Friday, October 23rd 2020
NVIDIA Readies New GeForce RTX 30-series SKU Positioned Between RTX 3070 and RTX 3080
Possibly unsure of the GeForce RTX 3070 tackling AMD's Radeon RX 6000 series parts, NVIDIA is designing a new RTX 30-series SKU positioned between the RTX 3070 and RTX 3080. This is not a 16 GB variant of the RTX 3070, but rather a new SKU based on the 8 nm "GA102" silicon, according to a reliable source with NVIDIA leaks, kopite7kimi. The SKU is based on the GA102 with the ASIC code "GA102-150-KD-A1." The silicon is configured with 7,424 CUDA cores across 58 streaming multiprocessors (29 TPCs), 232 tensor cores, 232 TMUs, 58 RT cores, and an unknown number of ROPs. According to kopite7kimi, the card is configured with a 320-bit wide memory interface, although it's not known if this is conventional GDDR6, like the RTX 3070 has, or faster GDDR6X, like that on the RTX 3080.
NVIDIA recently "cancelled" a future 16 GB variant of the RTX 3070, and 20 GB variant of the RTX 3080, which is possibly the company calibrating its response to the Radeon RX 6000 series. We theorize that doubling in memory amounts may not have hit the desired cost-performance targets; and the company probably believes the competitive outlook of the RTX 3080 10 GB is secure. This explains the need for a SKU with performance halfway between that of the RTX 3070 and RTX 3080. As for pricing, with the RTX 3070 positioned at $500 and the RTX 3080 at $700, the new SKU could be priced somewhere in between. AMD's RDNA2-based Radeon RX 6000 series GPUs are expected to feature DirectX 12 Ultimate logo compliance, meaning that there is a level playing ground between AMD and NVIDIA in the performance segment.
Source:
kopite7kimi (Twitter)
NVIDIA recently "cancelled" a future 16 GB variant of the RTX 3070, and 20 GB variant of the RTX 3080, which is possibly the company calibrating its response to the Radeon RX 6000 series. We theorize that doubling in memory amounts may not have hit the desired cost-performance targets; and the company probably believes the competitive outlook of the RTX 3080 10 GB is secure. This explains the need for a SKU with performance halfway between that of the RTX 3070 and RTX 3080. As for pricing, with the RTX 3070 positioned at $500 and the RTX 3080 at $700, the new SKU could be priced somewhere in between. AMD's RDNA2-based Radeon RX 6000 series GPUs are expected to feature DirectX 12 Ultimate logo compliance, meaning that there is a level playing ground between AMD and NVIDIA in the performance segment.

86 Comments on NVIDIA Readies New GeForce RTX 30-series SKU Positioned Between RTX 3070 and RTX 3080
All I'm saying is, you have to differentiate between features that your game requires, and features that you want. If you want to "future-proof" yourself with all the shine available, then of course you buy a high-tier GPU with all the current feature sets. There is nothing wrong with that - in fact, that's what I'm planning with my next build.
On the other hand, your game still needs certain processing power and an amount of RAM/VRAM even on the lowest settings. If you aim for just playing a game on basically any setting, then in this case, a graphics card with RT is not going to be any more future proof than one without. But you still need processing power and VRAM whether you use RT or not.
I hope I was clear enough this time.
1. rx5700 accourding to you was outdated upon release, your reasoning for this is the lack of RT support.
2. The RTX2000 is the first itteration of RT and thus is very inefficient at it, poor performance and mediocre looks, RT is not ready for prime-time so buying an RTX2000 series for RT is just silly.
3. RTX3000 is out now, much better at it and it can finally sorta properly play that handful of RT titles currently out, I mean sure if you liked mediocre framerate on a 700+ dollar card....enjoy I guess but I would say that is a poor investment for that experience.
Basically the RX5700 is a perfect purchase for the time period, when RT is nothing more then a gimmick, a techdemo, something to experiment with
By the time it actually becomes available properly the RTX2000 series wont be able to deal with that so you will have to upgrade anyway, and by that time people with an RX5700 will also upgrade.
