dual core bottlenecking 560ti?

[Capt.Toilet]

Ok to start off, here are specs

AMD phenom II x2 555 @ 3.2ghz
7 gigs pc6400 RAM
EVGA geforce 560ti
600 Watt PSU with 48amps to 12v
750gig 7200rpm HDD

A while back I posted about ideas on some possible g-cards I was thinking of upgrading to. Well I decided on the 560ti, plus it came free with Crysis 2 so it wasn't a bad deal. I get home and get it installed no problems, uninstalled old drivers and got the latest drivers from Nvidias site. The performance is most of my games just seems to be a little off, yet at the same time I am getting better performance. Sounds weird but here are a few examples

-Metro2033 runs roughly 20's and 30's at full HD and dx11 settings. Lowering settings results in increased fps. Various benchmarks show roughly the same fps. Should be no bottleneck.

-Crysis 2 certain areas will dip into the upper 20's to mid 30's. Changing resolution and graphic settings only seems to net me about 10 extra fps. On the same token if I turn around 180 degrees in the same area, my fps will double. Both circumstances are densely populated areas. I will post screenies to show in a bit. Possible bottleneck with this game?

-The original Crysis still runs like mud. I even enabled 8x AA and my fps stayed the same. This was really troubling for me.

-Portal 2 enabling or disabling AA yielding lower and higher fps, like it should. Portal 2 seems to not be bottlenecked.

-Borderlands. Still receive massive fps drops when looking out towards Tartarus in the claptrap dlc. Overall fps seems to be improved though.

-Battlefield BC2 never seems to go above 50 and can dip into the 30's on occasion. I can confirm on my old card, enabling HBAO would result in some really bad mouse lag, with this card it does not.

I checked the control panel and saw this card has a power management thing that downclocks the card when not in use. I switched it from adaptive to full power always. Didn't make a difference.

GPU-Z shows the correct specs of the card and even shows the card increasing in temp when in use, so I know it is being utilized.

So any suggestions on what I should be checking for? Anything I might have missed in regards to specs?

 

[Sir_Brizz]

What driver version?

 

[Kyllian]

What mobo do you have?
Either way, if you can get a quad-core in there, you'll probably see a difference
Phenom II x2 is an AM3 socket, but AM2 boards are forward complatible at a performance cost

That is rather odd that Crysis was going so low on FPS as your specs are well above the recommended

 

[NRG]

7 gigs pc6400 RAM

Do you have an uneven amount of RAM sticks installed? There's almost no point in having an extra stick installed if it isn't operating in dual channel. In theory, the system will be slower if that stick comes into use.

-Metro2033 runs roughly 20's and 30's at full HD and dx11 settings. Lowering settings results in increased fps. Various benchmarks show roughly the same fps. Should be no bottleneck.

Sounds normal. You need two high-end cards in SLI/Crossfire to get decent FPS in Metro while playing in DX11, as you can see here. They're using a much faster system and yet they only pulled ~25 fps in Metro with a 560 Ti. (They probably used higher settings than you though, so they could have pulled off closer to something like 35 with that computer) Play Metro in DX9. It'll double your fps or more and the only visuals you'll sacrifice is tessellation, physx and AAA. None of which are noticeable in that game unless you stop to look for it.

-Crysis 2 certain areas will dip into the upper 20's to mid 30's. Changing resolution and graphic settings only seems to net me about 10 extra fps. On the same token if I turn around 180 degrees in the same area, my fps will double. Both circumstances are densely populated areas. I will post screenies to show in a bit. Possible bottleneck with this game?

Sounds a little disappointing yes, but not too far off where I'd expect it to be. Tom's Hardware shows the 560 Ti eating up Crysis 2 at lower settings, but struggling once you turn them up. (There has been driver optimization since this article however) They also tested on a Sandy Bridge computer which again, is a lot faster than your setup. The fact you can't pull better FPS at lower settings suggests there's a bottleneck, but at the same time your CPU is much slower so it's possible that there's less CPU limitation from your computer just being plain slower and resulting in less skyrocketing FPS on low settings.

-The original Crysis still runs like mud. I even enabled 8x AA and my fps stayed the same. This was really troubling for me.

