Jump to content
You must now use your email address to sign in [click for more info] ×

Affinity 2.0 is Hideously Slow/Buggy on Windows


Recommended Posts

4 minutes ago, Nomad Raccoon said:

In fact, I have a X570 Aorus Master motherboard, it’s all pcie 4.0, even the the ssds. 

Ok, well THAT theory is soundly debunked then, since you're also on a x570, so it's not the PCIe bus slowing things down. I thought you were on a B550 for some reason - probably was someone else who replied in the thread, as I didn't go back and reread it

Link to comment
Share on other sites

1 hour ago, debraspicher said:

There was a 3060TI narrowly not beating a 3090 in the V1 benchmark charts lol. @MikeTO  (I believe) kept an extensive chart, but can't access the thread now. It seems correlated to CPU pairings, other factors. And tbf, I do have quite a bit running in the background that I could still clear away and my rig is not optimized to run benchmarks.

The 1.10.5 chart was in the Photo beta thread which is probably archived. Here it is again. Note, the numbers can't be directly compared between 1.10.5 and 2.0.x, but it will provide directional information for those looking at a hardware upgrade.

Benchmarks.pdf

Benchmarks.numbers

Benchmarks.xlsx

It was created in Numbers but I exported it to Excel for Windows users who want to fiddle with it. The green highlighting is applied automatically to the cell(s) with the fastest score with a margin of 5% since the same hardware can produce different results each time the benchmark is run.

Cheers

Download a free manual for Publisher 2.4 from this forum - expanded 300-page PDF

My system: Affinity 2.4.0 for macOS Sonoma 14.4, MacBook Pro 14" (M1 Pro)

Link to comment
Share on other sites

7 minutes ago, debraspicher said:

image.png.853d60700bd9f2a7bdac07b3ecbf58e0.png

image.png.d2b3dd6a060963a8dc92b4b342c8845c.png

image.png.7d3c5ec8abcac33fcb14eddf8151d6d6.png

Yes, that's as it should be: a lot faster than my 3070. Your compute benchmark results are similar to other 3080 users. 

So it's very perplexing that the 3080 scores lower on an Affinity benchmark than a 3070. And it's not just one person as we've seen - it's affecting two of you. Be useful to see if other 3080 forum users get similar benchmark results with the Affinity benchmark. 

This points to the performance being an Affinity issue it seems.

Link to comment
Share on other sites

1 hour ago, rvst said:

I'd definitely be curious to compare a Geekbench OpenCL compute benchmark with the one posted earlier in the thread. Your 3080 should blow my 3070 out of the water. 

Hard to tell what else it would be - your machine is overclocked pretty heavily as well and it's the major significant difference between your x470 chipset and my x570 chipset. The only other big difference is I have a lot more RAM, but that's more likely to slow things down, since I have all 4 banks populated (not sure what your memory config is - I see it's the same speed, but whether or not you have 2 or 4 banks populated I don't know). 

One other difference I note is I'm using Windows Pro and you're on Home, but I've never heard of any performance difference between the two. 

Sorry for the double reply, it is difficult to edit quotes out of order, much less write behind them. Ironically, found out by accident one way to do it after starting this post.

I have 2 DIMMs used. (Had to check BIOS to grab slot configuration, CPU-Z is being silly about showing DC)

image.png.c31bb5b311662380742670bfa588fdf9.png

Other information including memory timings:

image.png.bb49978911105b809220f719ecb61d02.pngimage.png.04d943bffda6399f80de08c05080971e.png

image.png.55afb0dde2121391d036d7a39acacf6e.png

17 minutes ago, rvst said:

Yes, that's as it should be: a lot faster than my 3070. Your compute benchmark results are similar to other 3080 users. 

So it's very perplexing that the 3080 scores lower on an Affinity benchmark than a 3070. And it's not just one person as we've seen - it's affecting two of you. Be useful to see if other 3080 forum users get similar benchmark results with the Affinity benchmark. 

This points to the performance being an Affinity issue it seems.

That would be my guess. My scores do bounce quite a bit for GPU results when testing during different periods. It's probable* I need to retest with a cleaner machine and go back and kill unnecessary processes. If you peak at Miketo's PDF he attached in an earlier post, it's an accumulation of V1 benches which features quite the hardware spread. If the forum were not archived, it might've still been possible to see some of the full configurations listed, but it gives a basic idea of how hardware scaled in Affinity V1.

