• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA GeForce RTX 3090 Ti to Feature 21 Gbps GDDR6X Memory

Having 12 less VRAM modules should cut down the power consumption by 30W, and no need to cool the VRAM modules on the backside either.

Hopefully these 3090 Ti are LHR :D
 
Hey, no. Bad. Let me stay on the top for a bit.

My VRAM better OC to that same speed, i'll just use a label printer and add the Ti myself.

HAHAH! I WIN! LABEL PRINTER TIME!
View attachment 226996
(+1440 Vram in afterburner, that active backplate gets the job done)
The original 3090 already had 21Gbps chips, just down clocked to 19.5 because most 3090 memory controllers suck.
 
I suppose the real test would be to run in New World-like situations. :cool:
 
This has to be the most pointless card in recent history.
 
The original 3090 already had 21Gbps chips, just down clocked to 19.5 because most 3090 memory controllers suck.
i disagree. there were 24 memory chips each running at 19.5 GT/sec on the og 3090. that would have cause some signalling problems with PCB tracing causing them to decrease the speed. with only half as many chips, the PCB traces should be better and we should be able to hit 21 GT/sec and maybe even OC ot to 23-24 GT/sec on the best models.
 
reduce the price of 3090 and make 3090ti the price of 3090 was. job done
 
Gelid uses the fakest thermal conductivity index in the entire market.
02-Final-bars.png


Reference from 7 to 14 W/mK are rebranded Fujipoly Sarcon aka Alphacool Eisschicht.
They used Gelid 1mm pad which is too thin. 1.5mm is recommended, which they didn’t test with.
 
They used Gelid 1mm pad which is too thin. 1.5mm is recommended, which they didn’t test with.
Less thickness means less thermal constraint: that's why thermal paste will always be better than pad where you can use it. And the thickness is predetermined by the tolerance of the cooler.
Also, comparing pads with different thickness just makes no sense.
 
Last edited:
Hey, no. Bad. Let me stay on the top for a bit.

My VRAM better OC to that same speed, i'll just use a label printer and add the Ti myself.

HAHAH! I WIN! LABEL PRINTER TIME!
View attachment 226996
(+1440 Vram in afterburner, that active backplate gets the job done)
You should also hope it doesn't fry itself or your PC :slap:
 
Nvidia, Nvidia. You're starting to mess your launches like AMD used to do.

I heard about a 3050 somewhere near H2/2022. Dafuq. We should be hearing about your new gen already. Not about the amount of 2060's you'll start glueing together again and your low-end which is obsolete by the time its out. And especially not about more cards that are level unobtanium-Ti...



Nah they'll just start over or maybe we'll get RTX where the R stands for Remastered. Geforce NEXT. RTX Ultra... Its so easy man... just include one or more X'es and all is well.
I dunno, RTXXX ULTRA 3090 Ti, seems like a chinese knockoff. :p
 
I loved the Ultra name. They should bring it back. Tho i suppose these days Titan has replaced that. I mean im not the target audience for it but still. Maybe it's just me but Nvidias marketing in the past few years has been worse than before.
 
Less thickness means less thermal constraint: that's why thermal paste will always be better than pad where you can use it. And the thickness is predetermined by the tolerance of the cooler.
Also, comparing pads with different thickness just makes no sense.
As someone who blew far too much money on thermal pads with his 3090...

the thickness has to vary, if the pads are softer. Softer pads need to be thicker, harder ones thinner. I had a lot of bad performance with some of the harder "better performing" pads because they'd damn well bend the heatsink or push it away preventing proper contact, but a larger softer one to squish into place was fine and gave better temps, despite worse thermal performance

I trust Igor. He may well have found different results by trying other thicknesses, but within the setup he used i'd trust those results
 
Nvidia, Nvidia. You're starting to mess your launches like AMD used to do.

I heard about a 3050 somewhere near H2/2022. Dafuq. We should be hearing about your new gen already. Not about the amount of 2060's you'll start glueing together again and your low-end which is obsolete by the time its out. And especially not about more cards that are level unobtanium-Ti...
From a business side view, there is no incentive at all to release or produce anything but enthusiast-level or halo products. Big margins and it gets sold regardless.
IMO, they are only refreshing the 3000-series (Ampere) to increase said margins for a greater cut of the crypto pie and maybe to have something "new" for OEMs in Q1-2 at the debut of Intel's Arc SKUs.
The RTX 4000-series (Lovelace) and Radeon 7000-series (RDNA 3) are still rumored to be scheduled for Q4 2022.
 
