One of the heaviest CPU bottlenecked locations I've found in game is just outside the Gridania Adventurer's Guild, right on the planks, with the camera looking inside. I managed 13-15 fps on my X4 620 @ 3.12 Ghz, while my friend, standing in the same spot, but with a i5 750 @ 4ghz, barely maintained 30 fps. If this is "working as intended" then I don't think we're gonna be fully capped on fps for awhile.
Hey I've also got a gtx460 1GB although mine is from evga and my temps are around 71-74 C while playing ffxiv. I'm using the drivers from my evga disk as I've been too lazy to update them yet.
My driver version is forcerware 258.96 My card isn't OC'd at all and the default core speed is 720 MHz with the shader at 1440MHz.
I was pretty worried when I first saw the temps that high in afterburner and did some googling. It seems the temps of the cards vary quite a bit under load but we're not the only two people experiencing temps in the 70's with the card under full load.
Sorry for the long winded answer; hopefully someone can provide some more info about the temps. they're experiencing.
No cause to worry about temps that low.
You need to look at your per-core CPU utilization, not aggregate. Aggregate does not tell the whole story because if even a single core gets to around 90% utilization in XIV, your performance will be capped regardless of what any other core is doing. Most tools report an average across CPUs which hides spikes of 100% utilization on a single core.
CPU performance cap can be seen in this graph (i7 860 @ 2.8GHz, no HT, no turbo, 2 cores enabled):
http://www.xariana.com/Benchmark/Test1.png
Note the dive in framerate AND GPU utilization in the middle of the graph (Bearded Rock) while the cores remain pretty much pegged.
Compare that to the same i7 860 @ 4.0GHz with no HT, no turbo, 4 cores enabled:
http://www.xariana.com/Benchmark/Test5.png
The game is clearly making use of 4 cores, albeit 2 heavily and 2 more lightly. With the system at 4GHz, there is no GPU sag around the middle of the graph - instead we see CPU utilization on the rise without capping at 95-100% which allows the GPU to be pushed much harder, thus keeping the framerate from taking a nosedive.
FFXIV is pretty heavily CPU-bound. While it's nice to have that GTX480 or 5870, one also has to have a CPU capable of driving it to get the full benefit, otherwise it's just sitting around waiting for work to do.
On the bright side, I already have my eyes on the Intel i7 2600K which Intel managed to push to 4.9GHz on stock air cooling. Pair that with a Noctua NH-D14 with a 3-fan configuration and you're looking at 5GHz. Having had great luck with a GA-P55A-UD7, it looks like the GA-P67A-UD7 Sandy Bridge motherboard is in my future unless AMD can pull a rabbit out of their hat. I'm praying for a competitive AMD!
7100 on high with a 920 @ 4.0 with 1 5970 max over volt 900/1200 raid 0 2 x 10K rpm drives
7211 on high with a 920 @ 4.0 with 2 295 gtx 690/1180 60gb SSD (room was 61F lol)
ATI and Nvidia control panels were set to performance. I'm choosing to play on the 5970 rig. I think it's score will pick up after release and I need to push it more. Each rig had a single 295 gtx and was scoring 62xx on high. I guess I'm bottle necked cpu wise. Hopefuly the 980x will be under 600 by x-mas. I was going to sell the 295 gtx's but They are keeping up with the 480's and the 295's are BFG (fail)so I think I will keep them as reminders of a lifetime warranty. I went with xfx for the 5970, lets see how long their lifetime is.
Finally put together my new PC, overclocked both the CPU and Graphics card - could only get a mild overclock with the CPU (175 BCLK, around 3.85 GHz), but got an incredible overclock on my Graphics card (Core clock from 607 -> 850 MHz)
Running the benchmark on high settings both before and after the CPU overclock, I got a score of ~3300.
I'll let the screenshot of the post gfx-overclock speak for itself:
http://img541.imageshack.us/img541/7...vbenchmark.png
Installed the new Nvidia drivers, Score went from 3220 to 3806 on High in Windowed Mode
Specs: i7 920 @ 4GHz, 6GB DDR3, GTX 460 1GB @ 875/2215
EDIT: Played in game as well, performance seems to have improved. Playing @ 1920x1080, Shadows on High, 16xAF, 16xCSAA, DOF on, Resolution = window.
Getting 60FPS in Thanalan, 30-40 in Ul'dah, 45-55 by Aetherythe (although not very populated, only 12ppl ish here)
I noticed something earlier when I was playing a bit. Have you checked your shadows since installing 10.9 drivers Izzy? It seems like the shadows on my character's face are better than they were.
