*RUMOUR* 4080/4070 Ti/4070 Super cards rumoured to draw 320/285/220W power and more specs recently released (10th nov)

[filpo]

Summary

Quotes

Quote

Recent reports regarding the RTX 40 SUPER cards suggest that the new models may not necessarily demand more power, at least for some of them. Kopite7kimi states that the power consumption values for the RTX 4080 SUPER and RTX 4070 Ti SUPER remain unchanged (it’s worth noting that there is still some debate about the latter SKU’s name). When referring to power not changing, this refers to the original non-SUPER variants. For the RTX 4080 SUPER, this implies a 320W TDP, while the RTX 4070 Ti SUPER is expected to feature a 285W TDP. The only exception is the RTX 4070 SUPER, which would see a 20W increase in power.

NVIDIA GeForce RTX 4080 SUPER Specifications (Rumored)

Quote

Starting at the top, we have the NVIDIA GeForce RTX 4080 SUPER which is rumored to use the AD103-400 GPU with the full 10,240 cores, 320 TMUs, 112 ROPs, and 64 MB of L2 cache. The chip will use a 256-bit bus interface which would suggest a 16 GB GDDR6X memory, same as the RTX 4080 Non-SUPER. So that suggests that NVIDIA has both 20 GB (AD103 / 320 bit) and 16 GB (AD103 / 256-bit) SKUs but what makes it to retail remains to be seen. The NVIDIA GeForce RTX 4080 Non-SUPER has a TBP of 320W so we can expect similar figures for the 4080 SUPER.

NVIDIA GeForce RTX 4070 Ti SUPER Specifications (Rumored)

Quote

Next up, we have the NVIDIA GeForce RTX 4070 Ti SUPER which Kopite7kimi believes is a really weird naming choice and to be honest, it kind of is. But as for the specifications, the card is said to feature AD103-275 / AD102-175 GPU SKUs with the same 8,448 cores across both variants and 48 MB of L2 cache. The GeForce RTX 4070 Ti SUPER should have a TBP close to the 285W mark of the existing 4070 Ti.

NVIDIA GeForce RTX 4070 SUPER Specifications (Rumored)

Quote

Lastly, there's the NVIDIA GeForce RTX 4070 SUPER which is said to come in AD104-350 or AD103-175 SKU configurations with the same 7,168 cores across both variants. The GPU is expected to get 48 MB of L2 cache. It looks like this card would stick with the 12 GB GDDR6X memory though we can expect NVIDIA to push the memory speeds up using better Micron DRAM which can reduce the stress on the lower 192-bit bus interface. For the RTX 4070 SUPER, the card will have a 20W higher TBP than the existing 4070 at 220W.

My thoughts

Obviously it's just a small bump in performance from the 4080 to a 4080 super but it's good to see that there's not a significant increase in power consumption. Great to see that the 4070 Ti super is rumoured to have 16gb of VRAM. Also the 4070 super looks more like a 4070 Ti than a 4070 super

 

Sources

NVIDIA GeForce RTX 4080/4070Ti/4070 SUPER GPUs rumored to require 320/285/220W power - VideoCardz.com

NVIDIA GeForce RTX 40 SUPER GPUs To Feature Similar Power Consumption As Non-SUPER Cards (wccftech.com)

Edited November 10, 2023 by filpo

[venomtail]

Still keeping the mass market model, most likely the 4070S on a leach, still within the milking aura. Guess they can't give the card 16GB incase it risks the consumer not prematurely upgrading to the 16GB standard. Disappointing.

[starsmine]

I have to agree with GN evaluation.

Nothing solid in these leaks other than super exists. 

Even Kopite is telling you the numbers as presented are tentative. 

[Stahlmann]

Same wattage and same architecture, just a few more CUDA cores. If all they offer is a 5-8% performance bump to keep the selling price high, then i really hope this flops. I hate the fact that voting with our wallet doesn't work anymore because their other products keep their numbers nice and inflated. It'll be interesting to see if i will ever be able to (significantly) upgrade my 3080 10G without spending more than $700. So far it seems that every performance jump comes with an equally high price hike.

 

I know it sounds very naive but i still have hope that the 5000 series will reset this whole GPU situation and go back to 3000 series launch pricing (before scalping, obviously). They don't really have any incentive to do that, but one can still hope. At the 3000 series launch no one expected them to make such a big price/performance jump.

