GTX 1080检测瓶颈...CPU还是RAM？
GTX 1080检测瓶颈...CPU还是RAM？
I recently switched from a GTX 970 to a GTX 1080 (MSI GTX 1080 Gaming 8G).
Even though early drivers were somewhat challenging, the overall performance has been satisfactory. However, this is the first time my GPU hasn’t been the main constraint in my setup.
I’ve been playing in 4K, and generally the experience has been smooth; but in certain multiplayer games with heavy player counts, I notice a brief drop from around 60 to about 45 frames per second, followed by a quick recovery to 60.
This pattern suggests a possible bottleneck, and I’m trying to figure out which component is causing it.
It might be related to GPU memory performance...
I’m currently using:
Mobo: Asus Maximus VII Impact
CPU: Intel i7 4790k, OC’d to 4.6ghz core (all cores—c states and speed step disabled) / 4.4ghz cache @ 1.20v each, w. HT enabled
RAM: G.Skill TridentX 16GB (8GB x2), 2400MHz w. CL 10 @ 1.65v
I’m monitoring everything with HWMonitor. My CPU never reaches 100% on any core or thread—maximum is around 70%.
The GPU is operating at full capacity, including core and VRAM (clock speeds are normal). I’m seeking advice on how to track VRAM usage more effectively.
I suspect this could be a RAM-related problem since usage stays under 30%, which would point to speed rather than capacity.
If I adjust the RAM timings, could I achieve 2800MHz?
Any suggestions on diagnosing the issue better?
Has anyone experienced insufficient RAM speeds for a 1080 at high load?
Dueprocessofflaw :
This happens in WoW / GW2 when I'm around 40 players, and another 40 join unexpectedly. A frame drops, then it resumes normally at 60.
The problem is likely that some new players arrived with models not already saved in the game or cache, causing delays while loading from HDD/SSD.
When the games you're using show performance issues, especially if they rely more on single-threaded tasks, consider turning off HT: threads. Competing threads on the same core can lead to slower performance.
Windows task manager can be misleading.
The activity of a single compute bound thread will be dispatched on all available threads.
MMO games typically are very compute intensive and depend on the single master core.
You are doing about as well as you can there.
It is not clear to me if windows can properly identify a slower hyperthread.
Disabling hyperthreading as suggested above might help.
Ram speed is not likely to make a difference. Here is info on haswell ram speed scaling:
http://www.anandtech.com/show/7364/memor...on-haswell
Are you using a SSD? sooner or later, a game will need to do autosaves or load textures from the storage drive.
A SSD can do this some 50x faster than a hard drive.
No, I don't believe it's causing the bottleneck. Which games are you playing? Some have a tendency to lose FPS even with different graphics cards. So, my advice would be to stop worrying about it. Simple!
There is a straightforward test:
Reduce the resolution (not related to quality or other settings) -> if your FPS stays consistent with higher resolution, your CPU has reached its limit.
Based on the description, it seems so.
Interaction between objects is typically handled by the CPU. The GPU only needs to render them.
😉
Recently I switched from a GTX 970 to a GTX 1080 (MSI GTX 1080 Gaming 8G). Early drivers were somewhat troublesome, but overall the performance has been satisfactory. However, this is the first time my GPU hasn’t been the main constraint in my setup.
I’ve been playing in 4K, and generally the experience is smooth; though in certain multiplayer sessions with heavy player counts, I notice a brief dip in FPS from around 60 to about 45, followed by a quick recovery to 60. This pattern suggests a possible bottleneck. I’m trying to figure out which component is causing the problem.
It might be related to GPU memory performance...
I’m using:
Mobo: Asus Maximus VII Impact
CPU: Intel i7 4790k, OC’d: 4.6ghz core (all cores—c states and speed step disabled) / 4.4ghz cache @ 1.20v each, w. HT enabled
RAM: G.Skill TridentX 16GB (8gb x2), 2400mhz w. CL 10 @ 1.65v
I’m monitoring everything via HWMonitor. My CPU never reaches 100% on any core or thread—maximum is around 70%. The GPU is at full capacity, with both core and VRAM usage at their limits (clock speeds are normal). I’m curious about VRAM usage; any recommendations for tools to track it?
I suspect this isn’t a CPU problem, since my CPU never hits 100% across any thread or core. If it were a CPU issue, based on current temperatures and voltages, I’d expect to see higher performance if I pushed the voltage slightly (I’m considering moving to 4.7 cores later).
My thoughts lean toward either RAM limitations—though usage stays under 30%—or GPU-related constraints. The sudden frame drops seem more like a memory bandwidth or VRAM bottleneck rather than a CPU limitation. It could also be tied to the 1080’s memory speed, especially if timings are not optimized.
I’m open to suggestions: have others found that 2400 MHz RAM is insufficient for the 1080 under load? I don’t think it’s my RAM. There are reviews suggesting that even quad-core CPUs like mine struggle to maintain minimum frame rates in certain games, especially when using multi-core processors. The performance drop can be abrupt and short, almost like a frame time issue.
It’s possible the memory bandwidth is the real issue, given the 1080’s reliance on VRAM. Adjusting RAM timings might help, potentially allowing 2800 MHz to be stable.
Any advice on diagnosing the problem further? Is there any evidence that others have experienced similar issues with this model and configuration?
And finally, does your budget allow for an upgrade? Would it be worthwhile to spend a few hundred dollars now for a noticeable improvement in gaming performance?
I'm not relying on Task Manager for tracking usage. It's a watered/boiled down report.
HWmonitor indicates no core or thread exceeds 70% capacity, but I plan to disable HT this weekend when I increase my core to 4.7, to observe any changes.
I see disabling HT can be beneficial here, though I'm worried about the core/cache/HT setup.
Currently, I use SSDs (slower, reliable Intel for OS / Samsung for games). They're Sata III, so it's possible I'm waiting on them... a NVME-based setup with PCIE 3x4 is next... but one upgrade at a time, correct?
It's comforting to think RAM isn't the issue. My OCD loved the mix of 2400mhz AND CL 10, so not needing to adjust timings to increase clocks will be nice.
This happens in WoW / GW2 when I'm around 40 players, and another 40 show up unexpectedly. A frame drops, then it recovers smoothly at 60.
With VSync off, framerates stay around 75-80, so it's not a GPU clock problem, though it might be related to VRAM.
I don't have the funds to upgrade my CPU atm.
Such an upgrade would mean replacing the mobo (chipset), RAM, etc.
Also, looking at the clock speed and voltages I'm receiving, I'm getting a solid 4790k (4.6 cores @ 1.20v / 4.4 cache @ 1.20v, with HT, meaning theoretically around 1.225v for 4.7 cores and 1.25+ for 4.8--), which is right near where Haswell hits its efficiency limit, so it would be a waste after about 1.5 years.
I plan to turn off HT and upgrade to 4.7 this weekend, then I'll share the outcome.
I don't have the funds to upgrade my CPU right now. Upgrading would mean getting a new motherboard (chipset), RAM, and more. Also, looking at the clock speed and voltages I'm seeing, I'm getting a solid 4790k (4.6 cores @ 1.20v / 4.4 cache @ 1.20v). That puts me around 1.225v for 4.7 cores and 1.25+ for 4.8—definitely near the efficiency limit of Haswell, so it would be a waste after about a year.
I plan to turn off HT and upgrade to 4.7 this weekend, then I'll share the results. A 100Mhz boost won't be obvious.
As you know, I already explained how to identify the system's bottleneck. Lower game resolution can help:
*FPS increases → GPU is limiting
*FPS stays steady → CPU is limiting
This happens because CPU workload doesn't change with resolution.*