The history of consoles is deeply linked to the evolution of graphics cards. During an early stage, which we can place in the eighties and early nineties, everything revolved around 2D. The graphic power of a console was measured in the size of the sprites and the amount of sprites it could reproduce, and also in things like the effects it was able to generate (such as parallax scrolling, rotations and zooming) and the amount of colors it could display on screen.
We can illustrate this perfectly with a very simple comparison. Mega Drive had a very powerful CPU, namely a Motorola 68000 at 7.62 MHz, but graphically it was more limited than Super Nintendo, as it could only reproduce 61 colors simultaneously against the 256 colors of the big N system, and lacked some of its graphic effects, such as the famous Mode 7, which allowed to rotate and scale textures.
Sega was able to overcome that limit in games like Ranger X (128 colors), and employed  “tricks” to make the games look incredible despite being limited to 61 simultaneous colors. Streets of Rage 2 is one of the best examples of this, as SEGA made clever use of the four 16-color palettes offered by the Mega Drive.
Use of the Mega Drive color palette in Streets of Rage 2. Image courtesy of  “strafefox”
Talking about 2D graphics and consoles implies making an obligatory stop at Neo Geo, a console that was, without any doubt, the queen at a technical level in that category. This console was capable of working with a large amount of large sprites, it could display up to 4.096 colors on screen simultaneously and had specialized 2D graphics chips that allowed it to offer an arcade-like experience in games where Saturn and PlayStation fell short.
And speaking of Saturn and PlayStation, with them the transition to 3D graphics took placealthough 2D continued to be very popular due to the limitations that both consoles had when working with fully polygonal titles, in fact I’m sure many of you will remember that many games from that era used pre-rendered graphics with 3D elementsand that in fact some of the most popular franchises of the time opted for that approach, such as Resident Evil and Final Fantasy.
Specialized hardware made a big difference
And it made it very difficult to make direct comparisons with the hardware we had on PC, especially in the 2D era, where the specialized chipsets used by the Neo Geo, LSPC2-A2 and NEO-B1, were truly unique. However, with the advent of PS1 it started to become a little easier to make such comparisons. Sega Saturn was an exception due to its internal design, it adopted a much more complex configuration than Sony’s console and this made it impossible to find “similar” hardware on PC.
Sega Saturn inside. We can see the two Hitachi SH2 processors.
Those of you who read us every day already know that as the years go by, the consoles were adopting a gradual approach to PC hardwarea process that culminated with PS4 and Xbox One, and has been enshrined with Xbox X Series and S Series, and PS5. However, during the long road we have traveled from the launch of PS1 to the arrival of the current generation we have seen the arrival of many consoles with different configurations, and I don’t think anyone has dared to try to shape a relationship of PC and console graphics card equivalencies that we can consider reliable.
The truth is that it is quite an interesting topic, although quite complicated. I had been thinking about taking the plunge and trying to make an article on this subject for a while, and I finally got up the courage to give it shape Keep in mind that some equivalences are very complicatedand that is why they are based on an estimation that starts from the potential demonstrated by the graphics solution of each console and the closest thing on PC. It should also be remembered that the concept of GPU underwent very important changes, and that today it does not have the same meaning as it had in the nineties.
PS1 used a simple but functional fully customized solution
Sony made a very wise decision when designing PS1, as they opted to use a simple and functional architecture focused primarily on 3D graphics, which made it was very easy to program for it. With Saturn it was precisely the opposite, as it was more focused on 2D and had a multi-chip configuration that forced parallelization.
PS1 inside. Image courtesy of iFixit.
PS1’s GPU was capable of complex 3D graphics, and was supported by a geometry transformation engine (GTE) that could handle up to 360,000 polygons per second with flat shading and 90.000 polygons per second with texture mapping, lighting and Gouraud shading. It had 1 MB of graphic memory, could generate advanced alpha blending effects and transparencies.
Considering that PS1 hit the market in 1994 there is no doubt that it was a very advanced system for the time, and that its 3D capabilities were revolutionary. At that time was when the 3D fever started on PC, and I must say that the first graphics cards that hit the market were not really at the level of the graphics solution that Sony’s console had The closest thing would be a Voodoo 1 from 3DFX, although saving the distances as this was far superior to the PS1 graphics architecture.
Nintendo 64 represented a clear evolution compared to PS1
Nintendo’s 64-bit console was weighed down by bad decisions by the big N, including the use of cartridges instead of CDs, which greatly limited the space available for games and forced developers to resort to compression and canned sound to port wonders like Resident Evil 2, and the memory structure is also questionable, but in spite of everything it was a more powerful system than PS1.
Its GPU was developed and customized by Silicon Graphics, was divided into two chips (Reality Signal Processor and Reality Display Processor) and could move up to 500,000 polygons per second with texture mapping. For comparison, PS1 moved 180,000 texture mapped polygons per second. It was very powerful for its time, but was limited by major design flaws, such as the inclusion of only 4 KB of VRAM for textures.
Nintendo 64 inside. Image courtesy of Copetti.org.
