ALERT: half H2D performance issue on TI83 TB3 enclosures (Node, Venus, Devil Box...
 
Notifications
Clear all

ALERT: half H2D performance issue on TI83 TB3 enclosures (Node, Venus, Devil Box, XG Station 2)  

 of  29
  RSS

kotlos
(@kotlos)
Trusted Member
Joined: 3 years ago
 
Posted by: Jetcopter

 

I have seen a difference in the games I have played.  But using something like 3DMark is general accepted as a good method (less subjective) to compare performance between setups and cards. It is essentially measuring FPS which is a good indicator of performance. 

 

I guess 3DMark could be used as well, but then the comparison should be between a TI82 and a TI83 case (which is done and shown no difference), not between an internal and an external configuration. 

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

I have AKiTiO Thunder2 and AKiTiO Node which I can test for the difference between Thunderbolt 2 and Thunderbolt 3 speed. Anyone with TI82 and TI83 enclosures to do comparable tests?

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2013 Mac Pro (FP D500 x2) [3rd,12C,E] + RX 5700 XT @ 16Gbps-TB2>TB3 (Mantiz Saturn Pro) + macOS 11 [build link]  


ReplyQuote
Mymantiz_John
(@mymantiz_john)
Vendor
Joined: 4 years ago
 
Posted by: kotlos

 

Posted by: Jetcopter

 

I have seen a difference in the games I have played.  But using something like 3DMark is general accepted as a good method (less subjective) to compare performance between setups and cards. It is essentially measuring FPS which is a good indicator of performance. 

   

I guess 3DMark could be used as well, but then the comparison should be between a TI82 and a TI83 case, not between an internal and an external configuration. 

   

My comparison was based on Razer TI82 ( 2200mb ) VS Mantiz Venus Ti83 ( 1100MB )

Mantiz: ● ●


kotlos liked
ReplyQuote
goalque
(@goalque)
Noble Member Admin
Joined: 4 years ago
 
Posted by: wimpzilla

 

It is really funny and amuse me to saw that nobody is able to to compile and perform decent ingame benchmark instead of throwing useless valley scores or basic benchmarks.

Would you explain why the Valley scores are useless? We need a tool to prove whether H2D bandwidth matters or not. None of the benchmarks on this thread were not any better, including 3DMark and Compubench. @Mymantiz_John did not see any performance gap between the 2200MiB/s vs 1100MiB/s.

The Matlab and CUDA-Z tests are correct. I always get roughly ~500MiB/s difference in bandwidth (Node vs Devil Box), but no any difference in OpenCL, CUDA or DirectX11 benchmarks.

A German review of the Devil Box and their numbers prove the same fact what I pointed out in a quick Valley bench: dual core CPU is useless in gaming if the aim is to maximize the performance:

https://www.computerbase.de/2016-10/powercolor-devil-box-test

Modern GPUs don’t require much bandwidth:

http://www.tested.com/tech/457440-theoretical-vs-actual-bandwidth-pci-express-and-thunderbolt/

"Everything down to x16 1.1 and its equivalents (x8 2.0, x4 3.0) provides sufficient gaming performance even with the latest graphics hardware, losing only 5% average in worst-case".

"Last year's most powerful graphics cards perform just fine at PCIe 2.0 x8 or even PCIe 3.0 x4"

The CPU/PCH makes the real difference. I tried to measure PCIe traffic but unfortunately my CPU (i5-6360U) is not supported:

 

./pcm-pcie.x

 Processor Counter Monitor: PCIe Bandwidth Monitoring Utility 
 This utility measures PCIe bandwidth in real-time

