A couple of days in the past, we had been studying the newest Nvidia RTX 50 collection GPU rumors, and one thing did not sound fairly proper to us. It wasn’t the knowledge itself – we have no thought whether or not it is true or not – it was the supposed configuration that Nvidia goes with for the “RTX 5080.”
The rumored specs for that mannequin, relative to the leaked specs for the RTX 5090, simply did not stack up based mostly on what we have normally come to count on from an “80-class” GPU. And if Nvidia does certainly lower down the RTX 5080 to such an enormous diploma as these rumors are suggesting, it will likely be the weakest 80-class card relative to the flagship mannequin in Nvidia’s historical past.
What does an Nvidia GPU in a given class usually appear like all through historical past?
And this obtained us considering: what does an Nvidia GPU in a given class usually appear like all through historical past? How does the typical 80-tier mannequin, 70-tier mannequin, 60-tier mannequin normally come configured relative to the flagship mannequin?
As a result of if we’re armed with this info forward of any next-gen GPU launches, we will floor our expectations in actuality and push again if needed towards a product that’s providing shoppers lower than they’ve traditionally gotten.
We will additionally use this knowledge to evaluate any GeForce RTX 50-series rumors to see how supposed configurations would examine to greater than a decade of releases.
10 Years of Nvidia GPU Configurations
The primary a part of this evaluation shall be figuring out which mannequin truly matches into every class, as a result of every era has seen a unique lineup.
We have ended up going with this labeling scheme:
- Excessive Flagship
- Flagship
- x80 Ti Class
- x80 Class
- x70 Class
- x60 Class
The 60, 70, 80, and 80 Ti class fashions are fairly simple, then the flagship and excessive flagship playing cards on the very prime. Granted, figuring out the precise flagship mannequin is troublesome as some fashions have come out a lot later than others; some fashions had been so ridiculously priced that nobody would have thought-about them the “smart” flagship in that era, so we ended up designating the flagship card because the GPU most individuals thought-about the most effective of that era.
For instance, within the GeForce RTX 20 collection, the flagship card was usually thought-about to be the RTX 2080 Ti at $1,200, however for those who actually needed to go loopy you possibly can have bought the Titan RTX as an excessive flagship mannequin.
Equally, within the GeForce RTX 30 collection we had the common flagship because the RTX 3090, and the intense flagship got here later – the $2,000 RTX 3090 Ti – for those who simply needed to have the most effective. For the needs of this text, we have additionally excluded the dual-GPU GTX Titan Z that was out there within the 700 collection, so we will analyze single-GPU fashions solely.
Nvidia GPUs: Generations
We have gone again so far as the GeForce 700 collection right here, which was first launched greater than 11 years in the past in 2013, as a result of that is the primary era through which Nvidia produced single-GPU fashions above the 80-class. That is the beginning of what we would describe because the “present” lineup construction. Previous to that, the highest mannequin tended to be the 80-class GPU, with the occasional dual-GPU mannequin thrown in.
Core Depend Configurations
Probably the most necessary points of a GPU that varies considerably between fashions is the shader core rely. Nvidia’s flagship fashions are usually both absolutely unlocked variations of their largest GPU die, or shut to completely unlocked, relying on the era. As we transfer down the product stack, new smaller dies are launched, and current dies are additional lower down to cut back the shader core rely.
Nvidia GPUs: Shader (CUDA) Core Depend
Usually, the 80-tier mannequin has between 60 and 80 % of the cores that the flagship mannequin has. On the excessive finish of the dimensions, the GTX 780 had 2,304 CUDA cores relative to 2,688 on the flagship GTX Titan, so the 80-tier mannequin again in 2013 was lower all the way down to 86% of the scale.
On the decrease finish of the dimensions, we have had the GeForce 900 collection, the place the two,048 CUDA core GTX 980 was simply 67% the scale of the GTX Titan X, and extra lately the RTX 4080 with 9,728 CUDA cores was simply 59% the scale of the RTX 4090 with its 16,384 cores – ouch.
