Don’t have any clue what this means
RAM tuning is quite time consuming if u want to get the last 10-20% out of your system. Much more so than CPU or GPU overclocking. So if u have the time and patience for lots of trial and error it can be worth it.
Here is a a recent article covering many aspects of RAM overclocking:
Some aspects are:
- 2 sticks (4 puts more strain on memory controller) of the right kind of RAM
- A motherboard which is in practice capable of good RAM speeds
- A CPU which doesn’t have a bad performing memory controller (maybe 80% of Ryzen 9000 CPUs have at least a decent memory controller)
- Extra fan for RAM is recommended since u will work with higher voltages = more heat which can make the sticks instable
- There is always the chance of corrupting your Windows installation when an overclock goes wrong, be prepared by having proper backup or using extra SSD with Windows installation just for testing
- RAM errors can be difficult to track down since there are many parameters and they can occur very infrequently even after hours of operations
3 Likes
rtx2080
Topaz Video AI v7.0.2
System Information
OS: Windows v29.09
CPU: Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz 15.944 GB
GPU: NVIDIA GeForce RTX 2080 7.8135 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 12.62 fps 2X: 07.58 fps 4X: 02.03 fps
Iris 1X: 12.45 fps 2X: 06.58 fps 4X: 01.95 fps
Proteus 1X: 12.42 fps 2X: 07.18 fps 4X: 02.13 fps
Gaia 1X: 04.28 fps 2X: 02.95 fps 4X: 01.82 fps
Nyx 1X: 00.63 fps 2X: 02.88 fps
Nyx Fast 1X: 06.48 fps
Rhea 4X: 01.49 fps
RXL 4X: 01.36 fps
Hyperion HDR 1X: 15.17 fps
4X Slowmo Apollo: 16.01 fps APFast: 28.55 fps Chronos: 10.03 fps CHFast: 15.13 fps
16X Slowmo Aion: 18.78 fps
1 Like
can someone post any laptop with a rtx 5080 benchmark a simple 1080 one will do.
I got the 5070 ti… now i wanna see the 5080
MOBILE only please not desktop
I have overclocked the processor and I see an improvement still will work more cause has not crashed yet will work more until crash and then go latest stable tune up ```
Topaz Video AI v7.0.2
System Information
OS: Windows v11.23
CPU: AMD Ryzen 9 7900 12-Core Processor 63.162 GB
GPU: NVIDIA GeForce RTX 4080 SUPER 15.671 GB
GPU: AMD Radeon™ Graphics 0.47438 GB
Processing Settings
device: 0 vram: 0.7 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 23.99 fps 2X: 14.79 fps 4X: 04.07 fps
Iris 1X: 23.76 fps 2X: 15.19 fps 4X: 04.03 fps
Proteus 1X: 23.78 fps 2X: 14.46 fps 4X: 04.69 fps
Gaia 1X: 08.79 fps 2X: 06.24 fps 4X: 04.08 fps
Nyx 1X: 10.02 fps 2X: 07.19 fps
Nyx Fast 1X: 20.30 fps
Rhea 4X: 03.43 fps
RXL 4X: 03.35 fps
Hyperion HDR 1X: 31.14 fps
4X Slowmo Apollo: 35.54 fps APFast: 61.74 fps Chronos: 18.09 fps CHFast: 21.24 fps
16X Slowmo Aion: 49.16 fps
1 Like
Will there be any performance improvement if I replace my motherboard with a new one that has a PCIe-5 slot?
First u need a GPU which supports PCIE 5. For NVIDIA, only the latest 5000 series supports it.
And then I wouldn’t expect anything but a miniscule performance increase, if any at all, since the data traffic to and from the GPU is not such an important part of the whole processing pipeline for TVAI.
Here is a PCIE comparison for LLMs workloads
1 Like
Just imagine, someone is trying to convince another one that a TODAY’s system that starts off at a minimum price tag, 4k stacks, is much better than a 7-year-old system that outperforms the 4K stack setup easily.
What’s even more comical… well, embarrassing to be honest, IMO, you could still increase your performance with that RTX card by just updating your MoBo, CPU, RAM & PSU at a fraction of the cost of these non-upgradeable 4k+ stack things.
I still use tvai 5.5.1 as version 6 UI is not to my liking and I haven’t upgraded to version 7 yet.
My results with 7950x3d and 3080ti looks very similar to yours, and except for 1X upscales (which I pretty much never use) it seems 4080 and 5080/5070ti results are in the +/-10% range for all reported benchmarks above?
Is it so that 4090 and 5090 is the only upgrade for a 3080 series GPU at this moment or is TVAI 5.5.1 just a lot faster than 7.0x?
The results that are within 10% are bottlenecked by memory bandwidth, which is why gpu matters less after a certain point when it comes to upscaling using generative ai models.
Just installed an Asus Tuf 4090 “OG” into my intel system. This is a smaller version than the Tuf they originally released, and the cooling is not as good as my Gigabyte 4090. After running SLM for half the day today on the OG, max GPU temp is 76.6c, max hotspot is 93.7c and max memory temp is 84c. Room has been around 23c all day.
Average temps are 67.5, 78.3 and 80.5 respectively.
Topaz Video AI v7.0.2
System Information
OS: Windows v10.22
CPU: Intel(R) Core™ i9-14900K 47.782 GB
GPU: NVIDIA GeForce RTX 4090 23.576 GB
Processing Settings
device: -2 vram: 0.85 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 39.33 fps 2X: 21.25 fps 4X: 05.68 fps
Iris 1X: 34.74 fps 2X: 22.39 fps 4X: 06.59 fps
Proteus 1X: 37.82 fps 2X: 24.50 fps 4X: 06.22 fps
Gaia 1X: 14.84 fps 2X: 11.07 fps 4X: 05.73 fps
Nyx 1X: 13.58 fps 2X: 10.62 fps
Nyx Fast 1X: 32.03 fps
Rhea 4X: 04.98 fps
RXL 4X: 04.74 fps
Hyperion HDR 1X: 29.37 fps
4X Slowmo Apollo: 42.70 fps APFast: 64.85 fps Chronos: 28.69 fps CHFast: 45.84 fps
16X Slowmo Aion: 32.65 fps
