-
Posts
150 -
Joined
-
Last visited
-
Days Won
3
Content Type
Profiles
Forums
Downloads
Posts posted by bloodhawk
-
-
1 minute ago, Tech Inferno Fan said:
Great. Looking lower but not x4 2.0 levels. Under the same load conditions, can you re-run CUDA-Z on the NGFF.M2 32Gbps interface?
I actually did last night, it was the same as posted in the OP, +/- 75-100. Even this hit around 2200 MiB/s.
Ill post a screen once i connect it over M.2 in about 20 mins.
-
Here is the CUDA-Z with GTX 1080 hooked up over TB3 (DSL6540 - Dual Port) to a P870DM-G (DSL6340 - Single Port)
Output :
SpoilerCUDA-Z Report
=============
Version: 0.10.251 64 bit http://cuda-z.sf.net/
OS Version: Windows x86 6.2.9200
Driver Version: 375.70
Driver Dll Version: 8.0 (6.14.13.7570)
Runtime Dll Version: 6.50Core Information
----------------
Name: GeForce GTX 1080
Compute Capability: 6.1
Clock Rate: 1733.5 MHz
PCI Location: 0:10:0
Multiprocessors: 20
Threads Per Multiproc.: 2048
Warp Size: 32
Regs Per Block: 65536
Threads Per Block: 1024
Threads Dimensions: 1024 x 1024 x 64
Grid Dimensions: 2147483647 x 65535 x 65535
Watchdog Enabled: Yes
Integrated GPU: No
Concurrent Kernels: Yes
Compute Mode: Default
Stream Priorities: YesMemory Information
------------------
Total Global: 8192 MiB
Bus Width: 256 bits
Clock Rate: 5005 MHz
Error Correction: No
L2 Cache Size: 48 KiB
Shared Per Block: 48 KiB
Pitch: 2048 MiB
Total Constant: 64 KiB
Texture Alignment: 512 B
Texture 1D Size: 131072
Texture 2D Size: 131072 x 65536
Texture 3D Size: 16384 x 16384 x 16384
GPU Overlap: Yes
Map Host Memory: Yes
Unified Addressing: Yes
Async Engine: Yes, BidirectionalPerformance Information
-----------------------
Memory Copy
Host Pinned to Device: 2147.95 MiB/s
Host Pageable to Device: 1919.76 MiB/s
Device to Host Pinned: 2678.09 MiB/s
Device to Host Pageable: 2304.06 MiB/s
Device to Device: 114.298 GiB/s
GPU Core Performance
Single-precision Float: 8701.88 Gflop/s
Double-precision Float: 284.502 Gflop/s
64-bit Integer: 470.519 Giop/s
32-bit Integer: 2619.8 Giop/s
24-bit Integer: 2016.1 Giop/sGenerated: Mon Oct 31 21:09:50 2016
- 1
-
-
Enable Hot Plugging under PCIe settings in the BIOS.
-
7 hours ago, benjaminlsr said:
Well maybe the issue only/mostly appears with small form factor computer (NUC6i7KYK). The issue is realy "gone" since I switched to ethernet and then to 5Ghz.
I've read this information somewhere online on a razer core discussion having disconnection issues.
Gotcha. Ill do some more testing tonight and see if it shows up on my end.
-
4 hours ago, Tech Inferno Fan said:
@bloodhawk, you wouldn't still have the TB3 enclosure? We need some CUDA-Z output to confirm if it's providing 20Gbps or 32Gbps. Without CUDA-Z your noticably faster 3dmark result using NGFF.M2 compared to TB3 favoring would suggest your TB3 interface was 20Gbps.
See the discussion RE: TB3 bandwidth at
I still do have it. Ill try to install it tonight and test it out.
-
7 minutes ago, rbut said:
Great work @GTANAdam and @bloodhawk.
I did read somewhere that Microsoft implemented Hot Plug for Thunderbolt 3 in Windows 10, so maybe that's why it works for you bloodhawk?
For us Thunderbolt 1-2 users though, it may be dependent on the age of our machines? For Apple users, their website states:
So my Late 2013 rMBP seems out of luck? Is anyone with a 2014+ Mac able to install the drivers?
I'm guessing playing with the PCI-E Slot Capabilities registers (to enable Hot Plug) is out of the question as it needs to be supported by the BIOS?
The only other option is to try and stop the installer from restarting the card somehow?
Over TB3, each and every driver works no matter what.
Over M.2 the latest drivers work only with HotPlug enabled.
-
So here are my findings in relation to AER and Hot Plug using the latest drivers 375.70 :
- AER - Enabled + Hot Plug Enabled - NO CODE 43 / ERRORS
- AER Disabled + Hot Plug Enabled - NO CODE 43 / ERRORS
- AER Disabled + Hot Plug Disabled - CODE 43
- AER Enabled + Hot Plug Disabled - CODE 43
So AER doesn't seem to the culprit here, its Hot Plugging. This would lead me to believe they are trying to optimize hot plugging of eGPU's. Since it seems to work 8 out of 10 times when i plug in my GPU over TB3.