 PCIe event definitions (each event counts as a transfer): 
   PCIe read events (PCI devices reading from memory - application writes to disk/network/PCIe device):
     PCIePRd   - PCIe UC read transfer (partial cache line)
     PCIeRdCur* - PCIe read current transfer (full cache line)
         On Haswell Server PCIeRdCur counts both full/partial cache lines
     RFO*      - Demand Data RFO
     CRd*      - Demand Code Read
     DRd       - Demand Data Read
     PCIeNSWr  - PCIe Non-snoop write transfer (partial cache line)
   PCIe write events (PCI devices writing to memory - application reads from disk/network/PCIe device):
     PCIeWiLF  - PCIe Write transfer (non-allocating) (full cache line)
     PCIeItoM  - PCIe Write transfer (allocating) (full cache line)
     PCIeNSWr  - PCIe Non-snoop write transfer (partial cache line)
     PCIeNSWrF - PCIe Non-snoop write transfer (full cache line)
     ItoM      - PCIe write full cache line
     RFO       - PCIe parial Write
   CPU MMIO events (CPU reading/writing to PCIe devices):
     PRd       - MMIO Read [Haswell Server only] (Partial Cache Line)
     WiL       - MMIO Write (Full/Partial)

 * - NOTE: Depending on the configuration of your BIOS, this tool may report '0' if the message
           has not been selected.

Number of physical cores: 4
Number of logical cores: 4
Number of online logical cores: 4
Threads (logical cores) per physical core: 1
Num sockets: 3
Physical cores per socket: 1
Core PMU (perfmon) version: 4
Number of core PMU generic (programmable) counters: 4
Width of generic (programmable) counters: 48 bits
Number of core PMU fixed counters: 3
Width of fixed counters: 48 bits
Nominal core frequency: 2000000000 Hz
Package thermal spec power: 15 Watt; Package minimum power: 0 Watt; Package maximum power: 0 Watt; 

Detected Intel(R) Core(TM) i5-6360U CPU @ 2.00GHz "Intel(r) microarchitecture codename Skylake"
Jaketown, Ivytown, Haswell, Broadwell-DE Server CPU is required for this tool! Program aborted
Cleaning up
 Zeroed PMU registers

The core count is incorrect. i5-6360U has only 2 cores:

http://ark.intel.com/products/91156/Intel-Core-i5-6360U-Processor-4M-Cache-up-to-3_10-GHz

automate-eGPU EFIapple_set_os.efi

Mid 2015 15-inch MacBook Pro eGPU Master Thread

 
2018 13" MacBook Pro [8th,4C,U] + Radeon VII @ 32Gbps-TB3 (ASUS XG Station Pro) + Win10 1809 [build link]  


ikir liked
ReplyQuote
wimpzilla
(@wimpzilla)
Honorable Member
Joined: 4 years ago
 

The answer is simple but didn't want to begin an off topic on benchmark vs games performances. I will simply point out one thing: in the top hardware sites reviews for a cpu or vga, witch part of the test when you read it, show more time spent testing: the 2 graphs showing 3DMark/Valley result or the 15 others graphs showing ingame results, with different resolution, with different games, etc?

Well you got your answer, i think. This is also the difference between cheap review and the well done ones, where you can make an opinion about and decide if buy or not!

Test protocol:

-Same laptop, possibly with a true quad. Same bios, ME firmware, same windows, mac install, same drivers.

-Same cable.

-Same video adapter, same clocks.

-Same power supply.

-Both T82/T83 TBE box adapters, each test should refer to a precise firmware, both with the eGPU one both with the pci-e one.

-Sames games with a built in benchmark like Tomb Raider, The Division, R6: Siege, Shadows of Mordor, GTAV, Grid Autosporp, Grid Rally, Grid 2, DOOM, Hitman, METRO, etc.

-Same resolution 1080p, 1440p, 4K.

-Repeat each test minimum 3 times, recording low/max/avg fps, if you can also record the frametime, t°, usage, etc.

 

=> This protocol should be applied on the same laptop, first with TB2 only laptop, then another time with a TB3 only laptop.

Hope it is clear. Smile

What make real difference in games it is the cpu speed/clock and how much cache it have. What allow i7 dual core to anyway perform not so bad, it is the high clock when boosting under the turbo. The best example of that, if you are interested, is the Intel I3 7350k!!! Perform extremely well when the multithreading is not the main thing to focus.

