Nvidia's GeForce RTX 3080 and 3090 could enter mass production in August

[Polycount]

Why it matters: We all know Nvidia's RTX 3000-series GPUs are likely to be revealed in just a few months, but thanks to new reports, we now have a better idea of when the cards might enter full production.

Sources familiar with the matter told Igor's Lab that Nvidia is currently in the Design Validation Test phase with its RTX 3080 and 3090. This information, combined with historical data gathered from past Nvidia launches, prompted the site to create a full production and testing timeline estimate for these highly-anticipated GPUs (see that below).

By the site's calculations, the Founders Edition models of the 3080 and 3090 could enter mass production in mid-August, with a full launch scheduled for the following month of September. Obviously, this timeline is speculative, but it's certainly not far-fetched, and it would line up with other industry predictions and rumors.

And, of course, a September launch would let Nvidia take full advantage of the massive hype surrounding Cyberpunk 2077, which is also launching in September and is set to include out-of-the-box support for a range of RTX features.

Igor's Lab's predicted RTX 3080 & 3090 release timeline.

Igor's Lab's sources also mentioned that the leaked image of an RTX 3080 heatsink -- which we've reported on in the past -- is actually one of two possible heatsink designs for the Founders Edition of the card. Apparently, Nvidia is testing designs from two different OEMs; likely to determine which one offers superior cooling performance.

According to Igor's Lab, these OEMs are not "newcomers" and have worked with Nvidia's engineers in the past. It remains to be seen which design will win out, but it's clear that Nvidia is at least contemplating a more radical look for its newest consumer hardware.

It remains to be seen whether Igor's Lab's predictions will prove accurate, but with September just a few months away, we won't have to wait long to find out.

Image credit: Kiklas, Igor's Lab

Permalink to story.

https://www.techspot.com/news/85658-nvidia-geforce-rtx-3080-3090-could-enter-mass.html

 

[QuantumPhysics]

It's going to be most enjoyable watching the 3000 series downward pressure on the 2000 series.

I am interested to see the pricing on the RTX Titan 3000 vs. the 3080Ti. then I'll make my decision on which to get.

 

[brucek]

As someone who either buys a new generation at launch, or decides to skip it, I'm wondering how production volume works over the life of that generation.

If all customers were like me, they'd have to pre-manufacture all the units in the weeks prior to launch, then have nothing more to build for the next two years.

Obviously not all customers are like me, but I doubt demand is constant over that two years either? Certainly for say phones there is/was a big initial rush of orders due to the way phone subsidies worked from the carriers (customers were entitled to free phone upgrades, and were just waiting for the phones to appear.)

To the extent it is front loaded, what does the fab do the rest of the year? Or is it just switching between big orders from different customers?

 

[kmo911]

Then the price in titan rtx would fall drastic in coming months.
more memory faster pc more ram needed into 4k-16k textures.

 

[Lew Zealand]

Here's something I wonder about early vs late production:

I have a 1060 6GB Gaming X, so a high end model 1060, bought early in the release cycle. That GPU doesn't tolerate undervolting well and really only likes to run stably at stock voltages. OCs OK with additional voltage but no golden sample.

I have a 1050Ti bought later in the prod. cycle and a 1080 bought at the very end after the 2000s were released. Those undervolt like champs, which is necessary as the 1050Ti is a slot-power only model so lower power means more performance. The 1080 is a PNY with a *crap* cooler so the undervolt also helps a lot there as well to keep temps under 75 at 1923MHz.

I wonder if early adopters get slightly cruddier silicon like my 1060, before gradual process improvements are made, or if I just happened to get a 1060 which won't undervolt because it's an outlier.

FWIW I have a very early 1660 Super which also undervolts like a champ but that card also came out after a year of Turing so it's not like the process was immature.

 

[grumblguts]

25% performance boost and people will buy it.
bunch of muppets

 

[brucek]

25% performance boost and people will buy it.
bunch of muppets

I think plenty of people skip generations, so make that 50%+. Even for those buying at the top end, say $1,000, spread over 4 years it's like $20/month. (And most people are buying at prices considerably lower down the line.)

Overall I think gaming is a high-value hobby that gives many hours of enjoyment at little cost compared to alternatives. Someone who goes for an evening out even just once more a month probably spends more.

 

[Squid Surprise]

I think plenty of people skip generations, so make that 50%+. Even for those buying at the top end, say $1,000, spread over 4 years it's like $20/month. (And most people are buying at prices considerably lower down the line.)

Overall I think gaming is a high-value hobby that gives many hours of enjoyment at little cost compared to alternatives. Someone who goes for an evening out even just once more a month probably spends more.

Yes... but aside from "uber-geeks", most people like to go out AND computer game... to compare the prices to each other is disingenuous at best...