[porina]

3 minutes ago, Stahlmann said:

Same wattage and same architecture. So how exactly are they supposed to deliver more performance? If all they offer is a 5% performance bump to keep the selling price high, then i really hope this flops. 

More cores at the same power limit gives an increase in perf as they're operating at a more efficient point. That's assuming same binning, they could get a bit more out if bins are better.

 

If we look back in history to the last time Super was used, the Super cards tended to be close to the Ti but priced closer to the base model. It is a way of addressing both competitive performance and pricing at the same time. Look at pricing some weeks after launch and they should be better value than their predecessors. I'm specifically excluding launch since sellers tend to price high to ride the hype train before adjusting to market conditions. To say you hope this flops is saying you don't want better value.

[starsmine]

10 minutes ago, Stahlmann said:

Same wattage and same architecture, just a few more CUDA cores. If all they offer is a 5-8% performance bump to keep the selling price high, then i really hope this flops. I hate the fact that voting with our wallet doesn't work anymore because their other products keep their numbers nice and inflated. It'll be interesting to see if i will ever be able to (significantly) upgrade my 3080 10G without spending more than $700. So far it seems that every performance jump comes with an equally high price hike.

 

I know it sounds very naive but i still have hope that the 5000 series will reset this whole GPU situation and go back to 3000 series launch pricing (before scalping, obviously). They don't really have any incentive to do that, but one can still hope. At the 3000 series launch no one expected them to make such a big price/performance jump.

Its a pure mid generation value gain. Same price, more performance, same electricity cost. 

Its not delaying a new architecture, its to not have AMD grab to much market share with the 7900xt @ 750USD.

[Somerandomtechyboi]

Well a not vramless 4070ti super that i can reccomend and a vramless 4070 super that i can reccomend if pricing isnt garbage

 

And why the hell has the 4070 been upgraded to near 4070ti in terms of cuda cores?

[porina]

2 minutes ago, Somerandomtechyboi said:

And why the hell has the 4070 been upgraded to near 4070ti in terms of cuda cores?

If the info is correct, it's 93.3% of a 4070 Ti. The 4070 is 76.7%. A good bump, but still not a Ti. 

 

Again if we look at historic Super models, they tended to get close to Ti performance for much less cash, so it is an overall improvement.

[Stahlmann]

13 minutes ago, starsmine said:

Its a pure mid generation value gain. Same price, more performance, same electricity cost. 

Its not delaying a new architecture, its to not have AMD grab to much market share with the 7900xt @ 750USD.

I've grown to love DLSS, Reflex and Nvidia's superior ray tracing performance. Unless AMD can significantly improve FSR and their low-latency mode in addition to a ~ 50% performance boost (in raster and ray tracing) i don't see myself upgrading to an AMD card. We'll see how the true next generation stacks up, but AMD has it's work cut out.

[HenrySalayne]

17 hours ago, porina said:

If we look back in history to the last time Super was used, the Super cards tended to be close to the Ti but priced closer to the base model. It is a way of addressing both competitive performance and pricing at the same time.

The last time "Super" was used, there were no Tis around.

The performance uplift was not uniform. The 2060S was around 5% slower than the 2070, the 2070S became basically the new 2080 and was around 5% slower. The 2080S could not at all bridge the gap to the 2080 Ti and was maybe 5% faster than the 2080. All these cards came with 8 GB of RAM.

 

This actually made th 2060S and the 2070S solid upgrades and massively improved the generally bad value proposition of this generation.

The 2080S struggled with a small performance gap in comparison to the 2070S and 2080, but a considerably higher price.

 

[porina]

1 hour ago, HenrySalayne said:

The last time "Super" was used, there were no Tis around.

The performance uplift was not uniform. The 2060S was around 5% slower than the 2070, the 2070S became basically the new 2080 and was around 5% slower. The 2080S could not at all bridge the gap to the 2080 Ti and was maybe 5% faster than the 2080. All these cards came with 8 GB of RAM.

It was a long time ago. Looking it up again using TechPowerUp rankings, relative to 1660 Super the 1660 was -8% and the Ti was +6%. It must have been that I was remembering.

 

I agree there weren't many other Ti models, but in general we can look at the next GPU up in the line.

1650 Super, 1650 was -26%. 1660 was +11%

2060 Super, 2060 was -11%, 2070 was +1%

2070 Super, 2070 was -12%, 2080 was +8%

2080 Super, 2080 was -5%, 2080 Ti was +15%

 

I stick by my generalisation even if the "Ti" part was slightly off. Perf went up, prices generally stayed same or dropped, value increased.