How a lot the 80 collection has been lower down over time has modified, however usually, so far as the core is worried, it is round a 30% discount relative to the flagship.
Then for the 70-tier fashions, normally the GPU core is about half the scale of the flagship, or barely greater than half. That is true for the GTX 770, GTX 970, GTX 1070, RTX 2070, and RTX 3070, however not the RTX 4070, which was lower down greater than standard to only 5,888 CUDA cores, making it simply 36% the scale of the flagship 4090.
As we transfer all the way down to the 60-tier, we discover that GPU cores are usually about one third the scale of the flagship, and once more this holds true for the 5 generations previous to Ada Lovelace; sadly, that RTX 4060 simply would not attain the standard customary.
Once we have a look at the scale of every GPU’s shader core (when it comes to core rely) relative to the flagship mannequin, we will create a pleasant six-generation common for every tier. On common, Nvidia’s 80 tier is 72% the scale of the flagship mannequin, the 70 tier is 54% the scale, and the 60 tier is 36% the scale.
Traditionally, we will additionally see that the 80 Ti tier could be very shut in its core configuration to the flagship. For instance, the RTX 3080 Ti with 10,240 CUDA cores was virtually an RTX 3090 with 10,496 CUDA cores.
Over time, this 80 Ti class has been Nvidia’s option to get virtually the identical stage of efficiency as their finest card, however in a greater worth mannequin.
Reminiscence Bandwidth Configurations
What additionally will get diminished between every tier is reminiscence bandwidth. That is normally achieved by means of narrower reminiscence buses on lower-tier playing cards, in addition to decrease clocked or worse spec reminiscence. This may be mixed right into a GB/s metric that we will use to match every era when it comes to general reminiscence bandwidth out there to the GPU.
Nvidia GPUs: Reminiscence Bandwidth (GB/s)
Reminiscence bandwidth scaling between every tier has usually been fairly just like the scaling seen within the shader core, although there are some exceptions.
For instance, again within the GeForce 700 days, every tier punched above its weight when it comes to bandwidth. The GTX Titan had 285 GB/s of bandwidth, however the GTX 760 retained 192 GB/s of bandwidth, only a 33% discount.
The RTX 2070 and RTX 2080 each shared the identical reminiscence configuration, so whereas the core was in the reduction of on the 70 tier card, it had a comparatively over-spec stage of reminiscence bandwidth.
However usually talking, what was seen within the GeForce 10 collection days is fairly typical. The Titan X and 1080 Ti had round 480 GB/s of bandwidth, lower to 320 GB/s on the GTX 1080, lower to 256 GB/s on the GTX 1070, after which lower to 192 GB/s on the GTX 1060 6GB.
This was largely achieved by means of reducing the reminiscence bus with every tier, mixed with dropping from GDDR5X reminiscence on the upper playing cards to GDDR5 on the decrease playing cards.
When taking a look at percentages relative to the flagship mannequin, we will once more create a median for what every tier usually seems like when it comes to reminiscence bandwidth. On common, the 80 tier will get 77% the reminiscence bandwidth of the flagship card, the 70 tier will get 61% the bandwidth, and the 60 tier will get 45% the bandwidth. So reminiscence bandwidth is not in the reduction of as a lot because the core on common, but it surely’s an identical cadence the place the 70 and 60 tier fashions find yourself with a little bit over and a little bit underneath half the reminiscence bandwidth of the flagship card. We even usually see a slight bump to reminiscence bandwidth for the 80 Ti class mannequin.
VRAM Capability Configurations
Then there’s VRAM capability. This is likely one of the areas the place Nvidia has traditionally given consumers of flagship fashions extra, particularly in generations just like the 900 collection, the place the GTX Titan X had 12GB versus 6GB on the GTX 980 Ti, and extra lately with 24GB on the RTX 3090 versus 12GB on the RTX 3080 Ti.