I could also say that taking a trip to Tahiti every year would be costlier than a gaming PC....

 

[brucek]

Not disingenuous. You can not be in two places at one time. It is not possible to spend the same three hours of free time in an evening both going out and playing a game. And like I said, if a good gaming setup causes me to choose in favor of the gaming even one more time a month, the financial effect is positive.

 

[Rayneofpayne]

Here's something I wonder about early vs late production:

I have a 1060 6GB Gaming X, so a high end model 1060, bought early in the release cycle. That GPU doesn't tolerate undervolting well and really only likes to run stably at stock voltages. OCs OK with additional voltage but no golden sample.

I have a 1050Ti bought later in the prod. cycle and a 1080 bought at the very end after the 2000s were released. Those undervolt like champs, which is necessary as the 1050Ti is a slot-power only model so lower power means more performance. The 1080 is a PNY with a *crap* cooler so the undervolt also helps a lot there as well to keep temps under 75 at 1923MHz.

I wonder if early adopters get slightly cruddier silicon like my 1060, before gradual process improvements are made, or if I just happened to get a 1060 which won't undervolt because it's an outlier.

FWIW I have a very early 1660 Super which also undervolts like a champ but that card also came out after a year of Turing so it's not like the process was immature.

That is a given,
Early silicon is always worse, before the process matures, meaning there are usually better dies harvested in greater numbers, that is what allows premium models like the super to be released or like the AMD RX 590 that is actually a RX 480+++++.
It's standard practice.
Honestly it's up to you, undervolting does not =more performance however there are only specific circumstances where it is effective where heatsoak causes performance degradation due to throttling.
For reference you can go to TSMC website where they have a calculator per wafer that you can see and calculate the expected premium versus average and worse dies per wafer an a NM. Or you can read the multiple articles from anandtech or watch Jim's videos on Adored TV where they explain this.

 

[redgarl]

If that is the case, don expect a launch before October, if not November.

 

[Squid Surprise]

Not disingenuous. You can not be in two places at one time. It is not possible to spend the same three hours of free time in an evening both going out and playing a game. And like I said, if a good gaming setup causes me to choose in favor of the gaming even one more time a month, the financial effect is positive.

I like the logic... but your numbers don't quite add up... your $1000 is JUST the video card. The entire gaming PC costs a fair amount more - if the GPU is $1000, the PC is probably $2-3000 at the least...

 

[Mr Majestyk]

25% performance boost and people will buy it.
bunch of muppets

Where on earth did the 25% number come from. It's going to be more like 40%-75% and RTX will have a minor hit on performance.

 

[brucek]

The rest of the PC is required anyway for work, at least for me. Anyway, we're kind of in the weeds. I'm sticking by my main point that gaming is an inexpensive hobby compared to many. (Particularly obvious where I live, where golf is common, and head-spinning as far as expense goes).

 

[Lew Zealand]

Honestly it's up to you, undervolting does not =more performance however there are only specific circumstances where it is effective where heatsoak causes performance degradation due to throttling.

As you say, UV is useless on the 1060 because the cooler is a beast, it instead begs to be overclocked for more performance.

Avoiding heatsoak is definitely the case in the 1080 because of the inadequate cooler. UV to 0.9v and running at 1923MHz @72C keeps the heat soak and throttling down as I'm now using ~150W instead of stock ~175W (at 1848MHz and 80C). It'll go to 1974MHz at 0.95v and ~165W but the performance difference IMO isn't worth the temp increase to ~77C. I'd rather keep my silicon unelectromigrated.

A 75W brick wall on the 1050Ti means that every Watt saved by undervolting gets more MHz out of the card. UV is most worth it in strictly power-limited situations.

The UV on the 1660 Super is simply not necessary, it was more to see what the tolerances of Turing were compared to my Pascal cards. Turns out it's just about the same.

 

[Evernessince]

As someone who either buys a new generation at launch, or decides to skip it, I'm wondering how production volume works over the life of that generation.

If all customers were like me, they'd have to pre-manufacture all the units in the weeks prior to launch, then have nothing more to build for the next two years.

Obviously not all customers are like me, but I doubt demand is constant over that two years either? Certainly for say phones there is/was a big initial rush of orders due to the way phone subsidies worked from the carriers (customers were entitled to free phone upgrades, and were just waiting for the phones to appear.)

To the extent it is front loaded, what does the fab do the rest of the year? Or is it just switching between big orders from different customers?

Enthusiasts are only a small portion of the market and as you stated, it's only a fraction of that market that even buy right away. People are buying cards throughout the year. In addition, there's the OEM and botique PC manufacturers.

 

[grumblguts]

Where on earth did the 25% number come from. It's going to be more like 40%-75% and RTX will have a minor hit on performance.