Link to comment
Share on other sites

20 minutes ago, debraspicher said:

My scores do bounce quite a bit for GPU results when testing during different periods. It's probable* I need to retest with a cleaner machine and go back and kill unnecessary processes

I doubt it will change anything. At most, it will make a few percent difference. I ran that earlier benchmark with about 20 windows open, WSL2 running a full gnome desktop environment and about 50 tabs open in my browser, all the while working on a huge Excel model and it still looked the same as earlier benchmarks I ran with a cleanly booted system.

You should be 20% higher than me on the Affinity benchmark and I doubt a clean boot is going to do that.

Link to comment
Share on other sites

Interesting results with a laptop with Radeon mobile card. Laptop is very old so not sure it is utilizing the GPU, but it shows up in Performance tab and OpenCL claims enabled.

221214_laptop_ap_perf-preferences.png.a84bf5f2ab07f7cee173d09fbb5b10ea.png

Benchmark Sandboxed (Notice Vector Single score)

221214_v2_laptop_sandboxed-benchy.png.e7772ce03b70b44ddef4e999d031d940.png

Benchmark Unsandboxed (Vector Single is lower --ran this 2x)

221214_v2_laptop_unsandboxed-benchy.png.908f38c15b542260a7eb1dc7e6170428.png

Hardware details:

Quote

221214_laptop_cpu-z-1.png.7c03a2153ebf922be0b2c98aabadcbc2.png221214_laptop_cpu-z-2.png.dc7cd7069e21838678c19d379341da28.png

221214_laptop_gpu-z.png.9823fe0eade8b1ad83598fc81d2e6ac1.png

Link to comment
Share on other sites

1 hour ago, debraspicher said:

Interesting results with a laptop with Radeon mobile card. Laptop is very old so not sure it is utilizing the GPU, but it shows up in Performance tab and OpenCL claims enabled

I see GeForce 960M not Radeon ;-) Plus looks like benchmark uses Intel integrated. On top of that, old Radeons pre 5XXX (Vega, RX5XX and RX4XX)series works well. AMD has a problem with OpenCL since 5XXX series up (I have no idea about just released 7XXX, but I guess they have same problems like 5 & 6 series)

Link to comment
Share on other sites

Well this is fun.

I installed a 5600XT in my main machine to do the benchmarks. Except, now I'm on the unreleased release candidate and the benchmark version is different on my sandboxed "retail" version now. Thankfully it matches beta, so I was able to unsandbox it to desktop and run a comparison.

 

"old" 2.0 for Entertainment Purposes Only (Unofficial unsandboxed 2.0.0.1640 benchmark 20000)

 

221214_V2.0.3.1640_amd-radeon-5600xt_unsandboxed.png.030b221139813617d17b9ea1fc597dba.png

New unreleased 2.0 (Sandboxed 2.0.3.1685 Benchmark Version 20300)

221214_V2.0.3.1685_amd-radeon-5600xt_sandboxed.png.f691026ab437e9bd0a520502b00281b0.png

Current Beta 2.0 (Sandboxed 2.0.3.1670 Benchmark Version 20300)

221214_V2.0.3.1670BETA_amd-radeon-5600xt_sandboxed.png.5baff3feab1eaa495661cbb64fba6f31.png

Current Beta 2.0 (Unofficial Unsandboxed 2.0.3.1670 Benchmark Version 20300)

10% change in GPU Single

221214_V2.0.3.1670BETA_amd-radeon-5600xt_unsandboxed.png.b932e14738ce2aed869d770d4d0cd82d.png

Link to comment
Share on other sites

10 minutes ago, slizgi said:

I see GeForce 960M not Radeon ;-) Plus looks like benchmark uses Intel integrated. On top of that, old Radeons pre 5XXX (Vega, RX5XX and RX4XX)series works well. AMD has a problem with OpenCL since 5XXX series up (I have no idea about just released 7XXX, but I guess they have same problems like 5 & 6 series)

I threw some 5600XT benchmarks below your post to check for CPU bottleneck. I just added the tested change, it's about +10% unsandboxed so it looks like the issue worsens with newer faster hardware?

I can retest a couple more times to be sure later and then put back in my NVIDIA for retests, but won't be until later.

Link to comment
Share on other sites

58 minutes ago, debraspicher said:

Well this is fun.

I installed a 5600XT in my main machine to do the benchmarks. Except, now I'm on the unreleased release candidate and the benchmark version is different on my sandboxed "retail" version now. Thankfully it matches beta, so I was able to unsandbox it to desktop and run a comparison.