So whereas these 80 Ti fashions are inclined to not be in the reduction of all that a lot in core configuration or reminiscence bandwidth, getting half the VRAM is frequent and that is factored into the value.
Nvidia GPUs: VRAM Capability (GB)
VRAM capacities are a bit far and wide relying on the era, however on common, the 80 tier has given us 55% the VRAM of the flagship mannequin, the 70 tier has halved the VRAM, and the 60 tier has given us 41% the quantity of VRAM.
Some generations are significantly better than others, although. The GTX 10 collection was an enormous win for shoppers with the flagship playing cards getting 12GB, the 80 and 70 class each getting 8GB, and the 60 class getting 6GB.
However within the era prior, we had the GTX Titan X providing 12GB, the 780 Ti providing 6GB, however this was lower down to only 2GB on the GTX 960. On this space, the RTX 40 collection is definitely across the mark of a median GPU era when it comes to relative spacing of VRAM configurations.
Pricing
One other essential issue is worth. Every tier has usually gotten dearer over time. Within the GeForce 700 to 10 collection days, you had been usually taking a look at round $1,000 for the flagship card. As of late, it is as much as $1,600.
The 80 class has gone from round $600 to ~$1,200, which is painful, thanks fully to the 40 collection. The 70 class was a little bit underneath $400 a decade in the past, now it is as much as $500 and even $600. And the value for the 60 class card has elevated as effectively.
Nvidia GPUs: Launch MSRP
With that stated, the $400 worth of the GTX 770 again in 2013 could be extra like $540 in at the moment’s cash, and the 700 GTX 780 Ti could be almost $950 at the moment. The RTX 4060 is definitely barely cheaper than it needs to be after inflation relative to the GTX 760. However on the entire, Nvidia GPU costs have certainly outpaced inflation in every tier.
What hasn’t modified as a lot is relative pricing. On common, the 80 tier card is simply 60% the price of the flagship mannequin, the 70 tier card is 38% the price of the flagship, and the 60 tier card is simply one-quarter of the value.
There are small abnormalities right here and there – the RTX 2080 and RTX 4080 had been each overpriced, the GTX 1070 a little bit cheaper than standard – however usually Nvidia sticks to this mannequin.
The primary driving issue of GPU costs has been that top-tier flagship mannequin getting dearer annually, which has pushed up the remainder of the lineup. A decade in the past, the GTX Titan was $1,000, and the GTX 770 got here in at 40% of that worth, slotting in at $400.
On this present era, the RTX 4070 launched at 38% the value of the flagship mannequin… it is simply that the RTX 4090 was $1,600 and that meant the 4070 arrived at $600.
In quite a lot of cases, what makes this pricing construction work effectively is the inclusion of that higher worth 80 Ti class card. When playing cards just like the GTX Titan, GTX Titan X, and RTX 3090 are ridiculous, the GTX 780 Ti, GTX 1080 Ti, and RTX 3080 Ti provide a greater deal, normally slotting in at simply 60 to 80% of the value of the flagship for less than small reductions in configuration.
The 2 most up-to-date generations that had worse than typical relative pricing – the 20 collection and 40 collection – each lacked a typical 80 Ti class card. The 40 collection had no 4080 Ti, and within the 20 collection, the 2080 Ti was successfully the flagship, leaving the RTX 2080 to creep up in worth and fill the hole the 80 Ti class normally fills relative to the flagship.
What a Typical Nvidia GPU Technology Seems to be Like
Once we common out the configurations of the final six generations, that is what a typical Nvidia GPU era seems like. Relative to the era’s flagship, the 80 collection card delivers on common 72% of the shader cores, 77% of the reminiscence bandwidth, and 55% of the VRAM capability for 60% of the flagship card’s worth.
The 70 collection card delivers on common 54% of the cores, 61% of the bandwidth, and 50% of the VRAM capability for 38% of the value. And the 60 collection has on common 36% of the cores, 45% of the bandwidth, and 41% of the VRAM capability for twenty-four% of the value.