All in all, the Nintendo 64 GPU demonstrated its power with games that would have been impossible on PS1such as the excellent version of Quake 1, the fantastic Turok and the incredible Perfect Dark, among others. Even the adaptation of Resident Evil 2 was very good despite the limitations imposed by the cartridge. It does not have a direct equivalent on PC, but the closest thing is again a Voodoo 1 from 3DFX, also saving the distances because this one is still more powerful. This is very clear when we see how well Turok runs on a Voodoo 1 compared to the Nintendo 64 version.
Dreamcast and PS2 raised the bar a lot, and the equivalences started to become a bit easier
In the case of Dreamcast we have a console that may have been the first to use PC hardware itself, as SEGA was considering using a Voodoo 2 solution from 3DFX, but in the end opted for a NEC PowerVR CLX2, which was capable of working with 6 million polygons with lighting, shading and texture mapping.
If you have been paying attention, you will have noticed that their superiority over previous generations was enormousand this was noticeable in the games, which finally began to use mostly fully 3D graphics that were not generated on the fly in a fog that tried to cover up an obvious void.
Dreamcast inside, a simple but functional design. Image courtesy of Copetti.org.
The graphics that Dreamcast was able to show were incredible for the time, especially for the degree of detail, modeling and texturing of objects, characters and other elements, and also for the higher quality of shadows, lighting and effects Its closest equivalent was a 3DFX Voodoo 2.
PS2 was clearly superior to Dreamcast, both in CPU and GPU, in fact we are talking about a generation in which the tasks of transformation and lighting were still performed at CPU level, and this helped to tip the balance in favor of the Sony console Its graphics unit could work with up to 25 million polygons with textures and effectsbut it was inferior to the Dreamcast graphics unit in some aspects, such as edge smoothing and texturing, as it only had 4 MB of graphics memory (Dreamcast had 8 MB). It’s very difficult to find a direct equivalence, but it would be halfway between a Voodoo 2 and a Voodoo 3 2000.
Game Cube and Xbox were the standard bearers of a revolution called GPU T&L
Both consoles were the most powerful consoles of their generation, although the outright winner in this regard was Xboxa console that I was lucky enough to enjoy at the time and which was truly impressive in terms of hardware and raw power. Both were the first to feature GPU T&L (transform and lighting), a job that was previously done at the CPU level (both Dreamcast and PS2 worked that way), and were able to achieve a graphical quality far superior to that of their generational counterparts from SEGA and Sony.
GameCube had a GPU designed by the now defunct ATi and known as Flipperwhich had four pixel shaders, one vertex shader, four texturing units, four raster units and 16 MB of VRAM accessed via a 64-bit bus. It had a power of 8 GFLOPs and accelerated T&L by hardware. The closest thing we could find in the PC world was the Radeon 7200 from ATi.
Xbox inside. Image courtesy of Copetti.org.
For its part, Xbox was clearly ahead thanks to its NV2A GPU, a version of the NVIDIA GeForce 3 which had four pixel shaders, two vertex shaders, eight texturing units, four raster units, and could access the system’s 64 MB of unified memory via a 128-bit bus. It also accelerated T&L by hardware and had a raw power of 20 GFLOPs. Its closest PC equivalent was a GeForce 200 Ti.
It is important to remember that Xbox was so powerful that developers were able to bring games like Half Life 2 and DOOM III to that consolesomething impressive that was totally impossible on the rest of the consoles of that generation. I don’t want to forget Far Cry Instincts, another title that for power reasons only came to Xbox, and that was impressive considering the PC needed to run it.
PS3, Xbox 360 and Wii repeated the bet on custom PC hardware
And this was especially evident with the PS3’s GPU which was, in essence a custom version of the GeForce 7900 GTas it was manufactured at the 65 nm node (90 nm in the PC graphics solution), and had a very similar, though not identical, configuration. The PS3 GPU had 24 pixel shaders, 8 vertex shaders, 24 texturing units, 8 raster units, had a 128-bit bus, and could access 256 MB of GDDR3 memory, which was 20.80 GB/s, a far cry from the 42.24 GB/s achieved by the GeForce 7900 GT. Its power was 230 GIGAFLOPs.
In the case of Xbox 360, it must be acknowledged that said console had one of the most advanced GPUs of its timeso much so that it was the first general-purpose system to use what we might consider as a a  “primitive” architecture of unified shaders, and it did so in 2005, long before AMD released the Radeon HD 2000 series and NVIDIA released the GeForce 8000 series. Both were the first generations to make the leap to this unified shader architecture.
Xbox 360 inside. Image courtesy of Copetti.org.
The Xbox 360 GPU had 240 shaders16 texturing units, could access that console’s 512 MB of GDDR3 unified memory via a 128-bit bus, resulting in a bandwidth of 22.40 GB/s, and had a power of 240 GIGAFLOPs. Its architecture was far superior to that of the PS3, and it had no concrete equivalent in the general consumer market, since it positioned between the Radeon HD 2600 XT and the Radeon HD 2900 Pro.
Wii was undoubtedly one of the least powerful consoles compared to what we could find at the time within its generation. Generally speaking, it was only slightly more powerful than Game Cube, offering a maximum power of 12 GIGAFLOPsfour more than that console. This was noticeable in its games, which could not compete with those of PS3 and Xbox 360, although the console was a success thanks to its revolutionary way of playing Its closest equivalent was a Radeon X1300but this one was much more powerful.
