the desktop 5070ti has 4GB more VRAM and is about 20% faster
stay away from mobile junk they are much slower and 5070 mobile only has 8GB vram
and has similar performance to the now 6yr old rtx 2070 desktop
https://www.techpowerup.com/gpu-specs/geforce-rtx-5070-mobile.c4237

20% more CUDA cores and higher bandwidth memory , usually also a bit more memory on board as well (12 vs 16 for the ti variant)

plus a few more minor changes , it averages about 15+% more real world performance. (though they state 20-30%)

Mobile versions are always slower due to them trying to keep the power down so it doesn't act like a space heater when running.(it's just best to stay away from , unless you really want one , then buy the highest one)

everything
https://technical.city/en/video/GeForce-RTX-5070-vs-GeForce-RTX-5070-mobile

gpu core, tmu, rop counts
core, ram clocks
ram bit bus width
ram amount
even the gpus l1/12 cache on the cores
power and cooling depending on the laptops board and cooling designs

figure the mobile version is about 1/3 slower than the desktop card, maybe more if its in a weak laptop

It's a gimmick, you wont stronger than 5080 but with a price of 5080, dont change my mind.

Nvidia Ti means better version that 5% ~ 20% in terms of performance, this includes specs higher than non Ti version.

Nvidia Super is just a refresh of the cards, meaning more performance, higher spec over previous version. There are Ti super as well.

For AMD it they have non xt, and xtx, where XT is basically Ti, and XTX is full flag ship GPU.

Anyways if you game at 1440p or higher get 5070 ti, it come with 16GB vram as that be important to you for future games.

ti used to mean more or faster cores or more tmus/rops

super used to mean faster ram

ti+super = more and faster core parts, with faster ram


anyhoo, ti is slightly better than vanilla

引用自 _I_：

ti used to mean more or faster cores or more tmus/rops

And more VRAM, also for a short time Ti was a slightly cut down Titan.

引用自 Shadow-79：

They're both considered RTX 5070 but one is the "normal" 5070 and the latter is 5070Ti. Despite that GPU having the exact same name, is there really any difference regarding performance between a desktop / mobile 5070 & desktop / mobile 5070Ti?

You're thinking of it backwards. You're approaching it as though "it has the same number so it's the same thing with minor differences" but in reality the name is just a name. They don't have the same exact name. One is a RTX 5070 and one is an RTX 5070 Ti.

History backstory

Traditionally, the "Ti" suffix has been used to denote a mid-generation refresh that was a bit better than the non-Ti of the same number.

In recent generations, for reasons that would warrant their own thread to cover, it's shifted to the Ti merely being a launch SKU that is "every other tier" (and the Super is the refresh now).

In other words, instead of an x50, x60, x70, and x80 launching, with some getting Ti refreshes... it's x50 (two versions like the old x30), x60, x60 Ti, x70, x70 Ti, x80, and x90 (sometimes x90 Ti) as launch models. And then some of them get Super refreshes. Of the three RTX 40 series refreshes that launched, two of them did not even improve the chip of their respective non-Super variant. The RTX 4070 Super, the lowest one, was the only one that did (and it did because the 7800 XT was giving its non-Super predecessor heavy competition). The RTX 4070 Ti Super just added VRAM, although it deserved it, and the RTX 4080 Super did... nothing. Its performance difference is literally margin of error compared to the non-Super. It was mostly just a price drop, but the new name makes it sound more interesting than just saying "we're dropping the RTX 4080 price 20%". Which, again, was needed because the RTX 4080 was originally a mediocre option.

So the structure has changed a lot in the last few generations, and a Ti isn't what it once was.

End of story.

Getting back to the RTX 5070 and RTX 5070 Ti, they simply aren't the same chip. The RTX 5070 Ti is actually closer to the RTX 5080 than it is to the RTX 5070.

Mobile parts are also different entirely. Mobile devices have lower power and thermal limitations, meaning those chips use much less power (meaning less performance) and sometimes they will also be smaller chips in addition to using less power. The result is a mobile chip with a 5070 number that isn't close to the same thing as a desktop part with that same number.

引用自 wing0zero：

引用自 _I_：

ti used to mean more or faster cores or more tmus/rops

And more VRAM, also for a short time Ti was a slightly cut down Titan.

every gt/gtx gpu is a cut down titan

引用自 _I_：

引用自 wing0zero：

And more VRAM, also for a short time Ti was a slightly cut down Titan.

every gt/gtx gpu is a cut down titan

GTX use a cut down version, Ti was a cut down Titan, same die, that's what I mean but then you knew that...

all from the same die/wafer for the series

binning determines if its a titan or what tier/trim including titan, ti, super or mobile

引用自 _I_：

all from the same die/wafer for the series

binning determines if its a titan or what tier/trim including titan, ti, super or mobile

Yeah that's not even remotely true, Google is your friend.

引用自 wing0zero：

引用自 _I_：

all from the same die/wafer for the series

binning determines if its a titan or what tier/trim including titan, ti, super or mobile

Yeah that's not even remotely true, Google is your friend.

do you really think they plan on making a wafer full of mobile gpus, or intel makes cpus missing the igpu components just for f cpus or pentium g?


the wafer is clipped into dies, the dies are tested and binned
near perfect are titan, and flawed or lower performing go down the line

same with cpus, thats how some intels with defective igpu get the f suffix
cores/cache clipped and glued together

Search "5070 / 5070ti TPU" in google. Click on techpowerup's site. There you find an extensive table that allows you to compare the differences between different GPUs. No need to rely on strangers again.

引用自 _I_：

引用自 wing0zero：

Yeah that's not even remotely true, Google is your friend.

do you really think they plan on making a wafer full of mobile gpus, or intel makes cpus missing the igpu components just for f cpus or pentium g?


the wafer is clipped into dies, the dies are tested and binned
near perfect are titan, and flawed or lower performing go down the line

same with cpus, thats how some intels with defective igpu get the f suffix
cores/cache clipped and glued together

Again actually Google and read instead of just saying words you think are true, show one piece of evidence to back your claim, I will show you a boat load to back mine.
Yeah they have different runs for different dies, that's how they get different cache sizes or different memory controllers ect, that's why the dies are a different size, you can't just cut a die in half and have a 5060 you just have a broken die.

Also why bring CPUs in to this, now they are done a bit different to GPUs and you know this and trying to use some simple tactic to subvert the main thing, that kind of thing doesn't really get by me.

Like I said Google is your friend, read and learn instead of just saying words you think are true,.