two threads at once, however this does not mean that a core suddenly equates to 2 units of performance, rather (depending on the application being run) it'll be closer to 1.5 units (again very simplistic).

Using this example then, if we take a quad core CPU with 2x HyperThreading, we have a CPU capable of 6 units of performance compared to 4 units if HyperThreading wasn't available.

When you program in parallel, you don't design for HyperThreading, you design for multi-threading. Therefore you detect that there are 8 concurrent threads available at once on our example system. Now if we had a non HT enabled CPU with 8 actual cores (therefore with 8 units of performance), a very well design multi-threaded application might be able to utilise about 6 units of the available performance at any one time (or 75%). Now if we take the same algorithm and run it on our 4 core HT enabled CPU, the application still executes 8 threads, however, the available performance units per thread is now less, therefore the application still utilises 75% of the available resource, however the underlying resource is less potent per percent, meaning the 4 core HT enabled CPU will probably perform at about the equivalent of a CPU with 6 actual cores.

What this does mean however is that the 4 core HT enabled CPU will be faster with the same algorithm than a 4 core CPU with no HT, so the technology definitely offers a performance benefit.

The problem though (as you can hopefully see) is that this all relies on the algorithm being run being well optimised for multi-threading, and unfortunately most software really isn't with scalability being the primary issue (i.e. the more threads you have the worse the performance improvement per thread is), so when you go from 4 threads on your non HT CPU to 8 threads on your HT enabled CPU, you don't get the ideal performance improvement that the extra 2 units of processing performance would promise, instead you get an ever diminishing return.
 
Gotta hold of for the time being though mate, still need to get a new ssd and change my ram first. Cant be doing with using a mechanical hdd as a primary drive.
 
tbh it depends on the games 99 percent dont use hyper threading to make a performance difference .even the ones that do use it only off a couple of fps over the ones that dont and if your minimums are above whats needed you paying for nothing.

crysis 3 2 fps difference maybe and bf3 2-5 fps difference . only games really worth mentioning that people actually play regular.


best thing is bf3 will be done in about 3-6 months total and be a grave yard and bf4 will be just a refined version of that . so id personally save your money and add that extra towards your better gpu where you will get a bigger fps different that the hyperthreading cpu.
 
Hyperthreading I noticed helps a lot for streaming/recording as well. The minimum framerate increase in BF3 is higher than 2-5 fps also.
 
benchmarked please ?

min,avg,max.

some people actually claim better fps bench marked with hyperthreading off in bf3.

Hyperthreading on ---------

Frames: 28457 - Time: 338553ms - Avg: 84.055 - Min: 51 - Max: 141

Frames: 28928 - Time: 344216ms - Avg: 84.040 - Min: 34 - Max: 142

Hyperthreading off ---------

Frames: 28084 - Time: 331798ms - Avg: 84.642 - Min: 59 - Max: 143

Frames: 28886 - Time: 333764ms - Avg: 86.546 - Min: 40 - Max: 144



also less stuttering.

http://www.tomshardware.com/reviews/battlefield-3-graphics-performance,3063-13.html
http://www.guru3d.com/articles-pages/battlefield-3-vga-and-cpu-performance-benchmark-test,3.html


also read xXxDieselxXx response here and how to check so you can see yourself in bf3

http://hardforum.com/showthread.php?p=1037945973

so just save money and put towards a better gpu or wait.
 
Last edited:
Don't have any benchmarks but when I used an i7 2700k over my 2500k my minimums in BF3 were 10-15 fps higher with a gtx680. I sold the 2700k for a 2500k as I didn't think the hyperthreading was benefiting me. As above you can see it helped LtMatt as well.
 
You must log in or register to reply here.
Back
Top Bottom