https://cdn.theatlantic.com/assets/media/img/photo/2019/03/brazil-carnival-2019-photos/c01_1128576841/main_1200.jpg?1551729661

That is just cool,wish I had the solder skills to do that.

Caesar:

https://cdn.theatlantic.com/assets/media/img/photo/2019/03/brazil-carnival-2019-photos/c01_1128576841/main_1200.jpg?1551729661

What the hell?

With current price of gpus, i would make sense, make them modular and enable at least memory chip upgrading replacement and adding.. we already had possibility to add more memory until 1997 or something like that.. and why no make gpu replacable too? So videocard would be something like 2nd level motherboard. It would be even more ecological too.. but wait, profits would be less.. so Nvidia or ATI would not ever enable it.. it would ends like modular cell phones in /dev/null

Hehe thats crazy! But i LOVE IT!

ruthan:

With current price of gpus, i would make sense, make them modular and enable at least memory chip upgrading replacement and adding.. we already had possibility to add more memory until 1997 or something like that.. and why no make gpu replacable too? So videocard would be something like 2nd level motherboard. It would be even more ecological too.. but wait, profits would be less.. so Nvidia or ATI would not ever enable it.. it would ends like modular cell phones in /dev/null

I'm not sure such cards would fit anymore the physical constraints of typical mobos and cases. Video cards from this decade and the 90's are quite different. Energy consumption and thus the heat produced have multiplied, which dictates a massive cooling solution must be employed, for both the GPU and the memory, preferably also for the regulators. As it is, the mainstream video cards must be both resilient enough and affordable enough to attract buyers. BGA is one way to make both the cooling optimal and to reduce production costs while ensuring quality. In fact every now and then there has been industry talk of wanting to switch to BGA, and the things going along with it, with the CPU+mobo combination as well, but it has encountered resistance due to historical reasons, even if it was, in fact, all the same maybe to most PC buyers. I've gone through quite a few PCs myself, as I wasn't born yesterday, but I've never only changed the CPU. I've always got a mobo and a CPU when building a new PC. Although I've messed with the memory more.

In fact, what they did was not take the HOF GDDR6 and get into a 2080Ti. They took the 16Gbps GDDR6 from the new 2080 Super, and put in a HOF 2080Ti 2080Ti comes with 14Gbps GDDR6, even the HOF model

I thought that with all that RTX speed, she/he would be naked by now...

Here in Brazil we call this a "gambiarra".

So they took the 16Gbps memory off one 2080 Ti, and put it on another 2080 Ti?

jbscotchman:

So they took the 16Gbps memory off one 2080 Ti, and put it on another 2080 Ti?

No, they did what DarknightOCR said, took memory from a 2080 Super HOF and put it on a 2080Ti.

Solfaur:

No, they did what DarknightOCR said, took memory from a 2080 Super HOF and put it on a 2080Ti.

Then I'm confused because this is what Hilbert's article says: "So here's what three musketeers in Brazil did, they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame, and then planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR6. Surprisingly enough, that actually worked."

Solfaur:

No, they did what DarknightOCR said, took memory from a 2080 Super HOF and put it on a 2080Ti.

That's not what the article says. "Brazil did, they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame, and then planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR' So either the article needs to be corrected or this was a complete waste of time.

Solfaur:

No, they did what DarknightOCR said, took memory from a 2080 Super HOF and put it on a 2080Ti.

Nope: "Brazil did, they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame, and then planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR" Either the article is wrong or they wasted their time.

HeavyHemi:

So either the article needs to be corrected or this was a complete waste of time.

The article got it all wrong, Hilbert is probably tired. They took memory from 2080 Super (two of them) and placed it on 2080 Ti HoF. The white HoF card is the TARGET, not the source. https:// youtu.be/UqlKrGFmxKY?t=478 (remove the space before the "y") In any case, those are some mad soldering skills !

wavetrex:

The article got it all wrong, Hilbert is probably tired. They took memory from 2080 Super (two of them) and placed it on 2080 Ti HoF. The white HoF card is the TARGET, not the source. https:// youtu.be/UqlKrGFmxKY?t=478 (remove the space before the "y") In any case, those are some mad soldering skills !

Yes I know this, I watched the video. It's more time consuming and having the proper equipment. Solder mask is the key for the IC. Having worked for then ran a assembly and test division for a major semi firm... It just shows you how much time automation saves you versus building up dut boards by hand. Ha...

I love the smell of flux and BGA soldering in the morning...;)

Hello I'm Ronaldo from Teclab! This is not correct, I will explain: So here's what three musketeers in Brazil did, they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame, and then planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR6. Surprisingly enough, that actually worked. "they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame" - no, we removed the 14Gbps memory (11 chips), from that card. "planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR6" - no, we planted the GDDR6 16Gbps, that we got from 2 x RTX 2080 SUPER (that has fast memories 16Gbps), and planted in the RTX 2080Ti HOF. Summary: We got 11 memory chips, from 2 brand new RTX 2080 Super (EX and WFF series), and put that memory modules in the RTX 2080Ti. The card have a good improvement, and we reached excellent score (it's even world record, and more 300 points higher the max score we get before the modding). Like I said in the video, we believe the GPU Memory Controller should be limited, but even this way, we noticed a nice increase in the performance. People need to understand we made a lot of experiences with the videocards (GTX 1050 4GB (desktop), GTX 1060 3GB to 1070 8 GB, GTX 1060 into a GTX 1070 HOF body, cards running with memory Micron and Samsung together, zombies and hardmoddings). Thanks for watching.

O TECLAB É FERA UM GRANDE ABRAÇO AO RONALDO E PESSOAL QUE ESTÁ NESSE PROJECTO BRAZIL PORTUGAL ESTAMOS JUNTOS

rbuass:

Hello I'm Ronaldo from Teclab! This is not correct, I will explain: So here's what three musketeers in Brazil did, they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame, and then planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR6. Surprisingly enough, that actually worked. "they removed 16 Gbps GDDR6 from a Galax RTX 2080 Ti Hall of Fame" - no, we removed the 14Gbps memory (11 chips), from that card. "planted that same GDDR6 memory on a GeForce GTX 2080 Ti which normally has 14 Gbps GDDR6" - no, we planted the GDDR6 16Gbps, that we got from 2 x RTX 2080 SUPER (that has fast memories 16Gbps), and planted in the RTX 2080Ti HOF. Summary: We got 11 memory chips, from 2 brand new RTX 2080 Super (EX and WFF series), and put that memory modules in the RTX 2080Ti. The card have a good improvement, and we reached excellent score (it's even world record, and more 300 points higher the max score we get before the modding). Like I said in the video, we believe the GPU Memory Controller should be limited, but even this way, we noticed a nice increase in the performance. People need to understand we made a lot of experiences with the videocards (GTX 1050 4GB (desktop), GTX 1060 3GB to 1070 8 GB, GTX 1060 into a GTX 1070 HOF body, cards running with memory Micron and Samsung together, zombies and hardmoddings). Thanks for watching.

Yes, the article just needs to be corrected to reflect the correct GPU model that was cannibalized for the RTX 2080 Ti.