So if you guys have the Hot Plug option in your option in your BIOS for PCIe ports, maybe that will help. Or if @Tech Inferno Fan can implement it in his setup.
Props to @GTANAdam for noticing the Hot Plug being enabled.
-
1 hour ago, benjaminlsr said:
Pay attention ! There are stability issues between TB3 (maybe 2 also) and Wifi 2.4Ghz ! If you're egpu is disconnecting randomly, try to use a 5Ghz network and/or ethernet wire.
It works
Hmm..no issues on my end. Used a 1080 + TB3 HDK (Akitio TB3 PCIe Box) + the Intel 8260. I used this setup for about 5 days, before switching over to M.2.
-
15 minutes ago, Mr. Fox said:
No, that is an NVIDIA driver problem. This has been happening off and on again for years. I first experienced it with GTX 580M SLI in the M18xR1. Try a different driver version. Some driver versions are absolutely horrible and others are truly outstanding. Be sure to run DDU to clean out all of the old files between driver version changes or you may not get a clear picture of whether a particular driver version is better or not. If you cannot find a Pascal driver that does not have DPC latency problems, the best place to report your issues with NVIDIA drivers is GeForce Community forum.
Yeap, and it just became way more severe with the release of Pascal.
As usual, Nvidia doesn't care. They released the so called hotfix that worked for maybe 10% of the users, and they marked it as fixed. They arent even mentioning it in their patch notes anymore.
- 1
-
1 hour ago, rbut said:
@bloodhawk Can you please clarify on your solution. If you disable AER, then install the drivers, then reboot and immediately enable AER, then boot into windows for the first time, does it work?
Or is one reboot of windows with AER disabled required?
Yeap, after the driver installs, toggling AER doesn't do anything. At least until the next update.
1 hour ago, GTANAdam said:What I actually noticed, he had Hotplug option enabled which I had disabled when I disabled all L1 substates, gotta try that out as it could be just it.
theoretically, it could be that nvidia tries to reset the device when installing but the system does not respond correctly due to protection it has against hot plugging in PCIe.. could be it but I am not sure.
That actually makes sense, ill give it a shot with hot plugging disabled.
I have never had ASPM enabled btw.
-
-
1 hour ago, GTANAdam said:
I have created an account here just to explain that isn't Macbook-only related.. It concerns every PC in general..
I have a GTX 1060 3Gb and an Asus S551LB that I have modded the BIOS by myself to unlock all the menus along with the hard to unlock chipset menu, and have successfully managed to disable all the cards except the PCIE (the eGPU ofc) card, but that hasn't fixed it.. the only solution which worked was this 372.70 driver which worked perfectly fine, disabling all the PCIe ASPM related features in BIOS did not fix this for me nor when installing the driver too.. seems like Nvidia has introduced some changes that actually break something which they could actually fix if we protest and report in their forums in masses, write emails or whatever works.. (I am getting some interruptions from time to time for unknown reasons like black screen for a second)
Thank you for the driver hint btw, that's an awesome solution cuz I thought I had faulty devices.
Has anyone tested the latest 28th of October driver release?
Yeah the latest driver is working as well, after i disabled the Advanced Error Reporting.
Were you able to unlock that option as well ?
Instead of ASPM try and check if there is something called "Advanced Error Reporting" or similar under the PCIe settings for the port your eGPU is connected to. Because is the only think that worked for me. ASPM didn't do anything.
The black screen could either be :
- Unstable link.
- Insufficient power.
- In sufficient Bandwidth, when other ports are being used.
Here is a thread i created on their drivers forum :
-
13 minutes ago, ld0891 said:
They are moving their production line from Shanghai to Taiwan, and increasing the price to cover costs.
I contacted them last week for PE4C V4.1 since I'm in Mainland China and their shipping cost is too high through official shop.
A week passed without any reply and I have to call their Taipei headquarter about purchasing.
Then the news came that they have just increased the price.
So instead I bought the last PE4C V4.1 on Mfactors, but have to wait much longer before delivery.
Ah. Thank you for the update.
Yeah I got mine from Mfactors as well.
I might sell mine in December though, as soon as my desktop is assembled.
-
37 minutes ago, Tech Inferno Fan said:
Just checked. PE4C V3.0 is now listed at US$162.
A google cache from Oct-25 has it at US$65 .
So yeah, 150% increase in 3 days.
It's US$85 on Amazon-US and ebay-US.
PE4C V4.1 @ $175.
DAMN.
Any idea why the price hike ?
- 1
-
2 hours ago, frozentax said:
Oh man, I think I need to upgrade my desk game - nice setup! Are you using the cable that came with the PE4C? Mine seems a lot shorter than yours, I don't think I could even rig it like that if I had the space.
Not really, its the standard length, i just routed it in a way that it neatly goes behind the laptop and under. Ill post a few pictures tomorrow morning.
-
Good stuff!
btw, try not to stress the M.2 end of the cable too much.
