Thread Rating:
  • 2 Vote(s) - 3 Average
  • 1
  • 2
  • 3
  • 4
  • 5
i7 performance with and without hyperthreading.
#1
At first I didn't really think that hyperthreading made that big of a difference, and a lot of people agreed, but after running some tests I value the i7 a lot higher than I did ever before. The tests I ran were actual video game tests, and they were not synthetic benchmarks, so the results that I got are real life results that you guys can actually see when playing games or doing other things(Well not as of now, but in the future)

After running a few tests in a few games, mainly GTA 4, with only 1 core activated. And experimenting with different configurations, I can see that hyperthreading makes a HUGE difference in games. It's really underrated. It's so big of a difference that it could be the difference of a game not running at all to a game running with pretty good graphics(Even though not maxed out).

I tested GTA 4 mainly. I'm giving my GTA 4 results now, and the results for the other game in a later post. ow GTA 4 is patched with version 1.0.3.0 so most performance issues are gone. The only problem is that the shadows look a lot uglier when put to medium settings(but now we have the option to turn them off which looks pretty good), and a few minor graphical glitches.


I first played with everything at the minimum settings. Yes everything, that includes resolution. I ran the benchmark and I got an FPS average of 36, now the fps average in GTA 4 is about 10 fps higher than what you are actually getting in, plus there aren't many cars or pedestrians in the benchmark so when you play you'll probably receive worst performance. Now after the benchmark finished, I tried to load the actual game, I waited for 10 minutes, and the game never loaded so I never actually played to see what it would be like. Now, maybe if I had waited for a little longer, the game might have loaded up, but do you think that waiting for ten minutes everytime try to play is playable? I don't think so. Now this is with everything minimized out. Everything so there is just no way I could have tweaked the settings for it to load.

This was the total report I got

Statistics
Average FPS: 36.69
Duration: 37.29 sec
CPU Usage: 99%
System memory usage: 58%
Video memory usage: 54%

Graphics Settings
Video Mode: 800 x 600 (60 Hz)
Texture Quality: Low
Texture Filter Quality: Low
View Distance: 1
Detail Distance: 1

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: benchmark.cli

Now just to test the benchmark I jack the settings up to what I normally play with

All in all the settings were either maxed out, or one setting away from being maxed out. The view distance slider is at 34, the detail distance slider is maxed, the traffic density is at 33% or 100% it's one of those since thats what I alternate between the two. 33% is what the console versions use. I also have shadow density at 1 or 16 depending on how I feel.

These are the results I got.

Statistics
Average FPS: 20.92
Duration: 36.91 sec
CPU Usage: 100%
System memory usage: 62%
Video memory usage: 93%

Graphics Settings
Video Mode: 1600 x 900 (60 Hz)
Texture Quality: Medium
Texture Filter Quality: Highest
View Distance: 34
Detail Distance: 100

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: benchmark.cli

Yes 20 fps. And thats what thw game says. You knows how low it really is. And of course the game never loaded.

Now lets look at the results that I get with hyperthreading enabled, with just one core. With hyperthreading I have 2 virtual cores, but only one actual core is activated. A lot of people say that it doesn't improve performance Because you are still on just one core. Even I believed that at one point.

My first benchmark was with everything minimized. I didn't think that hyperthreading would do anything since I had gotten some less than perfect results in other games when I tested earlier.

This is what I got.

Statistics
Average FPS: 52.12
Duration: 37.30 sec
CPU Usage: 96%
System memory usage: 60%
Video memory usage: 54%

Graphics Settings
Video Mode: 800 x 600 (60 Hz)
Texture Quality: Low
Texture Filter Quality: Low
View Distance: 1
Detail Distance: 1

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: benchmark.cli

Yes 52 FPS on one core(with hyperthreading), in GTA 4. 52 FPS. It's just amazing that I got these types of frames. Even though hyperthreading is active, it's still just one lone core. So to get 52 FPS was a big surprise. So big of a surprise that I just had to check that again. With the second benchmark and everything on the same settings, I still got 50 FPS so no this wasn't a fluke. Now this would mean nothing if the game couldn't load so I tried playing an actual game in it. The game started pretty quickly. It wasn't ultra ultra fast, but you'd have to be an absolute liar or very unpatient to say that it loaded even just a little long. When I played the game, The game was actually running well. There weren't any problems with textures missing, no texture pop ups, no stuttering. Well there was a stutter here or there but it's really minor and it barely happened anyway. I played for a long time trying to see what the maximums I could get were and the minimums. I was able to get to the 70s and maybe even lower 80s when inside certain buildings, when out and about I normally got the 40s. and in extreme extreme action(100% traffic density in times square while blowing up every car in the rode with a granade or rocket launcher while getting in shootouts with the police and swat team) I got normally upper to mid 20s, I did get to 17 or something though, but it was never unplayable. All in all it was pretty smooth. The framerate were never a real problem when actually playing the game, and the game was definitely very playable as far as framerate goes. I could go to my native resolution of 1600 x 900 without much of a performance hit as well, and that makes a very BIG difference in the game. It actually makes the game look decent when you put it at that resolution and use the right brightness settings.

