- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
i think it can only monitor it for nvidia cards
-
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.
getting low fps in games yet gpu's are only at about 50-60% usage Help :@
- Thread starter pinkpotato
- Start date
this is a example of mine
What is AMD VISION Engine Control Center? Mine says the same as, pinkpotato's on 11.2.
Soldato
- Joined
- 4 Nov 2002
- Posts
- 2,982
- Location
- England - Leeds
After reading your comments, so this is why basically the new graphics cards I see have 3 gig of ram to make up for these issues with people that want to use crossfire and 3 monitors?
So in a way you could just think of the ram on your graphics cards in that situation as overriding your system memory coz you said the vram on the graphics cards is faster then the DDR3 ram on our PC's so therfore is there any point buying loads of ram anymore on PC if we already got loads on graphics card?
Like maybe in future if we got 3 graphics cards with 3gig of ram on each then instead of buying 8gig or 16gig memory in our PC's we could just buy 4gig of memory? Or am I missing something?
So in a way you could just think of the ram on your graphics cards in that situation as overriding your system memory coz you said the vram on the graphics cards is faster then the DDR3 ram on our PC's so therfore is there any point buying loads of ram anymore on PC if we already got loads on graphics card?
Like maybe in future if we got 3 graphics cards with 3gig of ram on each then instead of buying 8gig or 16gig memory in our PC's we could just buy 4gig of memory? Or am I missing something?
vram is only for graphics.
system ram is for applications/OS, ect
even if u have 3 graphic cards with 3GB it'll still only be 3GB of graphics ram not 9GB
when ddr5 comes to system ram it'll be the same speed as ddr5 vram
system ram is for applications/OS, ect
even if u have 3 graphic cards with 3GB it'll still only be 3GB of graphics ram not 9GB
when ddr5 comes to system ram it'll be the same speed as ddr5 vram
Last edited:
- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
No, like it got said earlier you would only get the 3gb vram not 3x3gb because it only uses one cards worth
Tempted to buy a 6990 next year when the prices drop as that has 4 gb which would sort me out for a longgg time, i guess the 4gb actually works out as 4 gb or does it only use 2 gb because of the whole crossfire thing or is internal crossfire different?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
No, system ram is system ram, gpu ram is gpu ram. Theres zero interface which will let the system access gpu ram for its own use, and it would be slow.
GPU memory is incredibly fast, but very high latency, the problem is its on the card, across a slow interface pci-e bus through a chipset or a cpu pci-e controller.
A gpu on the same card has very fast access to this memory because its already across the pci-e bus, the system would need to go across a comparatively very slow pci-e bus to access that high latency memory and then send it back across a slow bus, it would bring about horrific performance.
When you load a game a large portion of that, if not the majority is loading graphics data off hdd's, passing it across the pci-e to the gpu, and then having it there ready to be used.
Take a 6970, it has 175gb/s bandwidth iirc, its insanely fast memory but in highly predictable work with very high latency. System memory has 10-20gb/s depending on the system(older dual channel vs newer triple channel) but its also very low latency for much more unpredictable work loads. System memory, cas latency anything from 6-9 on ddr3, gpu memory is a cas latency of 13-15 or so(ok, I made that up, thats what it was a while ago, probably not on gddr5, which could easily have gone up again by a large margin).
Now the issue being that pci-e connection is 16gb/s, so even if you did access gpu data, you'd only be using it at the same speed as system memory, but with far higher latency and a much more complicated and long winded path taking even longer.
Theres a reason gpu's have memory on the card, because the reverse would be true, if you had 32gb's of system memory in a few years with decent bandwidth, it would still be very slow across the pci-e bus. Localisation is the key to the speed in pretty much everything in computers. Remember a monolithic single chip cpu is faster than the same chip in two halfs on the same pcb, going "off die" and across the cpu package, mere millimetres away is magnitudes slower than going across cpu interconnects.
Intel's IGP's in the next couple years are having memory intergrated on die for this very reason despite it already having access to system memory, their igp's will gain insane bandwidth from it.
Anyway, your speed issues are almost certainly NOT from running out of vmem, from benchies on the web a single 6870 should get somewhere close to 40-50fps average at 3x1920x1080 eyefinity resolution.
When you run out of memory its usually a case of dropping from say 50fps to 0.5fps, its just the "in" answer to any problem at the moment, vmem is too low.
If you are getting minor hitching occasionally it can be when an engine goes over the memory limit briefly, if its over the memory limit consistantly, you'll be looking at sub 10fps most of the time, that isn't the case.
Two things, Cod is incredibly low in memory use, even in eyefinity, because frankly its a ruddy ruddy old engine now, its also not always the best in xfire for whatever reason.
You really haven't said enough about settings to know where the issue is, vsync on or off, in game or not, trip buffering on or off, aa/af settings.
Vsync is getting more and more problematic these days. Several games react very strangely to vsync dropping performance down WELL below where it should and could go.
Just Cause 2 did this on my single and xfire 5850 setup, on xfire with vsync I got low gpu usage, stuttering and frequent drops specifically when turning around. Without vsync I got circa 110fps average, and almost never below 80-90fps. Seen a few other games show this, GTA4, few others.
GPU memory is incredibly fast, but very high latency, the problem is its on the card, across a slow interface pci-e bus through a chipset or a cpu pci-e controller.
A gpu on the same card has very fast access to this memory because its already across the pci-e bus, the system would need to go across a comparatively very slow pci-e bus to access that high latency memory and then send it back across a slow bus, it would bring about horrific performance.