You have to understand, for games to actually REALLY go for RT, the world has to be able to make use of it, OR Nvidia would have to pump a TON of money in the developer to compensate them for the low amount of sales due to the limited customers that can actually buy/play the game.
It wont be until everyone, low, med and high end can do RT, including consoles, that games can be build on RT tech and again, by that time RTX2000, probably 3000 as well will be just as obsolete as the rx5700.
And probably max out CP2077 on an RTX2080Ti just means they are holding back on the RT in favor of a larger playerbase making use of it.
That is kinda my problem with the whole Vram debate, you see sure, maybe games dont need more then 8gb of Vram today for 4k gaming....but lets keep in mind that the only reason they dont need more is because they dont have more so developers cant make use of it so games are held back.
If all new cards had like 12gb and up, then devs could use higher quality/resolution assets because the cards out there in the hands of players can support it, basically if you are sticking the cards with a relatively low amount, then we are just held back in that regard for the near future, which is not a good thing imo which is why im disappointed in that.
But, nVIdia has played it dirty as usually does. Previous flagship, the 2080Ti has 11GB VRAM. Why on earth the next flagship and replacement has 1GB less? Because 3080 10GB is not what nVidia anounced.
The 3090 24GB is the flagship and replacement of 2080Ti 11GB.
nVidia could have gone with TSMC 7nm and build a real monster GPU. Create a 2080s 8GB replacement with at least 12GB (1 side of PCB fully filled) if not 24GB, call it 3080, with a performance +30% of a 2080Ti 11GB and give it an 800$ price tag. 7nm could have give nVidia headroom to uplift performance another 30% from 3080 12/24GB and create the 3080Ti with 20/24GB, give it a price tag around 1000~1100$ and call it a day... RDNA2 could not reach that level of performance hence the crown and the 1000++ price is justified for the fastest GPU in the world with +60% perf above the 2080Ti. I bet those cards would sell like hot cakes.
Instead nVidia...
Deliberately chopped VRAM off the 3080 and make it 10. Call that a flagship GPU with 1GB less than previous...
Has created another GPU with insingificant perf uplift, call it 3090 instead of 3080Ti, to present as a whole other tier when in reality is not, gave it 24GB and disquise it under the "Titan level performance" just to give it the 1500$ price, and mislead users to a fake 1000$ Titan discount. When in reality is a 2080Ti replacement with a +300~400$ price hike.
Because was also planning to promote a 3080s/Ti 20GB with +5% and a 1000+$ price tag.
But we all know the outcome...
If the 3090 is a Titan call it a Titan, give it Titan performance and Titan drivers and be done with it.
But no...
The master of deceptive marketing strikes again, and making plans to maximize profit margins without giving real performance uplifts beyond 3080 10GB, but taking a few hundread $$$ for every +5% of perf uplift.
The dumbest (rather cheap) decision of all was the choise of Samsungs 8nm(10++++) which lead them to terrible(small) perf/watt gains over Turing and that lead them to push TBP to unpresedent levels of power draw (for nVIdia) and... guess what... lead them to over sized and over engineered and expensive coolers for both FE and AIBs. Thus with the combination of GDDR6X diminishes profit for the 700~800$ price range. Thats why they wanted the 3080s/Ti 20GB and the fake 3080Titan/3090 24GB.
That was a real good plan and business.
Turing GPUs are very capable of handling RT and Ultra details at 1440p and below, not sure why you declare it dead when 4K gaming is still a long ways off.
Ultra details which require massive VRAM are rarely worth it visually, rasterization has long past its diminishing return. That why everyone is flocking to RT now, a less costly RT effect like Reflection and Transparent Reflection can make bigger visual impact than Ultra detail.I agree, so why people are bothered with future proofing 3070/3080 when they are just gonna upgrading to RTX 4000 anyways :D. Adding useless VRAM would just destroy 3070/3080 price to performance advantage.So you prefer another 1200usd 3080 Ti 11GB ?
Not sure if it's a better sell better than the 3080 at 700usd, probably a bit better than 3090.