[StDragon]

Pricing won't change for the existing lineup. These new Super cards will fill new pricing tiers that's currently vacant. So expect the perf:dollar ratio to remain relatively intact.

[porina]

8 hours ago, StDragon said:

Pricing won't change for the existing lineup. These new Super cards will fill new pricing tiers that's currently vacant. So expect the perf:dollar ratio to remain relatively intact.

It's pretty uncommon for MSRP to be changed for an identical product, but market pricing does vary over lifetime.

 

Putting the new models in context of surrounding models, there are gaps MSRP could go. I still think if Super happens this time, it'll be similar to last time and used to increase perceived value vs competition. Doesn't mean beat competition in value, just narrow the gap a bit. Nvidia still has significant brand value and wont give it away.

 

$299 4060

$399 4060 Ti 8GB

$499 4060 Ti 16GB

$599 4070

$??? 4070 Super

$799 4070 Ti

$??? 4070 Ti Super

$1199 4080

$??? 4080 Super

$1599 4090

[starsmine]

 

16 hours ago, StDragon said:

Pricing won't change for the existing lineup. These new Super cards will fill new pricing tiers that's currently vacant. So expect the perf:dollar ratio to remain relatively intact.

considering their biggest marketing threat at the high end is a 750USD 7900XT, they are not going to maintain the same price performance with the super line. a hypothetical 700 USD 4070 super is not going to be worth much when the 750 USD card from AMD is still better than their 1200 USD one in raster, and better then the 800 dollar one in simple ray traced games.

They need to lower price/performance to get those sales back to them. Otherwise, this refresh would do literally nothing. Their whole existence and being a refresh being worth having necessitates being launched near the msrp of the nonsuper varients. 

 

[SADS]

oh no... not the super moniker again.

Ti and Super...in one name... insanity.

[williamcll]

what about the rumors of a RTX 4080 Ti earlier?

[filpo]

Summary

 

 

Quotes

Quote

Kopite7kimi, a hardware leaker with a solid track record when it comes to NVIDIA leaks, claims that the company intends to provide an official update on these SKUs at the CES. This annual consumer electronics expo is held in Las Vegas, U.S., and has seen better times before COVID-19. The next CES will be a major return to the physical event, and all major PC hardware manufacturers are expected to participate.

 

Quote

NVIDIA is not a ‘featured’ exhibitor, but the company will be present and already has three conference rooms booked. However, there is no sign of any major media event yet. The RTX 40 SUPER may not need a major public event, the company can just as well focus partner designs, new technologies or the main topic for next year’s CES, which is artificial intelligence.

 

Quote

According to rumors, NVIDIA is expected to launch at least two new cards, possibly three. Each model is to feature more CUDA cores and in some cases more memory compared to existing models. With this release, NVIDIA will solidify its portfolio in the upper high-end segment.

My thoughts

Nice to see that these cards are coming so soon. Though quite confusing to have a Ti and a SUPER in the same name for one of the cards. But assuming they keep these cards at the same msrp as their non SUPER variants or only about $50 more then it'll be a decent buy according to the rumoured spefcs

 

Sources

NVIDIA's RTX 40 SUPER series reportedly making its debut at CES 2024 - VideoCardz.com

NVIDIA GeForce RTX 40 "SUPER" GPUs Rumored To Be Unveiled At CES 2024 (wccftech.com)

SUPER cards rumoured specs on my other thread:

[CommanderAlex]

Why not update the original post you created?

[filpo]

2 minutes ago, CommanderAlex said:

Why not update the original post you created?

Let's say a mod just helped me

[TVwazhere]

Just now, filpo said:

Let's say a mod just helped me

Hi, I'm mod.

 

-topics merged-

[CommanderAlex]

1 minute ago, filpo said:

Let's say a mod just helped me

Much better!!! Keeping it together helps especially on the same topic. Makes sense these cards will launch at CES after the holidays and mid-cycle product launch. 

[starsmine]

honestly, the biggest tradgey from this leak is the lack of 4060 super and 4060ti super. 

The fact that there has been ZERO price performance uplift at that price segment since the launch of the 3060 12gb that is still selling new for 300 USD. is a tragedy from Nvidia, and if there is none before the the launch of the 5060, thats fucked up. because that would be 4 years of stagnation at the 300-400 USD price bracket from Nvidia.
 