This is what my setup ended up like :
SpoilerFinal Setup (for now) :
-
Haha haha I'm Indian, and this shit is fucking hilarious. XD
-
-
3 hours ago, Mr. Fox said:
Does it happen under load? It should never do it under load. At idle this is normal with stock BIOS due to c-states being enabled. If you are not already, should be using ThrottleStop to smooth out the bumps. Stock Clevo BIOS is junk.
It may help in Windows High Performance power plan advanced settings to set minimum processor state to 100%.
@Georgel THIS ^^
As i mentioned on the other forum as well, do not disable C-States unless you absolutely need to. Not every one is comfortable with the increased thermals. They help save a lot of power and heat generated when the system doesn't need that full boost clock.
- 3
-
11 minutes ago, Tech Inferno Fan said:
Yes, per previous message pls disable ASPM in power profile and enable it in BIOS to see if this is a fix for users without extensive BIOS options like you have. Based on your previous comments, it appears ASPM is causing error 43 in the newest NViDia drivers.
That didnt help.
I set Advanced error reporting back to Enabled and Enabled ASPM as well, along with double checking if its enabled in the power profile as. With the drivers installed and working, there was no change and things were normal, but if i reinstalled the drivers, it would result in a code 43. I had to disable Advanced error reporting again in order for the drivers to install and then i could do whatever i please. Enabling or disabling either of the settings didnt cause any issues or changes, after a successful installation.
My best guess is that the driver checks for something during the install?
One thing i did observe was, when the driver was being installed with Advanced Error reporting turned off, the black screen that we get during the installation was way longer than it normally takes, and after which everything went normally.
- 2
-
58 minutes ago, Tech Inferno Fan said:
Great investigation. Rightly so, if don't have such a BIOS setting then the problem persists with newer NVidia drivers.
Can you disable ASPM in your power profile using pic below (PCI Express -> Link State Power Management -> off ), re-enable ASPM in the BIOS and see if this power profile setting gives an equivalent fix?
Ill give that a shot. But i have ASPM disabled everywhere in my BIOS. Do you think ASPM might be causing this? I have it disabled in my power profiles as well.
Never quite used it in a long time, since it caused some issues for me, a while back.
It was a total brain fart while doing something else in the BIOS, figured since the system says the card is reporting problems, just make the card NOT report those problems
-
I managed to get the new drivers working on my system (GTX 1080 eGPU), might or might not work for you guys :
Went into the BIOS and disabled - Advanced Error reporting for the PCIe slot i had my GTX 1080 connected to.
SpoilerAlso not sure if this is possible with every BIOS or not. Since i dont quite have the standard issue Prema BIOS.
-
I managed to get the new drivers working on my system, might or might not work for you guys :
Went into the BIOS and disabled - Advanced Error reporting for the PCIe slot i had my GTX 1080 connected to.
SpoilerAlso not sure if this is possible with every BIOS or not. Since i dont quite have the standard issue Prema BIOS.
17" Clevo P870DM-G + [email protected] (PE4C V4.1) + Win10 [bloodhawk]
in Implementation Guides - PC
Posted
@Tech Inferno Fan This is with M.2 NGFF
Output :
CUDA-Z Report
=============
Version: 0.10.251 64 bit http://cuda-z.sf.net/
OS Version: Windows x86 6.2.9200
Driver Version: 375.70
Driver Dll Version: 8.0 (6.14.13.7570)
Runtime Dll Version: 6.50
Core Information
----------------
Name: GeForce GTX 1080
Compute Capability: 6.1
Clock Rate: 1733.5 MHz
PCI Location: 0:62:0
Multiprocessors: 20
Threads Per Multiproc.: 2048
Warp Size: 32
Regs Per Block: 65536
Threads Per Block: 1024
Threads Dimensions: 1024 x 1024 x 64
Grid Dimensions: 2147483647 x 65535 x 65535
Watchdog Enabled: Yes
Integrated GPU: No
Concurrent Kernels: Yes
Compute Mode: Default
Stream Priorities: Yes
Memory Information
------------------
Total Global: 8192 MiB
Bus Width: 256 bits
Clock Rate: 5005 MHz
Error Correction: No
L2 Cache Size: 48 KiB
Shared Per Block: 48 KiB
Pitch: 2048 MiB
Total Constant: 64 KiB
Texture Alignment: 512 B
Texture 1D Size: 131072
Texture 2D Size: 131072 x 65536
Texture 3D Size: 16384 x 16384 x 16384
GPU Overlap: Yes
Map Host Memory: Yes
Unified Addressing: Yes
Async Engine: Yes, Bidirectional
Performance Information
-----------------------
Memory Copy
Host Pinned to Device: 2784.71 MiB/s
Host Pageable to Device: 2371.13 MiB/s
Device to Host Pinned: 2541.68 MiB/s
Device to Host Pageable: 2224.87 MiB/s
Device to Device: 118.4 GiB/s
GPU Core Performance
Single-precision Float: 7892.93 Gflop/s
Double-precision Float: 259.88 Gflop/s
64-bit Integer: 398.933 Giop/s
32-bit Integer: 2363.25 Giop/s
24-bit Integer: 1794.89 Giop/s
Generated: Mon Oct 31 22:28:55 2016