 

 

 

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²

 
2012 15" Lenovo Thinkpad T530 [2nd,4C,Q] + R9 270X @ 4Gbps-mPCIe2 (EXP GDC 8.4) + Win10 [build link]  


ReplyQuote
Poblopuablo
(@poblopuablo)
Eminent Member
Joined: 4 years ago
 
Posted by: wimpzilla

 

The answer is simple but didn't want to begin an off topic on benchmark vs games performances. I will simply point out one thing: in the top hardware sites reviews for a cpu or vga, witch part of the test when you read it, show more time spent testing: the 2 graphs showing 3DMark/Valley result or the 15 others graphs showing ingame results, with differents resolution, with different games, etc?

Well you got your answer, i think. This is also the difference between cheap review and the well done ones, where you can make an opinion about and decide if buy or not!

Test protocol:

-Same laptop, possibly with a true quad. Same bios, ME firmware, same windows, mac install, same drivers.

-Same cable.

-Same video adapter.

-Same power supply.

-Both T82/T83 TBE box adapters, each test should refer to a precise firmware, both with the eGPU one both with the pci-e one.

-Sames games with a built in benchmark like BF4 (kinda), Tomb Raider, The Division, R6: Siege, Shadows of Mordor, GTAV, Grid Autosporp, Grid Rally, Grid 2, DOOM, etc.

-Same resoltion 1080p, 1440p, 4K.

-Repeat the each test minimum 3 times, with low/max/avg fps, if you can also record the frame time, t°, usage, etc.

 

=> This protocol should be applied on the same laptop first with TB2 only, then another time with a TB3 only laptop.

 

Hope it is clear. Smile

 

 

 

   

I thought the clarification was not for TB3 vs TB2. You have to use different computers which will likely not have the same specs, because it's unlikely the only difference between 2 computers is just 1 port(as far I know).  For example, a Mac book with TB2 won't have the same cpu (and other specs) as a MacBook with TB3

 

Posted by: kotlos

 

I think the best test would be comparing real world performance (game performance or a professional application) with the same card and computer of a TB2TI82  and TB3TI83 case.

 

but rather TI82 vs TI83(you just switch out the GPU docks, this way all computer specs will stay the same) i didn't think it was the actual TB2 v TB3 ports (but rather the difference  speeds that resembles TB3(TI82) v TB2(TI83). Both of which actually utilize TB3, but have different results)

 

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
wimpzilla
(@wimpzilla)
Honorable Member
Joined: 4 years ago
 

You can use both box on a laptop that is TB2 or TB3, i hope! Since it was posted getting better result with a TB3 box on a TB2 laptop than a TB3 box on a TB3 laptop.

-TB3 laptop + TB3/TB2 box

-TB2 laptop + TB3/TB2 box

=> Compare results!

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²

 
2012 15" Lenovo Thinkpad T530 [2nd,4C,Q] + R9 270X @ 4Gbps-mPCIe2 (EXP GDC 8.4) + Win10 [build link]  


ReplyQuote
Poblopuablo
(@poblopuablo)
Eminent Member
Joined: 4 years ago
 

Those laptops won't have the same specs(cpu I can almost gaurentee is different.) And that matters. Along with other specs. 2 different PCs will have different bios and drivers.

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
Poblopuablo
(@poblopuablo)
Eminent Member
Joined: 4 years ago
 

I see what your saying, but can somebody go from TB2 egpu to TB3 laptop? 

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
wimpzilla
(@wimpzilla)
Honorable Member
Joined: 4 years ago
 

It is not hard to swap a cpu/ram/ssd between laptops, for the sake of keeping everithing same!

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²

 
2012 15" Lenovo Thinkpad T530 [2nd,4C,Q] + R9 270X @ 4Gbps-mPCIe2 (EXP GDC 8.4) + Win10 [build link]  


ReplyQuote
 of  29