When you load a game a large portion of that, if not the majority is loading graphics data off hdd's, passing it across the pci-e to the gpu, and then having it there ready to be used.
Take a 6970, it has 175gb/s bandwidth iirc, its insanely fast memory but in highly predictable work with very high latency. System memory has 10-20gb/s depending on the system(older dual channel vs newer triple channel) but its also very low latency for much more unpredictable work loads. System memory, cas latency anything from 6-9 on ddr3, gpu memory is a cas latency of 13-15 or so(ok, I made that up, thats what it was a while ago, probably not on gddr5, which could easily have gone up again by a large margin).
Now the issue being that pci-e connection is 16gb/s, so even if you did access gpu data, you'd only be using it at the same speed as system memory, but with far higher latency and a much more complicated and long winded path taking even longer.
Theres a reason gpu's have memory on the card, because the reverse would be true, if you had 32gb's of system memory in a few years with decent bandwidth, it would still be very slow across the pci-e bus. Localisation is the key to the speed in pretty much everything in computers. Remember a monolithic single chip cpu is faster than the same chip in two halfs on the same pcb, going "off die" and across the cpu package, mere millimetres away is magnitudes slower than going across cpu interconnects.
Intel's IGP's in the next couple years are having memory intergrated on die for this very reason despite it already having access to system memory, their igp's will gain insane bandwidth from it.
Anyway, your speed issues are almost certainly NOT from running out of vmem, from benchies on the web a single 6870 should get somewhere close to 40-50fps average at 3x1920x1080 eyefinity resolution.
When you run out of memory its usually a case of dropping from say 50fps to 0.5fps, its just the "in" answer to any problem at the moment, vmem is too low.
If you are getting minor hitching occasionally it can be when an engine goes over the memory limit briefly, if its over the memory limit consistantly, you'll be looking at sub 10fps most of the time, that isn't the case.
Two things, Cod is incredibly low in memory use, even in eyefinity, because frankly its a ruddy ruddy old engine now, its also not always the best in xfire for whatever reason.
You really haven't said enough about settings to know where the issue is, vsync on or off, in game or not, trip buffering on or off, aa/af settings.
Vsync is getting more and more problematic these days. Several games react very strangely to vsync dropping performance down WELL below where it should and could go.
Just Cause 2 did this on my single and xfire 5850 setup, on xfire with vsync I got low gpu usage, stuttering and frequent drops specifically when turning around. Without vsync I got circa 110fps average, and almost never below 80-90fps. Seen a few other games show this, GTA4, few others.
- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
vsync off
triple buffereing disabled
4x AA via game settings
all settings in game maxed
also same thing is happening with company of heroes, i get like 20-40 fps yet gpus arent maxed.
I think your probably right about the old game engine idea. As i didnt think 3 x 1680x1050 was too huge really compared to some people running 3 2560x1600 monitors
Although i think i do run into i vram bottleneck when playing dirt2 with 8xmsaa as it drops to like 4 fps if that and i cant even navigate that goddam menu interface. Although a 2xmsaa is perfectly playable.
Also if i do run into a bottleneck from my vram in newer titles what do you suggest? I saw you wrote lower draw distances and turn down textures, and im guessing from my own experience turn down aa. Anything else?
triple buffereing disabled
4x AA via game settings
all settings in game maxed
also same thing is happening with company of heroes, i get like 20-40 fps yet gpus arent maxed.
I think your probably right about the old game engine idea. As i didnt think 3 x 1680x1050 was too huge really compared to some people running 3 2560x1600 monitors
Although i think i do run into i vram bottleneck when playing dirt2 with 8xmsaa as it drops to like 4 fps if that and i cant even navigate that goddam menu interface. Although a 2xmsaa is perfectly playable.
Also if i do run into a bottleneck from my vram in newer titles what do you suggest? I saw you wrote lower draw distances and turn down textures, and im guessing from my own experience turn down aa. Anything else?
Last edited:
yea, crysis onmy system at 2560 goes from a nice silky smooth fps to something like a frame every few seconds when I up the AA too high and run out of VRAM (like 2x to 4x or 4x to 8x, cant remember which)
BFBC2 on the other hand, 2x aa and I sit at 750 odd usage, 4xaa and its at 894/896 but still in the 40s fps wise, but does have stutter down to like 20fps when new things have to load, so it much be just on the edge or something.
- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
Same thing is still happening with cat 11.3. Tried crysis warhead earlier all settings at mainstream with no aa @ 5040x1050, at only like 40-50fps and my gpus were only at like 75-85% usage and this is a game that definately supports crossfire well. vsync triple buffering etc are definately off
Soldato
- Joined
- 6 Feb 2010
- Posts
- 14,583
I think it is still because of CPU bottleneck.
It's not so much that a i7 930 at 4.4GHz being not fast enough, but more to do with Crysis/Crysis Warhead are written to run as only dual-threaded (if I'm not mistaken), so the 3rd and 4th cores on a Quad-core CPU doesn't help in the game.
It's not so much that a i7 930 at 4.4GHz being not fast enough, but more to do with Crysis/Crysis Warhead are written to run as only dual-threaded (if I'm not mistaken), so the 3rd and 4th cores on a Quad-core CPU doesn't help in the game.
- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
No its not maxing any cores, ohwell :/
- Joined
- 28 Oct 2010
- Posts
- 299
- Location
- South east
Ports being the wrong speed? How can i check and what ports?