Mine are based on facts not wishfull thinking. where did you get thiss 40 to 75% from ???? the 980 is only half the speed as the 2080 on this graph and how many generations is that.....

 

[grumblguts]

People think they are going to get 75% performance increases are living in a dream world. Its isnt like that anymore Nvidia now knows they can charge you what they want for a tiny % gain and even a reduction in performance when you look at the 1080ti vs 2080

 

[jpuroila]

As someone who either buys a new generation at launch, or decides to skip it, I'm wondering how production volume works over the life of that generation.

If all customers were like me, they'd have to pre-manufacture all the units in the weeks prior to launch, then have nothing more to build for the next two years.

Obviously not all customers are like me, but I doubt demand is constant over that two years either? Certainly for say phones there is/was a big initial rush of orders due to the way phone subsidies worked from the carriers (customers were entitled to free phone upgrades, and were just waiting for the phones to appear.)

To the extent it is front loaded, what does the fab do the rest of the year? Or is it just switching between big orders from different customers?

I would imagine that people buying at launch are a small minority. Most people buy when they feel the need to upgrade(either because they're having problems with their current card, or its performance is falling below what they consider acceptable), or when they build a new rig. In those situations, they might delay for a time, but if they want/need a new computer(or GPU) now and might have to wait for the next gen for six months or more, they'll just get current gen instead. And waiting for weeks or months might not be an option if their current card has just failed and is out of warranty.

And of course, for people who want most bang for their buck, everyone knows that GPU prices tend downwards after release. Even if you know want an Nvidia card, you might want to wait until AMD launches their new cards, because that's when Nvidia will slash prices.

And then you have people who don't want to buy things at launch in the first place, either because AMD is known to have less than stellar drivers at launch, or because they don't trust companies not to **** up(anyone else remember the issues EVGA's 1070 and 1080 cards had at launch?).

And that's the enthusiast market. As Evernessince already pointed out, the majority of market is people who buy prebuilt computers or ask for help from their friends when they need a new computer.

 

[Rjmachine]

People think they are going to get 75% performance increases are living in a dream world. Its isnt like that anymore Nvidia now knows they can charge you what they want for a tiny % gain and even a reduction in performance when you look at the 1080ti vs 2080

You will be surprised. Brand new arch on 7nm TSMC = Big gains.

You sound like someone who bought 2000 series recently.

2000 series never delivered a true generational leap (and was not meant for 12nm to begin with). 3000 series will.

My 1080 Ti will be replaced by 3080/3080Ti/3090 very soon.

 

[OortCloud]

Mine are based on facts not wishfull thinking. where did you get thiss 40 to 75% from ???? the 980 is only half the speed as the 2080 on this graph and how many generations is that.....

Err 2...

 

[Mr Majestyk]

Mine are based on facts not wishfull thinking. where did you get thiss 40 to 75% from ???? the 980 is only half the speed as the 2080 on this graph and how many generations is that.....

The leaks have alreadsy been out, 40% for 3080Ti or I guess 3090 for unoptimized games, 75% for optimized games. Clocks speeds are going to be much higher, 2.2GHz+, IPC improvements, faster memory, new architecture, 40% will be easy. Dream on if you think it's going to be a lame 25%, Big Navi will be 30-50% faster on the second fastest model in the lineup the 6900XT than 2080TI and less than 10% hit when DXR is enabled. 25% would see AMD crush Nvidia this time around and that will not happen.

 

[BoboOOZ]

People think they are going to get 75% performance increases are living in a dream world. Its isnt like that anymore Nvidia now knows they can charge you what they want for a tiny % gain and even a reduction in performance when you look at the 1080ti vs 2080

But now there is competition...
The RDNA graphics in the PS5 boosts to 2.3GHz. Nvidia is working seriously to be sure they keep the performance crown this time.
That's why no Titan, too, it wouldn't do to have one and risk being beaten by AMD.
So performance is significantly up this generation.

 

[grumblguts]

We will wait and see and I will remember to say I told you so when the new cards are less than 25% faster than the top 2080ti
75 % honestly

 

[Rjmachine]

We will wait and see and I will remember to say I told you so when the new cards are less than 25% faster than the top 2080ti
75 % honestly

Even the 3080 will beat 2080 Ti, at 500 bucks less. Over time, 3080 will pull more and more ahead too.

You must be new at this if you think the biggest Ampere consumer card is not going to be more than 25% faster than 2080 Ti hahah.

Rumours suggests "3090" will be 60-90% faster than 2080 Ti and this is also my expectation considering new arch and 7nm/7nm+ TSMC. Huge leap incoming.

Also, AMD actually will bring a top-card this time. Meaning Nvidia needs to deliver a decent bump. You won't see a crappy 1000 -> 2000 series step this time. Mark my words. C ya here in a month or two when we know specs.