Last edited:
From a business side view, there is no incentive at all to release or produce anything but enthusiast-level or halo products. Big margins and it gets sold regardless.
IMO, they are only refreshing the 3000-series (Ampere) to increase said margins for a bigger cut of the crypto pie and maybe to have something "new" for OEMs in Q1-2 at the debut of Intel's Arc SKUs.
The RTX 4000-series (Lovelace) and Radeon 7000-series (RDNA 3) are still rumored to be scheduled for Q4 2022.

Depends. Nvidia hopefully also does have strategic planning and their roadmaps are several years ahead of reality. High end SKUs only means you have no way to get product segmentation but more importantly, you can barely get your yields in order. The bigger the chip, the higher the risk. Failed chips turn up in lower tier products. Effectively increasing margins over the whole stack.

Burn those bridges and youre also heading to a fast dwindling market share. Todays lower tier buyers are tomorrow's opportunity to have them 'trade up'. The Geforce stack is no longer profitable with only high end chips. Those margins are big because of scale, not because they are cheap or easy to make.
 
so, let me see...... this so far.....

rtx 3060 12gb
rtx 3060 ti 8gb
rtx 3070 8gb
rtx 3070 ti 8gb
rtx 3070 ti 16gb
rtx 3080 10gb
rtx 3080 ti 12gb
rtx 3090 24gb
rtx 3090 ti 24gb


but, where is the rtx 3080 20gb ................ ???
 
so, let me see...... this so far.....

rtx 3060 12gb
rtx 3060 ti 8gb
rtx 3070 8gb
rtx 3070 ti 8gb
rtx 3070 ti 16gb
rtx 3080 10gb
rtx 3080 ti 12gb
rtx 3090 24gb
rtx 3090 ti 24gb


but, where is the rtx 3080 20gb ................ ???
waiting for sales on all those other products to drop, of course

Nvidia want people undecided on which to buy, they get more buys that way
 
waiting for sales on all those other products to drop, of course

Nvidia want people undecided on which to buy, they get more buys that way

How? When everything is sold on launch date and some people on pre-order haven't even gotten their RTX 3080 cards yet.

Read on a danish forum earlier this week that a person finally received his RTX 3080 he had pre-ordered since launch so seriously Nvidia should focus on getting cards out ito all the pre-prdered since crypto mining and the demands outside normal computer and gaming doesn't look to stop which doesn't make anyone feel better that nice a new product that feels like a paper launch since only a small group of people outside reviewer can get a card.
 
How? When everything is sold on launch date and some people on pre-order haven't even gotten their RTX 3080 cards yet.

Read on a danish forum earlier this week that a person finally received his RTX 3080 he had pre-ordered since launch so seriously Nvidia should focus on getting cards out ito all the pre-prdered since crypto mining and the demands outside normal computer and gaming doesn't look to stop which doesn't make anyone feel better that nice a new product that feels like a paper launch since only a small group of people outside reviewer can get a card.
because those people who want a 3080, buy something else in the meantime. They end up buying a second one.

They want you confused between the 3060ti and 3070, because if theres a clear winner or clear popular card, you get the 3080 - sold out everywhere. Better to spread the demand out over barely different cards.
 
because those people who want a 3080, buy something else in the meantime. They end up buying a second one.

They want you confused between the 3060ti and 3070, because if theres a clear winner or clear popular card, you get the 3080 - sold out everywhere. Better to spread the demand out over barely different cards.

No always true, because a lot of people in these crazy times cannot afraid another card because of the reserved money accounted on there account or shops do that here.

Also it's almost impossible to get a Nvidia for a real not fantasy price and if you want AMD go to their own shop you can get lucky getting a reference card for almost MSRP price.

But not a lot of people like the AMD cards even I think the performance is awesome in this generation and you get something for your money.

I can say this because I had 2x RX6800XT cards and now I got a RTX 3090, but sadly I was never able because of availability to get a Sapphire RX 6800/6900 XT Nitro card they are not easy to find in my country and prices ain't good either but I don't regret my trade.
 
2mm Gelid works for all parts of the 3090 founders AFAIK and drops memory temps by 20C. I have a 3080TI (1st world problems) so can't be 100% sure.
 
Back
Top