Normal, for several reasons. AA in Crysis (especially if you select the AA from the in-game menus) wont hurt FPS drastically. This is just how good the new nvidia cards are at AA. Second, I'm sure it's definitely affecting FPS still. The in-engine FPS counter (r_displayinfo) is very in accurate for the same reason you don't notice the FPS dropping. It's all very technical what's happening when you lose FPS in Cryengine 2 but it's not as easy to notice as in most games. All you really see when FPS drop is movement become slower as opposed to everything becoming increasingly sluggish or jerky in a lot of games. When your FPS drop in Crysis, it's a lot like turning down the game speed (slomo) in Unreal Tournament. Thus, the easiest way to check is to do things like looking at the sky or ground and seeing if Nomad is suddenly running a lot faster.

-Portal 2 enabling or disabling AA yielding lower and higher fps, like it should. Portal 2 seems to not be bottlenecked.

Technically, you shouldn't notice a huge drop when using smaller amounts like 2x/4x. (less than 25% I'd say) Modern cards have just gotten that good at AA. Source is however one of the few engines left that let us use ridiculous amounts of true, driver-level AA. So if you're really cranking the AA settings then yeah, that should hurt.

-Borderlands. Still receive massive fps drops when looking out towards Tartarus in the claptrap dlc. Overall fps seems to be improved though.

I'd guess a similar scenario as Crysis 2: a mix of a small bottleneck and the rest of your system showing it's age.

-Battlefield BC2 never seems to go above 50 and can dip into the 30's on occasion. I can confirm on my old card, enabling HBAO would result in some really bad mouse lag, with this card it does not.

Welcome to the Frostbite engine. Your fps is going to suck regardless of your settings unless you have a sweet, well-rounded system. Got a bottleneck? Lowered settings? Mid-end to low-end video card? Too bad, your FPS is going to suck.

I checked the control panel and saw this card has a power management thing that downclocks the card when not in use. I switched it from adaptive to full power always. Didn't make a difference.

Unnecessary. The downclocks don't happen until your computer enters power save mode. You sure should it back on because it really does help the computer suck down less power.

So any suggestions on what I should be checking for? Anything I might have missed in regards to specs?

Motherboard? GPU bottlenecks often come from not just the CPU being too slow, but the platform being old and communication between the GPU and the CPU being too slow for newer video cards.

 

[Kyllian]

From what I hear, BC2 for PC is built off of the PS3 version, so you'll need a beefier CPU to get better framerates

 

[Capt.Toilet]

Do you have an uneven amount of RAM sticks installed? There's almost no point in having an extra stick installed if it isn't operating in dual channel. In theory, the system will be slower if that stick comes into use.

Sounds normal. You need two high-end cards in SLI/Crossfire to get decent FPS in Metro while playing in DX11, as you can see here. They're using a much faster system and yet they only pulled ~25 fps in Metro with a 560 Ti. (They probably used higher settings than you though, so they could have pulled off closer to something like 35 with that computer) Play Metro in DX9. It'll double your fps or more and the only visuals you'll sacrifice is tessellation, physx and AAA. None of which are noticeable in that game unless you stop to look for it.

Sounds a little disappointing yes, but not too far off where I'd expect it to be. Tom's Hardware shows the 560 Ti eating up Crysis 2 at lower settings, but struggling once you turn them up. (There has been driver optimization since this article however) They also tested on a Sandy Bridge computer which again, is a lot faster than your setup. The fact you can't pull better FPS at lower settings suggests there's a bottleneck, but at the same time your CPU is much slower so it's possible that there's less CPU limitation from your computer just being plain slower and resulting in less skyrocketing FPS on low settings.

Normal, for several reasons. AA in Crysis (especially if you select the AA from the in-game menus) wont hurt FPS drastically. This is just how good the new nvidia cards are at AA. Second, I'm sure it's definitely affecting FPS still. The in-engine FPS counter (r_displayinfo) is very in accurate for the same reason you don't notice the FPS dropping. It's all very technical what's happening when you lose FPS in Cryengine 2 but it's not as easy to notice as in most games. All you really see when FPS drop is movement become slower as opposed to everything becoming increasingly sluggish or jerky in a lot of games. When your FPS drop in Crysis, it's a lot like turning down the game speed (slomo) in Unreal Tournament. Thus, the easiest way to check is to do things like looking at the sky or ground and seeing if Nomad is suddenly running a lot faster.

