Jump to content

SAAB340

Members
  • Content Count

    684
  • Donations

    $0.00 
  • Joined

  • Last visited

Everything posted by SAAB340

  1. Nice one! First result over 60FPS in FSXmark11. I knew iit would be you Hasse=) Congrats
  2. That's a huge difference cased just by the RAM. I guess FSXmark11 is almost 100% CPU limited with a Titan. I just love how you can throw hardware at FSX and it happily eat it all=) How high did the IMC in IvyBridge manage to overclock? And how does Haswells IMC perform under high overclock? Is it better, a lot worse or just slightly worse than Ivy? Ps, if you want high FSXmark11 numbers, figure out where you should exit the previous run to achieve a high FPS on the next one. The throttle setting you had when you exited the previous flight is the throttle setting the next flight will start at and changed by the autothrottle once the benchmark starts as FSX is not closed down between the benchmarks. This results in the plane covering slightly different ground during the different benchmarks and that causes the fairly large fluctuations between them... Or might that be called cheating? =) Ds
  3. This thread is an interesting one. It's always good to see what settings that are good for ATI users being discussed. Not that I'm one of them, but there are plenty of ATI users around. Looking forward to se what Dario has to day about the difference between the GTX580 and the 7950. FSXs so called "fibers" (or" other threads" that's a more correct name for them) can run on any core. They are not reserved to just the first core available. Where they are scheduled is decided by Windows. (Source Phil Taylor)
  4. 255 is a no-no affinity mask. Anything putting a t&t loader on the same physical core as the Main-thread is. Just because it shows activity in the task manager doesn't mean that it's good activity...
  5. To answer the OPs question. It depends what you are after when it comes to performance. My reccomendations are: If you value groundtexture loading and loadtimes more than FPS, leave HT on and use an apropriate affinity mask. 249, 241 and 244 are 3 good ones to try with. If you use your computer to other things that make good use of HT but wan't the best FPS and can live with a increased risk of stutters and lowerd FPS every now and again leave HT on and set affinitymask 84. If you want the highest FPS and least risk of stutters in FSX all the time turn HT off and use affinitymask 14, 13, 11 or 7. If you can't be bothered and just want to fly, leave HT on and let FSX use it's default affinitymask (85). It still does a good job in a lot of scenarios and in general does a slightly better job than HToff with default affinitymask (15). I personally use my computer set at stock speeds normally with HT on and have affinitymask 249 set in the .cfg. I don't have a rig exclusivly for FSX. If I know I'm gonna sim with a lot of photoscenery i overclock my CPU with HT on and leave my cfg untouched. If i know I'm gonna sim with complex aircraft and big airports I overclock my CPU with HT off and change the affinitymask to 13.
  6. 1. Correct. you always get more stutters and most of the time lower FPS. 2. Well, I guess the answer is actually no. You can lower the risk a lot by manually tweaking the affinity mask when you have HT on. But as soon as HT is activated in BIOS there's a 2nd virtual core exposed that will share the same physical CPU core as FSXs Main-Thread. The risk here is that Windows shedules something to run on that core, because if it does, it will steal CPU time from FSXs main-thread - 'colliding' with it. Yes, Windows could always shedule other threads (from both FSX and other applications) to run on the same physical core as FSXs main-thread even when HT is selected off in BIOS. But it's very very unlikely. Because FSXs main-thread will run at full core load all the time. As long as Windows has a core with lower load availible it will most likely shedule any other threads there. Why put more work to a core thats already occupied when there's other cores that have spare cycles? (Thats why we de-select a core in the Affinitymask even on a quad-core without HT to make sure Windows always has a core with low load on it to shedule other-threads on.) Having HT on in BIOS always exposes a Core, that shares physical core with FSXs main-thread, with no load on it that Windows might see as a god place to schedule other threads on. You can always reduce that risk by tweaking the affinitymask but the risk will always be there as long as HT is on in the BIOS.
  7. It's implemented in the way that FSX default affinitymask only assigns every other core when it detects a Hyperthreaded CPU. So you end up with the same number of texture&terrain loaders as you would on the same CPU with HT turned off. So on a single core with HT it won't assign any texture&terrain loader and the default affinitymask=1. The same as on a pure single-core. The performance with HT on is however greatly improved as Windows will shedule some threads that arn't FSX mainthread to run on the HyperThreaded core. If you were to treat it as a dual-core CPU by setting affinitymask=3 and in that way assign a texture&terrain loader on the HyperThreaded core you'll get a lot of stutters from thread collisions between the mainthread and the texture&terrain loader sharing the same physical core.