People whine about Turing high price points too, but what are high end PC gamers gonna do, buy an AMD GPU ? :roll:. Let just hope Big Navi bring big performance back to the table so high end PC gamers will have more choices this time.
I truly hope they can make a top tier version of the 5600 XT .... but every new generation it's been "AMD is gonna". They were the Chicago Cubs of the technology world.... hadn't won it in all in 100 years ... but this was gonna be there year. After 100 years, you couldn't help but root for the cubbies and Im sitting here rooting for AMD ... I would love to see them do it.. But I just can't get invested in the same ole hype we see every new generation only to be left with disappointment ... yet again .... I don't know how peeps can go thru this every generation .... shouting to the rooftops for 2 months before the release how "this is gonna change everything" and than ya gotta show ya face when it's another 290x, Fury, vega or Radeon VIIl.
Its going to begin and end with the GPU ...every time 2 VRAM versions of a card have come out it's been shown that uness you work really hartd to create a special situation, VRAM only makes a significant difference in fps when the frame rates are so low as to make the game unplayable.
www.pugetsystems.com/labs/articles/Video-Card-Performance-2GB-vs-4GB-Memory-154/
alienbabeltech.com/main/gtx-770-4gb-vs-2gb-tested/3/
www.guru3d.com/articles_pages/gigabyte_geforce_gtx_960_g1_gaming_4gb_review,12.html
www.extremetech.com/gaming/213069-is-4gb-of-vram-enough-amds-fury-x-faces-off-with-nvidias-gtx-980-ti-titan-x
Yes ya can find a game that appears to show a difference... a poor console port ... but most games are unaffected ... you can also find games where the performance difference between same cards w/ different VRAM goes DOWN when you increase resolution ... which makes no sense.
But I will play the “broken record mode” and try to break it down for others to understand it better. You understood it just fine, but trying to confuse public. Couldn’t care less why.
nVidia should have gone with 7nm to have better perf/watt ratio so could increase performance even higher on the same/less TBP.
And do...
3080 (at least)12GB, 250W, 30+% perf over 2080Ti 11GB, 700~800$
3080Ti 20/24GB, 300+W, 30+% over 3080 10GB, 1100 or even 1200$.
Titan, 48GB, 2500~3000$, with real Titan performance and drivers.
Instead, they did...
3080 10GB, 320W, +20~30% perf over 2080Ti 11GB, 700~800$
3090 24GB, 350+W, +10% perf over 3080 10GB, 1500+$
Also wanted...
3080Ti 20GB, 320+W, +5% perf over 3080 10GB, 1000+$
So they want/ed after 3080 10GB to charge +300~500$ for every +5% perf uplift.
So to sum up.
Gone cheap with Samsung, forced to over engineer coolers, restrict 3080 to 10GB because they can’t make good enough profit margins, and so push users to 1000+~1500$ GPUs with insignificant (+5%, +5%) perf uplift.
The canceled “3080Ti 20GB” and 3090 24GB.
Faked the 3090 24GB as the next Titan with a fake 1000$ discount, with no Titan performance or drivers. That is no Titan...
As I said, it was a good business plan to maximize profit margins. Sure they will find a way to survive this. But their original plan back-fired pretty much everywhere. From their faces to Jensen’s arse...
They could have won the “fight” single handed and be the clear King of performance and features, still with good enough margins.
But no... they choose lies, fake titles and user exploitation to the max level.
Anything more outrageous that you are demanding in the process ? world peace ? end of pandemic ?
The current 3080/3090 are not good enough ? don't buy it, problem solved.
FYI TSMC is also charging ridiculous prices for their foundry, nothing wrong with supporting Samsung foundry to bring in competition, sounds very much like people want AMD to be competitive eh :D
Also AMD is not a charity either, they will try to increase their margins once they have the performance crown, Zen 3 would like to say Hi :roll:
My 2 examples of nVidia of what should have done and what has done has a meaning.
They’ve choose to do what they have done, and try to present it as my “should” option. That’s was all about.
At least the last 3-4 years any price uplift from AMD was attended by something. A nice performance uplift, and not fake titles(not all but some of them).