Technically, you shouldn't notice a huge drop when using smaller amounts like 2x/4x. (less than 25% I'd say) Modern cards have just gotten that good at AA. Source is however one of the few engines left that let us use ridiculous amounts of true, driver-level AA. So if you're really cranking the AA settings then yeah, that should hurt.

I'd guess a similar scenario as Crysis 2: a mix of a small bottleneck and the rest of your system showing it's age.

Welcome to the Frostbite engine. Your fps is going to suck regardless of your settings unless you have a sweet, well-rounded system. Got a bottleneck? Lowered settings? Mid-end to low-end video card? Too bad, your FPS is going to suck.

Unnecessary. The downclocks don't happen until your computer enters power save mode. You sure should it back on because it really does help the computer suck down less power.

Motherboard? GPU bottlenecks often come from not just the CPU being too slow, but the platform being old and communication between the GPU and the CPU being too slow for newer video cards.

Ok answering your questions from top to bottom

-Memory I believe is running in dual channel. I attached a screenshot so you can take a look. When my computer I built back in 07 died, I was on a very tight budget so I bought the current prebuilt machine I have. If it is indeed running dual channel, I have no problem removing the odd stick. 6 gig should be plenty for gaming anyway.

-Sounds good about Metro2033. I was running Very high settings DX11 at 1920*1200.

-Yeah without AA I was reaching Sources 300fps cap, whereas with my 8800gt I wasn't.

-Borderlands and Crysis 2 I guess I will try a new proc when I get the chance, or remove that extra mem stick.

-I turned "adaptive" back on in the power settings of the card. One thing of note. Even with that setting off, GPU-z will downclock the card on the desktop. I monitored it while playing and it never downclocked so its all moot, just wanted to give you a heads up.

-Motherboard is an MSI-Aspen 780g AM2+. Manufactured back in '09 sometime it has a pci express 16x gen 2 slot, which according to google is another way of saying pci express 2.0. So motherboard shouldn't be considered old for this card I wouldn't think. Bios has been updated to support AM3 processors. Course I am stuck in dual channel mode, but I realized that when I got the system.

For Brizz and Kyllian. Driver 275.33. Drivers that came with the card were 266.xx but they yielded the same result.

 

[Bi()ha2arD]

my c2d e6850 @ 3.5ghz bottlenecked the **** out of my gtx470. esp on win7.

 

[Sir_Brizz]

The thing is, even bottlenecked, all of those games should run at top settings just fine. I have an ATI 5750 with a Q6600 and even I can run everything maxed out with no problems on almost any new game.

 

[Capt.Toilet]

So judging from the screenshot I posted, I am assuming my ram is running in dual channel correct? 4 slots filled 2x2x2x1 so would it be wise to remove that lone 1 gig and stick with 6?

 

[MrLoathsome]

Pretty sure that indicates your ram could be running in dual channel mode, but is
not. Yank that odd 1g out of there, and retest things.

 

[Bi()ha2arD]

dual channel only works with eitehr 2 sticks or 4 sticks and they have to be the same. Or at least the pairs have to be the same.

 

[NRG]

So judging from the screenshot I posted, I am assuming my ram is running in dual channel correct? 4 slots filled 2x2x2x1 so would it be wise to remove that lone 1 gig and stick with 6?

Yes, notice how CPUZ says "Unganged" for DC Mode. There is a lot of confusion over how identical the RAM sticks have to be to operate in dual channel. Obviously, the rule of thumb has always been to get identical sticks to ensure compatibility. However, it is possible to mix two different sticks of memory depending on the motherboard's memory controller's compatibility and if both sticks can operate stably at the same frequency/voltage/timings as suggested by wikipedia:

Modules rated at different speeds can be run in dual-channel mode, although the motherboard will then run all memory modules at the speed of the slowest module. Some motherboards, however, have compatibility issues with certain brands or models of memory when attempting to use them in dual-channel mode. For this reason, it is generally advised to use identical pairs of memory modules, which is why most memory manufacturers now sell "kits" of matched-pair DIMMs.

I've done that myself on both AMD and Intel platforms with great success. What I'm NOT sure about is if the memory size of the sticks can be different. People every where will tell you they have be, but I'm honestly not sure if that's entirely true. I've never tried, and the same wikipedia article suggests it's possible on some motherboards:

If the motherboard has two pairs of differently colored DIMM sockets (the colors indicate which bank they belong to, bank 0 or bank 1), then one can place a matched pair of memory modules in bank 0, but a different-capacity pair of modules in bank 1, as long as they are of the same speed. Using this scheme, a pair of 1 GB memory modules in bank 0 and a pair of matched 512 MB modules in bank 1 would be acceptable for dual-channel operation.

