Enclosure & Hardware Discussions
iMac Pro Vega 56 and eGPU RX 5700 XT why?
 

iMac Pro Vega 56 and eGPU RX 5700 XT why?  

  RSS

(@blackmamba_lu)
New Member
Joined: 1 month ago
 
Processor Information  
Name Intel Xeon W-2140B
Topology 1 Processor, 8 Cores, 16 Threads
Identifier GenuineIntel Family 6 Model 85 Stepping 4
Base Frequency 3.20 GHz
Package  
Codename  
L1 Instruction Cache 32.0 KB x 8
L1 Data Cache 32.0 KB x 8
L2 Cache 1.00 MB x 8
L3 Cache 11.0 MB x 1
OpenCL Information  
Platform Vendor Apple
Platform Name Apple
Device Vendor AMD
Device Name AMD Radeon Pro Vega 56 Compute Engine
Compute Units 56
Maximum Frequency 1.25 GHz
Device Memory 7.98 GB

OpenCL Performance

Score 160716  
Sobel 399765
17.6 Gpixels/sec
 
Histogram Equalization 105719
3.30 Gpixels/sec
 
SFFT 35508
88.5 Gflops
 
Gaussian Blur 404890
7.09 Gpixels/sec
 
Face Detection 31029
9.06 Msubwindows/sec
 
RAW 1124986
10.9 Gpixels/sec
 
Depth of Field 750092
2.18 Gpixels/sec
 
Particle Physics 27979
4423.2 FPS
Processor Information  
Name Intel Xeon W-2140B
Topology 1 Processor, 8 Cores, 16 Threads
Identifier GenuineIntel Family 6 Model 85 Stepping 4
Base Frequency 3.20 GHz
Package  
Codename  
L1 Instruction Cache 32.0 KB x 8
L1 Data Cache 32.0 KB x 8
L2 Cache 1.00 MB x 8
L3 Cache 11.0 MB x 1
OpenCL Information  
Platform Vendor Apple
Platform Name Apple
Device Vendor AMD
Device Name AMD Radeon RX 5700 XT Compute Engine
Compute Units 40
Maximum Frequency 2.10 GHz
Device Memory 7.98 GB

OpenCL Performance

Score 109706  
Sobel 104883
4.62 Gpixels/sec
 
Histogram Equalization 77984
2.44 Gpixels/sec
 
SFFT 24385
60.8 Gflops
 
Gaussian Blur 189284
3.32 Gpixels/sec
 
Face Detection 49030
14.3 Msubwindows/sec
 
RAW 515125
4.99 Gpixels/sec
 
Depth of Field 571264
1.66 Gpixels/sec
 
Particle Physics 38520
6089.6 FPS
 
This topic was modified 1 month ago

Pending: Add my system information and expected eGPU configuration to my signature to give context to my posts


ReplyQuote
joevt
(@joevt)
Prominent Member
Joined: 2 years ago
 

I think @mac_editor had a similar result comparing Vega 64 and 5700 XT (both in eGPUs connected to a MacBook Pro 2018).
2018-15-macbook-pro-rp560x-8th6ch-rx-5700-xt-32gbps-tb3-mantiz-venus-macos-10-15-1-win10-mac_editor

You have a slower Vega 56 but it gains more speed because it's not in an eGPU.

Does anyone know what the Open CL Performance numbers mean? Are any of them directly constrained by the Thunderbolt 3 limit of 22 Gbps? I think some of them must be. Particle Physics works best on 5700 XT at least. 

Mac mini (2018), Mac Pro (Early 2008), MacBook Pro (Retina, 15-inch, Mid 2015), GA-Z170X-Gaming 7, Sapphire Pulse Radeon RX 580 8GB GDDR5, Sonnet Echo Express III-D, Trebleet Thunderbolt 3 to NVMe M.2 case


ReplyQuote
Anthlenv
(@anthlenv)
New Member
Joined: 9 months ago
 

Huh, I have had a Vega 56 in eGPU and a 5700 XT Annv in egpu (Developer beta 2) and the 5700 XT scored about 25 percent more with Metal with Geekbench 5. In Geekbench 4 I got 154348 with my Mac mini and my Vega 56 in egpu using OpenGL and with my 5700 XT in eGPU I got 176941. So do with that what you will. Your 5700 XT numbers seem way off. 

Pending: Add my system information and expected eGPU configuration to my signature to give context to my posts


ReplyQuote
psonice
(@psonice)
Estimable Member
Joined: 2 years ago
 

From what I remember from a quick look at the 5700 series, they're quite weak in terms of compute but fast for other things. The OpenCL tests will be more compute oriented, so it's not surprising at all to see the Vega 56 win there. For most games it’ll probably go the other way, but it’ll depend still. What it really means is if you’re doing rendering and such the Vega is a better bet. I’m guessing Vega II will beat the 5700XT in the gaming benchmarks, while also offering more compute power. 

Pending: Add my system information and expected eGPU configuration to my signature to give context to my posts


ReplyQuote
mac_editor
(@mac_editor)
Famed Member Moderator
Joined: 3 years ago
 

Geekbench has always been inconsistent across versions (even minor updates) - I know major updates set new baselines and scores decrease, but I’m not too convinced about their GPU tests. In Luxmark, currently the Vegas would outperform Navi on macOS. The Vega GPUs still offer higher peak flops vs. the current Navi lineup. Higher end Navi GPUs should surpass them. My initial impression of Navi is that it is probably the most NVIDIA-like card from AMD, appropriately configured for good gaming performance (as it should be). Even so, the compute performance decrease for day-to-day use in video editing for example should be minimal vs Vega. 

purge-wrangler.shpurge-nvda.shset-eGPU.shautomate-eGPU EFI Installer
2018 MacBook Pro 15" RP560X + RX 5700 XT (Mantiz Venus)


ReplyQuote
(@christopher_straser)
Active Member
Joined: 1 month ago
 

Geekbench is broken since macOS 10.14.5. Scores are way too low on most AMD cards and the new Navi cards are no exception. This also hasn't been fixed in Catalina or the new Geekbench 5. I have no Idea why they don't fix this.

For example: Back when running a Vega 64 in my hackintosh it scored roughly 200k points in Geekbench 4. This was reduced to roughly 150k points with macOS 10.14.5. Yet none of the other Benchmarks confirms a degradation in performance - so definitely a Geekbench issue and not a driver issue.

Same is happening with my Radeon VII. It also scores roughly 150k points in Geekbench 4 throughout all current macOS releases and around 57k points in Geekbench 5. Running Windows on the same machine with the same voltage and clock settings I score 214k in Geekbench 4 and 82k in Geekbench 5.

Long story short: Geekbench on macOS is crap. Avoid it and don't trust the scores. Use other benchmaks like Luxmark or Indigo Benchmark.

Pending: Add my system information and expected eGPU configuration to my signature to give context to my posts


ReplyQuote