values for these gpus

[jason7861]

do you have values for 1070, 1070 ti and 3060 ti please? the correct values

[Etayson]

How much free memory do these cards have on Windows?
For 1070 try -w 30.25 -htsz 28 -t 256 -b 90 -p300
For 1070ti try -w 30.25 -htsz 28 -t 256 -b 114 -p280
For 3060ti try -w 30.25 -htsz 28 -t 256 -b 152 -p210

Due to using the same values -w and -htsz files ...htCPUv0.BIN, ..htGPUv0.BIN will be the same for this GPUs so you can create one time and copy this files to all this GPUs. Only file _g2.BIN you need create for every type of this GPUs.

[jason7861]

do you have telegram?

[spyder403]

what about 4090 the program recommends values to use are these accurate / suggested ?

[Etayson]

Let me know how many free memory do these card have on Windows.

[spyder403]

22.9 I believe

…

On Sun, Jul 23, 2023, 12:38 a.m. Etayson ***@***.***> wrote: free memory do these cards have on Windows? Let me know how many free memory do these card have on Windows. — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AN577KXA2HI6YSMYVOHZ3OLXRTBEXANCNFSM6AAAAAAZ2YGNGU> . You are receiving this because you commented.Message ID: ***@***.***>

[Etayson]

For RTX4090 try -w 31.51 -htsz 30 -t 512 -b 384 -p 360
Requires a lot of memory to generate files.
You can use Onlygen app at PC that have enough RAM and than copy file to PC where GPU installed.
After the files are generated, you only need 16Gb to run searching.

[spyder403]

Thank you

…

On Mon, Jul 24, 2023, 12:05 a.m. Etayson ***@***.***> wrote: For RTX4090 try -w 31.51 -htsz 30 -t 512 -b 384 -p 360 Requires a lot of memory to generate files. You can use Onlygen app at PC that have enough RAM and than copy file to PC where GPU installed. After the files are generated, you only need 16Gb to run searching. — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AN577KWZ4TZ5I5X4VCC2OA3XRYGD3ANCNFSM6AAAAAAZ2YGNGU> . You are receiving this because you commented.Message ID: ***@***.***>

[Maxim-MDProjects]

Hello.
Thanks for the great work. I have 2 questions.

1. where to find the program: onlygen_1_9_7File0.exe or similar
2. why does the source file in your versions (bsgs cuda etc) not change? Could you post the source of the current versions? This will help in product development and finding bugs.
   Thank you

[Etayson]

You are right. Added source files for 1.9.7 version and for onlygen

[Maxim-MDProjects]

I'm still studying your products, such a great job has been done, it's super. I have a question.
parameter -w greater than 31.51 cannot be raised?
It's just that the situation is this, I have 64 RAM or more for example and 12 video memory on the card.
I want to use all the RAM. As I did not try to change the parameters, more than 50 GB could not be used.
What should i change in the code so i can use more RAM?

[Etayson]

-w 31.51 this is the maximum value of the number of х-points in the hashtable. If you do more, then the table will have the same x coordinates of different points. Since the table does not store 256 bits of each x coordinate, but only 32 bits + htsz.
And since the comparison in the hashtable takes used a binary search, cannot allow the same value in the hashtable.
Host RAM is not used for searching, only GPU RAM. Host RAM used when first time generated hashtable. giant table etc.

[Maxim-MDProjects]

-w 31.51 this is the maximum value of the number of х-points in the hashtable. If you do more, then the table will have the same x coordinates of different points. Since the table does not store 256 bits of each x coordinate, but only 32 bits + htsz. And since the comparison in the hashtable takes used a binary search, cannot allow the same value in the hashtable. Host RAM is not used for searching, only GPU RAM. Host RAM used when first time generated hashtable. giant table etc.

I understand that this is a different solution for a video card and a cpu, but still, as the keyhunt implemented support for up to 8 terabytes RAM, hashes and dP there in the bloom filter? Is it possible to do something similar? make a big table like in a keyhunt. And store the largest in RAM, and what is currently being loaded into video memory?
email me please [email protected]

[Etayson]

Keyhunt is for cpu usage. It is very different things cpu & gpu.
Bloom filter for gpu is bad idea - too many memory requests.
Using host memory and make gpu request to it is bad idea - slow down process.
When you are using keyhunt you need a lot of RAM due to using small amount of cpu threads. GPU have a lot of threads but limited memory resource.
For ex, speed 2^60 you can reach at single low cost 1660S with 6GB of GPU memory and how much memory and threads is needed to get the same speed on keyhunt?
The author of the keyhunt has plans to develop for the GPU, you should wait for his decision. Bsgs cuda is for gpu only.

[Alexxino]

Hello @Etayson

• How can I get the right values of threads and pparam for my GPU(RTX 3060 12GB) ? (Any link for this?)

• Also, I'm performing some tests with the puzzle#110 but I got confused with the -pos (position) parameter, how can I calculate it and what values can I set as a position? I just want to see that depending of the position, the key is discovered quicker.

• One last question: bsgs-fractions can be compiled for Iinux too or is it only for Windows? (I have checked in the PureBasic compiler but no option for linux)

Thanks