 

"old" 2.0 for Entertainment Purposes Only (Unofficial unsandboxed 2.0.0.1640 benchmark 20000)

 

221214_V2.0.3.1640_amd-radeon-5600xt_unsandboxed.png.030b221139813617d17b9ea1fc597dba.png

New unreleased 2.0 (Sandboxed 2.0.3.1685 Benchmark Version 20300)

221214_V2.0.3.1685_amd-radeon-5600xt_sandboxed.png.f691026ab437e9bd0a520502b00281b0.png

Current Beta 2.0 (Sandboxed 2.0.3.1670 Benchmark Version 20300)

221214_V2.0.3.1670BETA_amd-radeon-5600xt_sandboxed.png.5baff3feab1eaa495661cbb64fba6f31.png

Current Beta 2.0 (Unofficial Unsandboxed 2.0.3.1670 Benchmark Version 20300)

10% change in GPU Single

221214_V2.0.3.1670BETA_amd-radeon-5600xt_unsandboxed.png.b932e14738ce2aed869d770d4d0cd82d.png

Wait a sec, 5600XT has a 4020 pts in 200 and 4700 pts in 203, it is an improvement. I am unable to test 203 beta, because I did not buy V2 yet because of this AMD GPU thing. I am just testing on a trial and trial is not eligible to test beta. Most funny thing is I have newer generation of GPU and 2 class higher and have 880 pts in 200. What is even going on here AMD 🤣

Link to comment
Share on other sites

31 minutes ago, rvst said:

We can't really say that. The benchmark across different versions is not directly comparable, even minor versions.

Well I understand that 1.9.10 and 2.0.0 are not compatible, but 2.0.0 and 2.0.3 should be...

Link to comment
Share on other sites

  • Staff
3 minutes ago, slizgi said:

Well I understand that 1.9.10 and 2.0.0 are not compatible, but 2.0.0 and 2.0.3 should be...

Why? The benchmark calculation is for our diagnostic purposes, we can change it when we wish.

Patrick Connor
Serif Europe Ltd

"There is nothing noble in being superior to your fellow man. True nobility lies in being superior to your previous self."  W. L. Sheldon

 

Link to comment
Share on other sites

1 hour ago, slizgi said:

Wait a sec, 5600XT has a 4020 pts in 200 and 4700 pts in 203, it is an improvement. I am unable to test 203 beta, because I did not buy V2 yet because of this AMD GPU thing. I am just testing on a trial and trial is not eligible to test beta. Most funny thing is I have newer generation of GPU and 2 class higher and have 880 pts in 200. What is even going on here AMD 🤣

If you look on 20300 bench results, my Vector single actually went down a good chunk (430s versus 480s). I certainly didn't change CPUs lol.

Link to comment
Share on other sites

Ok. So I retested the AMD 5600XT.
I updated my drivers for the NVIDIA 3080 and retested again on new benchmark.

I did 3x for averaged results and highlighted the top score for Single GPU.

Both are tested on the current official beta (2.0.3.1670 Benchmark Version 20300).

Here's the results.

Hardware:

Quote

Microsoft Windows 10 Home (Build 19045)
AMD Ryzen 7 5800X @ 3.8Ghz (-30 all core +200mhz PBO); Mobo: Asus X470 Prime Pro
32GB DDR4 (3600Mhz)
Monitor 1&2 4K @ 150%

AMD 5600XT 6G Results

Variance: 5% ~ 6%

Sandboxed (2.0.3.1670 Beta Benchmark Version 20300)

221214_V2.0.3.1670BETA_amd-radeon-5600xt_sandboxed_RETESTS.png.c1d73f84a7d20eb01a8fd3d5cce181fa.png

Unofficial Unsandboxed (2.0.3.1670 Beta Benchmark Version 20300)

221214_V2.0.3.1670BETA_amd-radeon-5600xt_unsandboxed_RETESTS.png.a9ee93b630663f9babd32f9596ffbfa8.png

 

NVIDIA RTX3080 12GB Results

Variance: 18% ~ 20%

Sandboxed (2.0.3.1670 Beta Benchmark Version 20300)

221214_V2.0.3.1670BETA_nvidia-rtx-3080-12g_sandboxed_RETESTS.png.c15f662398dcb7a03bd1617abbc79e86.png

Unofficial Unsandboxed (2.0.3.1670 Beta Benchmark Version 20300)

221214_V2.0.3.1670BETA_nvidia-rtx-3080-12g_unsandboxed_RETESTS.png.7b2926e38d064d9fb8fe3bdcd630e369.png


Summary: Similar drop in performance being sanded versus unsanded  on the 20300 benchmark as the 20000 benchmark in the NVIDIA 3080.