  8. This quote is one of the single most misinterpreted quote in FSX history when people try to say that FSX can't use Hyper treading. But you are getting it right. Thanks. The one thing to add is: Hyperthreading should be off to avoid the risk of thread collisions on CPUs with more than 2 cores. On Dual-core and Single-core with HyperThreading it's better to leave HT on. The default affinitymask in FSX will make sure we get very few thread collisions in these cases. Turning HT off in these cases will create even worse performance. When we're taking Quad-cores (or more) we can avoid most of these thread collisions by turning HT off. By also tweaking the affinitymask setting manually and de-selecting one core we can avoid these thread collisions almost entierly at a small penalty in texture loading and load time. FSX can still make good use of hyperthreading for better textureloading and load times. (That's what FSX is designed to use multi core for) But activating HT on a Quad-core allways exposes the risk of thread collissions . The risk of these can once again be reduced a lot by tweaking the affinitymask manually. The default affinitymask is not optimal here. Hardly surprising that the FSX code works this way given that HyperThreaded quad-cores and above didn't exist when the latest update with SP2 was released.
  9. Hi There. Cool video. As you are one of few people that uses an AMD FX processor may I ask you if you could do a quick test for me? It involves running FSXmark11 but with different affinity mask from the official one. I take it the tweaking tool will have assigned affinity mask 254 for your FSXmark11 test? I would be really pleased if you could try with non standard Affinity mask 9 and 3 as well. Don't worry if you don't have a clean install any more. As long as you use the FSXmark11 flight it'll be fine. I'm trying to see how AMDs 'module' design works within FSX in comparison to Intels Hyper threading.
  10. Thanks, for reproducing the Ivy results. We don't have that many yet and the interesting bit is how much faster IvyBridge is clock for clock Vs SandyBridge in FSX.
  11. Can I just add to my previous post. It's clearly shown that @ 28.8 Avg FPS were clearly limited by a x8 PCIe 2.0 bus in FSXmark11. Should that not mean that @ doubble that FPS we'll clearly be limited by a x16 PCIe 2.0 bus or a x8 PCIe 3.0 bus? The highest entries we have so far, are in that region. So would it not be a good idea to include the PCIe bus width/type you used when you report a result now?
  12. Hi, Can I just chime in with these official results confirming that in FSXmark11 1 HT on/off has no effect (It's just the clockspeed you get that will affect) 2 Your storage has no effect 3 Being limited by PCIe bus does affect the result HT: Off, Slow 5400rpm 2.5" HDD HT: On, Slow 5400rpm 2.5" HDD HT: Off, Fast 10000rpm 2.5" HDD HT: Off, Fast 10000rpm 2.5" HDD x8 PCIe 2.0 bus I don't have any official results to show but RAM speeds/timings do influence performance in FSX. It has the same effect as upping the CPU clockspeed so it's most noticable when we're only CPU limited. It doesn't do much if you just change it a little, but the memory controller in IvyBridge seems to be a very competent one enabling us to use a lot faster RAM. And increasing the bandwidth with actual speed is always better than increasing it by adding more channels. Looking through my old notes I can see a 2.6% improvement by running a single chanel RAM @ twise the speed of Dual Chanel while keeping the latency the same. They both have the same theoretical bandwidth. (1600Mhz Single Chanel C8 vs 800Mhz Dual Chanel C4) I can also see a 6.1% improvement by doubling the bandwidth with RAM speed while keeping latency the same. (1600Mhz Dual Chanel C8 vs 800Mhz Dual Chanel C4) But only 3.4% improvment by doubling the bandwith by going Dual Chanel instead of Single Chanel (1600Mhz Dual Chanel C8 vs 800Mhz Dual Chanel C4) I can also see a 2.5% improvement by tightning the timings from C9 to C7 @1528Mhz Dual Chanel. So it wouldn't be surprising to see the equivalent of going from 4.5 to 5.0 Ghz on Ivy bridge by going from a 1333Mhz C9 to a 2666Mhz C10 kit with the CPU staying at 4.5Ghz. But don't pay crazy money for an extremely high performance RAM kit. Go for sensible priced performance kits. Huge changes in RAM will give a small noticable improvement, but smaller changes might be measurable but they are so small so you won't notice it. Hardly worth loads of money.
  13. Are you sure you didn't change anything else as well? BP=0 should not have a negative impact.
×
×
  • Create New...