Nvidia GPUs: Typical Technology (2013-2022)
You may clearly see right here how every soar down within the product stack sees a bigger discount to the value in comparison with its discount in configuration. This preserves the worth proposition of lower-tier fashions and means there’s traditionally been a worth premium to get a sooner card. It additionally means we will decide every era based mostly on this common.
The RTX 40 collection was poorly acquired as a result of every card was lower down greater than it ought to have been, and within the case of the 4080, was each lower down extra and priced increased than standard.
The RTX 30 collection was a terrific success outdoors of crypto mining points, with every mannequin providing decrease relative pricing at every of its standard configurations. The RTX 20 collection had a typical set of configurations however increased relative pricing, making it underwhelming.
Then additional again, the legendary GTX 10 collection was bang on common when it comes to core configurations, provided once more at favorable relative pricing for lower-tier fashions. The 900 collection was related, although not fairly as sturdy as the ten collection, and the 700 collection wasn’t as lower down as standard throughout the lineup, however this got here with common or barely worse than common relative pricing.
How Do RTX 5080 and RTX 5090 Leaked Specs Stack Up?
So the place do these rumors of the GeForce RTX 5090 and RTX 5080 slot in? Properly, issues aren’t wanting good based mostly on these rumored specs. Identified leaker kopite is suggesting the 5090 has 21,760 shader cores and 32GB of GDDR7 on a 512-bit bus. Nevertheless, the RTX 5080 has 10,752 shader cores and 16GB of GDDR7 on a 256-bit bus.
This info means that the RTX 5080 has each the core and reminiscence configuration lower in half relative to the flagship card, although we do not know what reminiscence bandwidth to count on as no GDDR7 pace was listed.
Primarily based on our historic knowledge wanting again at six generations of Nvidia GPUs, the 70 tier is usually the category of card that’s half the configuration of the flagship. Once more, on common, we discovered that tier to have 54% of the core rely, 61% of the reminiscence bandwidth, and 50% of the VRAM measurement at 38% of the value.
Primarily based on latest rumored RTX 5080 specs, this seems to us to be extra of an “RTX 5070” sort of product. We’re not even making an attempt to counsel that the rumor is fallacious and that configuration is definitely a 5070, simply that if that configuration is labeled as a 5080, it could be probably the most cut-down 80 tier mannequin in Nvidia’s historical past.
However extra necessary than the title is the value. The benchmark is de facto set with the value: over the past six generations, when Nvidia provides you a card that is half the flagship’s configuration, they’ve offered that for 62% lower than the flagship.
Which means that if the RTX 5090 is priced at $1,600, to be across the mark of earlier generations, the RTX 5080 as configured needs to be a $600 GPU. If the 5090 is $2,000, the 5080 needs to be within the $750 to $800 vary.
This evaluation has centered on relative configurations and relative pricing, which is only one method to take a look at a graphics card era. Worth-to-performance ratios are equally, if no more, necessary, so a sure GPU configuration that is weak relative to the most effective mannequin may very well nonetheless be good general relying on how its price-to-performance ratio stacks up in comparison with earlier era fashions.
However a giant GPU configuration discount will solely succeed if both the flagship mannequin is so highly effective {that a} huge cutback would nonetheless be very highly effective, or if Nvidia is prepared to cut back the value of a given product tier.
If the RTX 5080 is definitely extra like a 5070 in configuration and generational efficiency – however can be priced like a 5070 – it would not actually matter what it is known as. But when it is a 5070 configuration priced and named like a 5080… effectively, now we’re in bother.
Our final purpose right here was to arm you with historic configuration and pricing info, in order we get nearer to Nvidia’s official launch of the RTX 50 collection you will be higher geared up to investigate the brand new fashions. With a clearer understanding of earlier GPU lineups, you’ll decide whether or not the brand new era is cheap or is ripping you off. It is undoubtedly one thing we’ll be holding a detailed eye on.