[filpo]

THESE ARE ALL LEAKED

Summary

We've got updated leaked specs of the 40 series SUPER cards from WCCFtech

All these cards are expected to be revealed at Nvidia's CES show

 

Quotes

Nvidia GeForce RTX 4080 SUPER Specs and Expected Performance

Quote

Starting at the top, we have the NVIDIA GeForce RTX 4080 SUPER which is expected to use the AD103-400 GPU (PG139 SKU 355) with the full 10,240 cores, 320 TMUs, 112 ROPs, and 64 MB of L2 cache. The chip will use a 256-bit bus interface which would suggest a 16 GB GDDR6X memory, same as the RTX 4080 Non-SUPER. The NVIDIA GeForce RTX 4080 SUPER will feature the same TGP as the RTX 4080 Non-SUPER at 320W

 

In terms of performance, the graphics card is expected to be 3-5% faster than the RTX 4080 Non-SUPER which isn't a huge gain but custom models will deliver close to 10% performance boosts. NVIDIA hasn't decided on the pricing of the SUPER series lineup yet but this small performance boost should help the card be faster than the Radeon RX 7900 XTX and if it comes at anything around $1000-$1100 US, then that ends up being very competitive.

Nvidia GeForce RTX 4070 Ti SUPER Specs and Expected Performance

Quote

Next up, we have the NVIDIA GeForce RTX 4070 Ti SUPER which will feature a choice of AD103-275 / AD102-175 GPU SKUs (PG141 SKU 323) with 8,448 cores across both variants & 48 MB of L2 cache. The graphics card will also feature 16 GB of GDDR6 memory which will be a 4 GB VRAM upgrade vs the RTX 4070 Ti Non-SUPER and support a 256-bit bus interface. The graphics card will retain a 285 Watt TGP which is the same as the Non-SUPER variant.

 

For performance, the NVIDIA GeForce RTX 4070 Ti SUPER 16 GB GPU will offer an average 15% boost in performance over the Non-SUPER variant and should compete well against the Radeon RX 7900 XT at a possible lower price point.

Nvidia GeForce RTX 4070 SUPER Specs and Expected Performance

Quote

Lastly, there's the NVIDIA GeForce RTX 4070 SUPER which is said to come in AD104-350 or AD103-175 SKU (PG141 SKU 335) configurations with the same 7,168 cores across both variants. The GPU is expected to get 48 MB of L2 cache. It looks like this card would stick with the 12 GB GDDR6X memory across the same 192-bit bus interface. For the RTX 4070 SUPER, the card will have a 20W higher TBP than the existing 4070 at 220W.

 

As for the performance, the NVIDIA GeForce RTX 4070 SUPER should end up being around 14% faster than the RTX 4070 Non-SUPER. The graphics card will also get the new CEM 5.1 Gen5 power connector. The existing RTX 4070 graphics cards either come with an 8-pin connector or a 12-pin power connector (supplied with an adapter).

Overall expected performance increase over non-SUPER cards

Quote

RTX 4080 SUPER vs RTX 4080: +3-5% Faster On Average

 

RTX 4070 Ti SUPER vs RTX 4070 Ti: +15% Faster On Average

 

RTX 4070 SUPER vs RTX 4070: +14% Faster On Average

 

4090 SUPER? What we know so far

Quote

We have also come to know that there were indeed plans for a GeForce RTX 4090 SUPER though those were dropped due to the recent sanctions which limit the release of GPUs hitting over 4800 TOPs in China and other regions. Currently, there are no plans to do a RTX 4090 SUPER or Ti GPU but that might change if restrictions soften up in the coming year (which seems unlikely at the moment).

 

My thoughts

Loving the fact that the '4070 Ti SUPER' (though that's still a silly name) is gonna get 16gb of ram and the rumour of a 4090 SUPER could be interesting too

 

Sources

NVIDIA GeForce RTX 40 SUPER GPU Specs & Performance Leak: 4080 SUPER With Bigger AD103, 4070 Ti SUPER With AD103 & 16 GB, 4070 SUPER With AD104 & Gen5 Connector (wccftech.com)

NVIDIA RTX 4080 SUPER and RTX 4070Ti SUPER to feature AD103 GPU, RTX 4070 SUPER gets AD104 - VideoCardz.com