Radeon RX480 faster than a GTX 1080Ti?
21 May 2017 7:19:08 UTC
Topic 207775
(moderation:
Is it normal that my Sapphire RX480 is faster than a GTX 980 and 1080Ti for the app Gamma-ray pulsar binary search #1 on GPUs v1.18? The RX480 takes about 700-750 seconds for a task, while GTX 980 ~1600 seconds and the GTX 1080Ti ~900 seconds.
Is there a difference between NVIDIA and ATI applications?
https://einsteinathome.org/it-it/host/12441031/ -> GTX 980
https://einsteinathome.org/it-it/host/12321953/ -> GTX 1080Ti
https://einsteinathome.org/it-it/host/12525172/tasks/ -> my RX480
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz
[img]http://boincstats.com/signature/-1/user/3515553/sig.png[/img]
There is a difference. I do
)
There is a difference. I do about 15 minutes (900 sec) on a GTX1060 little OC'd when my screen is in standby.
Did you leave a core free on your CPU to support your GPU? It needs 1 core free for max performance.
When comparing times, please
)
When comparing times, please keep in mind that you may run two tasks at a time on one card, which saves a bit of time.
Yes, I have the CPU
)
Yes, I have the CPU percentage set on 75% so the GPU app has one core free. However, I'm so satisfied of my GPU
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz
Alessio Susi wrote:Is it
)
I belive you not see the entire picture, for example the 1080Ti you pick to compare, runs 3 WU at a time, so it´s actualy crunch a WU in about 330 secs! A lot faster than the 700 secs you say your 480 is doing. Since i not know how many WU/time you run to be sure if the 700 is for one or 2 WU, but seems the 1080Ti it´s a lot faster.
In my case crunching 2 WU/time the 1070 crunches a WU in 1100 secs (550 secs per WU) and the 1080 (not Ti) crunches in 930 secs (for 465 secs/WU). So the Ti model is a lot faster than both but uses a lot more power to do that.
Ok ok. I run only 1 WU at
)
Ok ok. I run only 1 WU at time, so 700 seconds is the real time it takes to complete a task. However, thanks for the explanation ;P
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz
I find the rx-480 on linux
)
I find the rx-480 on linux (running three tasks) average time is ~1440s / 3 = ~480s
You should be able to make more from your rx-480 running x2 (0.5) or x3 (0.33)
Thanks Juan, I haven't been
)
Thanks Juan,
I haven't been in home to look at this thread. Juan is correct that 1080Ti was running 3 at a time. However, it's run into an issue that I am trying to address (either the OpenCl here at Einstein or the Nvidia Driver) It's starting to look like a Einstein issue. That computer is on time out by the server here, but should get some more work tonight. Should know more later this week.
Zalster
You have an Intel 4650K and
)
You have an Intel 4690K and the RX480 8GB. I have an AMD A10-7800 and RX480 4GB. Maybe your takes less time because of more CPU power.
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz
Alessio Susi wrote:Maybe your
)
In my experience, the effect is quite small. The elapsed time for a GPU task seems to be relatively independent (within reason) of the CPU that supports it.
You will see a bigger improvement in the average time per task just by running two concurrently. You may even get a further gain by running three. With AMD GPUs, you wont have to have a 'free' core for each GPU task. The only way to find out for sure what works best for you is to do proper experiments with different settings.
Cheers,
Gary.