Now to test how far I could go I put the graphical settings back up to where I normally play at This is what I got. I did 2 benchmarks with the same settings.

Statistics
Average FPS: 29.81
Duration: 37.14 sec
CPU Usage: 99%
System memory usage: 65%
Video memory usage: 81%

Graphics Settings
Video Mode: 1600 x 900 (60 Hz)
Texture Quality: Medium
Texture Filter Quality: Highest
View Distance: 34
Detail Distance: 100

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: Benchmark.cli

Statistics
Average FPS: 30.19
Duration: 36.90 sec
CPU Usage: 97%
System memory usage: 57%
Video memory usage: 93%

Graphics Settings
Video Mode: 1600 x 900 (60 Hz)
Texture Quality: Medium
Texture Filter Quality: Highest
View Distance: 34
Detail Distance: 100

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: benchmark.cli

The game had a choppy framerate once you got ingame. It looked beautiful, but it was choppy. It was playable, but still choppy. the lowest framerate I got was 8fps, the highest was 40 something. It normally stayed in the upper teens though. Despite that, I still considered it playable. If that was the only way the only frame rate that I could play it on, then I would have played it like that. Luckily, you can still lower the settings. The settings I had are still very high all across the boards. The only thing that I have that said medium was the textures, and medium is actually the second highest setting for textures since the highest setting for textures is high. You could easily lower some settings and get some visuals that are decent with decent framerates. The detail distance and traffic density really stresses the cpu so lowering it gives some fast results. Besides, not many people even like playing with 100 percent traffic density, and I don't even notice the increase in detail density once you get to the 20s or 30s. And the consoles had a detail distance somewhere around 15 so its all good.
You can still get decent graphics with decent framerates.


Now here are my benchmarks with two physical cores activated and no hyperthreading enabled.

First a benchmark with everything minimized.

Statistics
Average FPS: 64.12
Duration: 37.02 sec
CPU Usage: 90%
System memory usage: 63%
Video memory usage: 58%

Graphics Settings
Video Mode: 800 x 600 (60 Hz)
Texture Quality: Low
Texture Filter Quality: Low
View Distance: 1
Detail Distance: 1

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: Benchmark.cli

64 FPS. That's a big difference from the 52 fps we got with just hyperthreading on one core. The game ran very fluid. It ran more fluid than when I play GTA 4 normally. When I play with my settings, I usually get lower 54 fps from the becnhmark even though I use 4 cores with hyperthreading which makes 8 logical cores. The game was very fluid. No choppiness, no pop ups nothing. Now lets see what it is with my normal settings.

Statistics
Average FPS: 46.72
Duration: 37.84 sec
CPU Usage: 96%
System memory usage: 61%
Video memory usage: 94%

Graphics Settings
Video Mode: 1600 x 900 (60 Hz)
Texture Quality: Medium
Texture Filter Quality: Highest
View Distance: 34
Detail Distance: 100

Hardware
Microsoft® Windows Vista" Home Premium
Service Pack 1
Video Adapter: ATI Radeon HD 4800 Series
Video Driver version: 8.14.10.655
Audio Adapter: Speakers (Realtek High Definition Audio)
Intel® Core™ i7 CPU 920 @ 2.67GHz

File ID: benchmark.cli

46 fps according to the benchmark. The game was very playable. And very fluid. You don't notice anything with the frame rate unless you use fraps.

So all in all even though physical cores are still of course much better than simply using hyperthreading. There is still a very big difference when using hyperthreading. Hyperthreading however doesn't add nearly as much heat as a whole new physical core. At stock clock rates, there really is no difference in heat produced, and the is probably no difference in wattage use. It's kinda like a free upgrade, It also doesn't use the extra space in a cpu that a physical core uses. As of now we won't really see much of a difference in performance with hyperthreading since there really aren't many programs that use 8 threads, but later on it could be the difference between not running a game at all, and running the game with decent graphics.

I’m thinking about doing some tests experimenting with pcsx2 as well
Reply

Sponsored links





Users browsing this thread: 1 Guest(s)