Any max-packet-size SMALL work units available for GPUs?

[Joe_H]

The "ti" versions do not have an additional GPU chip, just a different one than used on the non-ti version of the card. Sometimes the "ti" version is released at the same time, but more often several months later. In general the "ti" card will have a version of the GPU chip that has more shaders, sometimes more L2 cache, than the non-ti card.

[JimboPalmer]

I often wondered if the 'ti' cards were regular cards, with an additional GPU chip built in?

[This is my speculation]

Lets pretend that we want to make a new GPU (lets call it GP107) if we give it 6 sets of 128 corers each, we would have 640 cores.
If we design it so that if one set of 128 cores is defective, we can by pass it, we can make a working GP107 with 5 sets of 128 cores each.

Early in production we will have many defective chips that can be salvaged by turning off 1/6. We make GTX 1050 out of them. The rare perfect chips become GTX 1050 ti.
As we get more experience making GP107, the ratio of imperfect to perfect rises, and we get chips that have defects in other areas. (memory bus, lowered speed, etc.)

On the GTX 1060 (GP106) there are 10 groups of 128 cores. the GTX 1060 6 gig has all 10 in use, the GTX 1060 3 gig has 9 out of the 10 in use. So besides less memory (not real important for F@H) it has only 90% of the compute power.

[bruce]

People asking questions about which GPU make their decisions about which GPU to use for several reasons (or a combination of reasons)
1) It's a GPU that I already own so the initial investment is zero.
2) My long term costs are based on the price of electricity to run it; initial costs don't matter
3) I'm not going to lay out $X for the newest and most expensive card on the market; I'm going to wait until the prices come down.
4) I want the fastest single card on the market.
5) etc.

[There are spreadsheets available for PPD/initial_cost (item 3) and PPD/Watt (item 2) and some weighted average of the two.]

This topic is based on #1 and that's not a subject for debate. You are entitled to feel differently, but that doesn't make the initial assumption wrong -- just different.

[vahid.rakhshan]

I have the same problem. My GPU is a GeForce GT750 with 4 GB of RAM and its Work Units expire at percentages below 10%. Can you give me some links about how to use this "GPUGRID"?

[JimboPalmer]

https://www.gpugrid.net/

I have never used it.

[vahid.rakhshan]

https://www.gpugrid.net/

I have never used it.

Thanks.

[toTOW]

GPUGRID is very elitist ... it usually requires high end modern GPUs and drops support of older models quite quickly ...