So to add to weirdness: On the latest driver for the Nvidia (Nov 13 -> Dec 5), my OpenCL performance went up 8.7% (184566 vs 210300) according to geekbench. That could've explained the variance between @rvst's 3070 & my 3080 benches. However, I retested the old v2 unsanded benchmark a few times and didn't get much change from the last result. I included the best result from old benchmark for Single GPU. Unsanded-only because I don't have access to the sanded v2 now on this machine:

image.png.37ba3238db5447bc90dd58185cb6e057.png

12809 vs  13342 now. Not much change.

EDIT 12/16: Retested geekbench when I had more time and it does vary, so above result should be considered normal.

AMD 5600XT Info

221214_geekbench-OpenCL_5800x-rig.png.cd8a9dc9a6174e70187aced9eb08975e.png

221214_gpu-z_5600xt_5800x-rig.png.d207427320afe0625f2b1b413e4d3b7d.png

221214_adrenaline-driver-info_5600xt_5800x-rig.png.b816edba197e477d71df430b1b0bb0c4.png

 

NVIDIA RTX3080 12G Info (Updated Drivers Dec 05 2022)

221214_NVIDIA_3080_benchmark-V2_2.0.3.1670_BETA_BENCH20300_geekbench-updated-drivers.png.dbd100ee9425da5e7be19df3d09255f3.png

221214_NVIDIA_3080_benchmark-V2_2.0.3.1670_BETA_BENCH20300_gpu-z-updated-drivers.png.6b4161a1c1f53b38ce84e5464f5bdb9a.png

Edited by debraspicher
Rechecked geekbench
Link to comment
Share on other sites

5 hours ago, Patrick Connor said:

Why? The benchmark calculation is for our diagnostic purposes, we can change it when we wish.

Are you happy with the performance of your apps on PC Workstations and the new M1/M2 Macs?

Why do these apps appear to only use a modicum of a GPU's potential and (at most) two cores of a CPU?

What is being done to improve the scaling of performance of the apps on higher end hardware (many cored CPUs and high end GPUs)

Has the supposed rewrite of the apps been done with a view to multithreading such that you can saturate any number of CPU cores and more fully utilise modern GPU power and performance? 

Link to comment
Share on other sites

9 hours ago, Patrick Connor said:

Why? The benchmark calculation is for our diagnostic purposes, we can change it when we wish.

Because it confuses people, what is for example visible one post above this. But honestly I really don't care, it is just automatically what I would expect because it's called "benchmark" 😀 It is just semantics.

Link to comment
Share on other sites

4 hours ago, deeds said:

Are you happy with the performance of your apps on PC Workstations and the new M1/M2 Macs?

Why do these apps appear to only use a modicum of a GPU's potential and (at most) two cores of a CPU?

What is being done to improve the scaling of performance of the apps on higher end hardware (many cored CPUs and high end GPUs)

Has the supposed rewrite of the apps been done with a view to multithreading such that you can saturate any number of CPU cores and more fully utilise modern GPU power and performance? 

Second this, for a rework of the app it seems just as bad if not worse performance for Windows based systems, Amd especially. But my 3080 is barely used, my cpu is barely used.

And the savants on the thread are discussing UMA on Apple when Infinity Fabric is literally staring them in their face along with Pci 4.0 x16 . 
 

It doesn’t take a geekbench to realize this app is currently aimed to perform only at macs/ios? 
 

Will Serif do anything about this or not?

Link to comment
Share on other sites

The problem is that Metal is way better than the available OpenCL implementations for such workloads.

And concerning UMA and PCIE4: As fast as PCIE may become it always need 2 copy operations when cpu had to do something, on UMA the ram is used directly without any slow operation betwenn cpu/gpu access.

Here are current speed rates:

PCIE4/16 - 31,5GB/s
M1 - 66,67GB/s
M1Pro - 200GB/s
M1Max - 400GB/s
M1Ultra - 800GB/s

Mac mini M1 / Ryzen 5600H & RTX3050 mobile / iPad Pro 1st - all with latest non beta release of Affinity

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...

Important Information

Terms of Use | Privacy Policy | Guidelines | We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.