Basically, it's possible that that the 2GB and 1GB may be able to operate in dual channel. The fact CPUZ shows "unganged" suggests it's not. I believe you can change it somewhere in the BIOS. Remember though, if you remove the 1 GB stick, there will no longer be two pairs of memory, only one pair of two 2GB sticks and a lone 2GB stick. That last 2GB stick obviously wont be in dual channel mode without another stick in the corresponding DIMM slot.

To make things even more confusing, people frequently argue over whether "unganged" or "ganged" is faster. Ganged is only possible in dual channel mode AFAIK.

Ganged = A four-lane, one-way highway for data. (A lot of data can fit on the memory bus, but can't send/receive simultaneously)
Unganged = two highways with two lanes. (can send/receive at the same time, but not in as large quantities)

Personally, I'm pretty confident that ganged is faster for various technicalities. The whole unganged vs ganged thing is only on AMD platforms though. Intel does things differently.

So, after thinking about all this, I would honestly leave the 1GB stick in, because that's going to be faster than just having the third 2GB stick alone because that last 2GB stick obviously wont be operating in dual channel. If you really want to run ganged+dual channel mode, buy another 2GB stick to replace the 1GB stick or take out both the 2GB and 1GB sticks.

 

[Sir_Brizz]

"Dual channel" can work with three sticks.

 

[NRG]

The third stick wont be operating in dual channel though. In theory, any data being addressed in that third stick will be slower than data in the dual channel pair which is why I originally said to take out the 1 GB stick.

 

[Sir_Brizz]

Maybe they call it triple channel, but it's specific motherboards that require a matched triple of ram to work that way.

 

[NRG]

No AMD platforms are triple channel.

 

[Capt.Toilet]

Found this thread about the unganged mode

http://hardforum.com/showthread.php?t=1393322

But yeah i think I will leave the 1 gig in. I haven't run into anything pertaining to unstability, no crashing or anything of that nature. Could still be a possibility that my cpu is bottlenecking, but at 3.2ghz I figured it would have been clocked high enough?

 

[NRG]

I don't know how your proc compares to the E8400 though.

Pretty hard to compare. AMD platforms have always handled the AGP/PCIE bandwidth right on the processor. Before the i5 and i7's, Intel was doing that through the northbridge which is definitely slower. You can see how ridiculous the northbridges were getting for that kinda load on motherboards like the 780i and 790i and their huge heatsinks.

 

[Capt.Toilet]

What settings? The best way to run Crysis/Warhead is in DX9 mode with all the in game settings to 'High'. Autoexec in sunshafts, and the game will look and perform very well.

Even if you manually .ini enable all the "DX10 Very High" details in DX9 mode, it'll perform much better than in the actual DX10 mode.

I wouldn't expect to get comfortable framerates in Crysis with 8xAA enabled.

Just on a side note, a 3.0GHz E8400 will definitely bottleneck a GTX 560 Ti in Crysis - I just tested it a few days ago:

I don't know how your proc compares to the E8400 though.

I was running Very high settings 1920*1200. It felt like I was getting the exact same fps as my 8800gt was getting, which was around upper teens to mid 20's depending on situation.

On a side note, appears the newest 275.xx drivers have a display bug in Crysis 2 that causes your weapons to glow white only on rain type levels. I downgraded to the 266.xx drivers that came with the card and noticed my fps seemed to improve a bit more when lowering details. Lowest I have dipped so far on Extreme was lower 30's, but i usually hover around upper 40's to 50's, which seems to match the benchmark for this card.

I guess its 100% safe to say I have a per game bottleneck. Battlefield BC2 most definitely, and I am surprised Crysis would be. I thought that game was confirmed as only using 2 cores?

 

[Kantham]

Sounds like a similar problem I was/am having with my computer, and you even said how ****ed up it is.

So you have just upgraded your GPU right? It could be the motherboard if the benchmarks are showing your new card should run the game at much higher FPS.

In cases of mass disappointment, you could always RMA the card. Especially if you've ordered from Newegg. I've done it twice until I figured my motherboard is getting kung fu'd once I get a new pc.

PC hardware is a waste of time and resources the more we go on.