Spoiler: show
This may be a bad screenshot but I noticed when I was playing around in the character creator that I could see proper shadows being cast on my face from the strands of hair and decided to check in game. Sure enough it seems betterish?
edit: Forums are auto-resizing my picture. Right click > view picture to see a closer shot of the face and shadows.
I'll have to take a look at it. I turned them off when I finally got annoyed at how ugly they looked.
Can you force AA/AF now?
I'm wondering if something with my computer settings is wrong. I have a i7 OC'd to 3.3, 2x 5770s and 6 gigs of ram and I can only get about 2400 on high. I see some of you guys with much higher scores with very similar specs. I got the newest ATI drivers, thoughts?
2.4k sounds about right for a single card (it's what I get and I have same specs except CPU is at 3.8GHz), benchmark is windowed so doesn't support CF, and the .dll hack doesn't reflect actual boost you get from CF in-game (when it works that is, when it doesn't then you get same performance/worse as a single card).
would overclocking a cpu from 3.8 to 3.9 make a big difference? I could prolly push mine to 4mhz, but kind of lazy and not really convinced the difference from 3.84 to 4.0 would be worth the extra heat the cpu generates (its about 80c max on prime95 at 3.84). I'm using a Noctua D14
I tell myself if I run within Intel specifications it's all good.
80s seem kinda high for that speed with a Noctua D14, maybe need better TIM application, HSF installation, case airflow, etc. Not sure if you got a chip that needs higher voltages, a different stepped model or w/e.
As for the temp itself, 80 on a stress test is far as I'm willing to go for a 24/7 overclock.
If your running high ram speeds, much more VTT voltage is needed than normal. On my Mobo, it needs stupid high VTT voltage to run anything higher than 1800 DD3, so what I did was lower the multiplier on CPU, increase the BCLK, lower the ram divider to get a nice 1600 DDR3 speed along with a lower uncore which in turn lowered my VTT. I got a nice 5~ degree decrease in turn at the same clock speed for 4 Ghz on my i7 930.
I'm sure it's the TRIM, I'm just lazy for re-applying it. I had a guy do it before I learned how to do everything myself. And I almost passed out when he said "I just put some of my stuff over what was alrdy on it," when I had Arctic5 right next to him... that, and he didn't wipe the old stuff off (w/im prolly glad of because knowing him he would of used a rag or some shit lol).
It's not the case, its a Nvidia thermaltake case w/ 5 fans and really great airflow. And the case sits on wood.
I also think my i7 965 is like 1.4 voltage stock. Figured that was high and could be lowered, but don't really want to mess w/ voltage incase I fuck anything up
Very hard to fuck up a overclock as long as you take baby steps on voltage increases and decreases. Paper and Pencil is your best friend. Stay within intel specs and you're golden. Takes a long time to do though.
http://download.intel.com/design/pro...hts/320834.pdf <---- For i7 900s
I'm pretty much copying and pasting this from another forum but for the most part it's correct.
Error codes while running prime95, linpack, w/e.
0x101 = increase vcore
0x124 = increase/decrease vcore or QPI/VTT...have to test to see which one it is
0x0A = unstable RAM/IMC, increase QPI first, if that doesn't work increase vcore
0x1E = increase vcore
0x3B = increase vcore
0xD1 = QPI/VTT, increase/decrease as necessary
0x9C = QPI/VTT most likely, but increasing vcore has helped in some instances
0x50 = RAM timings/Frequency or uncore multi unstable, increase RAM voltage or adjust QPI/VTT, or lower uncore if you're higher than 2x
0x116 = Low IOH (NB) voltage, GPU issue (most common when running multi-GPU/overclocking GPU)
As for the TIM,
The "correct" way to apply AS5 on i7s is to actually tint both the Heat Sink, and CPU Chip by placing a small amount on both, then use a straight edge spread it out. Afterwards, you wipe it off with a coffee filter so a small film is left on the surfaces. Next is to use the line method across the middle of the chip, then finally install the HSF. Some people swear by using the pea method, but the line method is the instructions from the makers. (This is also assuming you actually clean the fucker off with 90%+ alcohol + coffee filter, or something comparable. :X )
So I updated Catalyst drivers from 10.4 to 10.9. It added 20 points to my 1080p benchmark score, while increasing idle temperature by 10 C... Ho-hum. Catalyst me no gusto.
Idle temps on all the recent CCC drivers will be higher as they increased the idle clocks (afaik it was to prevent crashes). You're lucky it's only 10c, mine increased by almost 20c from ~40 to 55~60c