PS4 and Xbox One marked the farewell to specialized hardware and the  “conversion” of consoles into PCs for gaming
Both Sony and Microsoft forgot about specialized hardware with these two consoles. Both mounted an APU, an AMD solution that integrates CPU and GPU into the same package, and used a x86 processor and a GPU based on the Radeon HD 7000 from AMD, albeit with different approaches.
PS4 mounts a GPU based on GCN 2.0 architecture that totals 1,152 shaders, 72 texturing units, 32 raster units and uses a 256-bit bus to access the 8 GB of unified GDDR5 memory that the console has, of which 5 GB was initially free for developers. Its power is 1.84 TFLOPs, and in general its closest equivalence is the Radeon HD 7850.
For its part, Xbox One uses a GPU based on GCN 1.0 architecture, but configured only with 768 shaders, 48 texturing units, 16 raster units and with a 128-bit bus through which it accesses 8 GB of unified DDR3-type memory, slower than that of PS4, which translated into a bandwidth of only 68.22 GB/s compared to 176 GB/s of PS4. To compensate, Microsoft incorporated 32 MB of eSRAM memory that worked similarly to the infinite cache that AMD employs today in the Radeon RX 6000. The Xbox One GPU reached 1.31 TFLOPsand its closest equivalent was the Radeon HD 7770 GHz Editionwith a power of 1.28 TFLOPs.
PS4 inside. Image courtesy of iFixit.
Both consoles had two major intergenerational refreshers, the PS4 Pro, [whichdoubledthenumberofshadersandusedamoreadvancedarchitectureandXbox One Xwhich had a much more advanced and powerful GPU. The former had a configuration of 2,304 shaders and a power of 4.19 TFLOPs, while the latter went up to 2,560 shaders and had a raw power of 6 TFLOPs. Its closest equivalents were the Radeon RX 470 (4.9 TFLOPs) in the case of the PS4 Pro, and the Radeon RX 580 (6.17 TFLOPs) in the case of the Xbox One X.
Again, Wii U was a console that swam against the tide and tried to introduce a new way to play games, but this time Nintendo did not succeed, and this console was a failure. Its GPU was no big deal either, in fact it was less powerful than the Xbox 360’s, since it it barely reached 176 GFLOPs. It was based on a custom AMD design that featured 160 shaders, 16 texturing units, 8 raster units, and could access the system’s 2 GB of unified memory via a 64-bit bus. Its closest equivalent would be a Radeon HD 2600 XT, which had a power of 192 GFLOPs.
And speaking of consoles from the big N, I cannot close this section without talking about Nintendo Switch. The Japanese company’s handheld uses a Tegra X1 SoC featuring an NVIDIA GPU based on the Maxwell architecture and equipped with 256 shaders. There is no direct equivalent, but just to give you an idea, the GeForce GTX 750 uses the first-generation Maxwell architecture and features 512 shaders.
PS5 and Xbox Series X-S repeated the bet of the previous generation
Both have again mounted an AMD APU, which means that they repeat x86 CPU and Radeon GPU, although of new generation, obviously. In the case of PS5 we have a GPU with 2.304 shaders144 texturing units, 64 raster units, and 36 cores for accelerating ray tracing. This GPU communicates with the system’s 16 GB of GDDR6 memory via a 256-bit bus, and is capable of up to 10.29 TFLOPs.
By specs it looks like the PS5 GPU is directly equivalent to a Radeon RX 6700, but it lacks infinite cache and has significant thermal limitations by sharing packaging with the CPU, so it’s really more in the league of the Radeon RX 6600 XTa graphics card that has a power rating of 10.60 TFLOPs.
Xbox X Series inside. Image courtesy of iFixit.
In the case of Xbox Series X, this is the most powerful console available today. Just like PS5, it uses a Radeon GPU, but configured with 3.328 shaders, 208 texturing units, 80 raster units, 52 cores for accelerating ray tracing, and 10 GB of GDDR6 memory accessed via a 320-bit bus. The other 6 GB use a lower bus and have a lower bandwidth. The raw power of this GPU reaches 12.15 TFLOPs in FP32, but it also lacks infinite cache and has the same limitations as the previous one as it shares a package with the CPU, so its closest equivalent in performance is the Radeon RX 6700which has a raw power of 11.29 TFLOPs.
Xbox Series S is a less powerful, and less expensive, version of Xbox Series X, and is very curious, as it maintains the next-generation architecture (RDNA2), but its GPU is much more modest, having a configuration of 1.280 shaders80 texturing units, 32 raster units, 20 cores for accelerating ray tracing, and can access 8 GB of unified GDDR6 over a 128-bit bus. The other 2 GB of GDDR6 works on a much slower bus, and has a lower bandwidth.
In raw power, this GPU achieves the following 4 TFLOPsand has the same limitations as above because it also shares packaging with the CPU and lacks infinite cache. It has as closest equivalent to the Radeon RX 6400whose raw power is 3.56 TFLOPs.
