GeForce RTX 5050: can you even play on this?

Recently, NVIDIA officially introduced the GeForce RTX 5050 graphics cards, which took the starting position in the RTX 50 series based on the Blackwell architecture. The developers do not particularly focus on the appearance of basic solutions, so this time they did without preliminary testing and the appearance of detailed reviews on the eve of the start of sales. However, of course, for ordinary users who plan to purchase a model of this level, the real capabilities of the new product are of considerable interest. The reviewers of the TechpowerUp resource have already had the opportunity to explore the potential of the GeForce RTX 5050, so we will use the results to form a first impression of the new product.

NVIDIA GeForce RTX 5050

Before analyzing the direct test results, we offer a short excursion into the technical characteristics of the new product. The GeForce RTX 5050 uses a very compact GB207 graphics processor with Blackwell architecture, the die area of which is only 149 mm² . Despite the fact that the same TSMC N4P (5 nm) process is used to manufacture the GPU, the "workpiece" itself includes a relatively small number of CUDA cores, which allows you to reduce the total die area. In the case of the GeForce RTX 5050, the GB207-300 processor version with 2560 compute units, 80 texture units and 32 rasterizers (ROPs) is used. The number of tensor cores is 80, and 20 RT cores are also provided for ray tracing acceleration.

It is not known for certain what the maximum number of processors GB207 provides. Perhaps some reserve remains "under steam" for further upgraded versions of the video cards. In general, only a mobile modification of the GeForce RTX 5050 Laptop video card is currently offered based on this compact GPU.

In the GB207-300 configuration, the processor also has 24 MB of L2 cache. As for operating frequencies, the processor received a base value of 2317 MHz and an average dynamic acceleration of 2572 MHz. The indicators are relatively low, so, given the already perfectly tuned technical process, you can count on at least an additional +300 MHz if you want to play with GPU overclocking.

If you look at the specifications of the GeForce RTX 3050, an interesting fact is the complete coincidence of the number of main functional blocks with that of the RTX 5050. However, the Ampere graphics architecture was used for the representative of the RTX 30. At the same time, the GeForce RTX 3050 uses a "truncated" version of the rather bulky GA106 chip (276 mm² ), which is manufactured by Samsung according to 8 nm standards and in its "full-fledged" version is also used for the older RTX 3060 models. Of course, here the operating frequencies are much more modest - 1552/1777 MHz, and the GPU has an L2 buffer of only 2 MB.

Returning to the specifications of the GeForce RTX 5050, we note the 128-bit memory bus. In general, it is quite expected for this category of video cards. Probably, the memory capacity of 8 GB is also unlikely to surprise anyone, considering that even the older GeForce RTX 5060 is also equipped with a buffer of identical capacity. Because of this, the latter received a lot of complaints, but in an even lower class of adapters, you can't expect any revelations.

But what came as a bit of an unpleasant surprise was NVIDIA's decision to use GDDR6 chips for the GeForce RTX 5050. There were some hopes that the younger model would also receive GDDR7 chips, which are equipped with all other RTX 50 models on the Blackwell architecture. But apparently, in the case of the GeForce RTX 5050, the developers tried to reduce the cost of manufacturing video cards as much as possible. Hence the compact GPU and more affordable memory chips.

Although, in fairness, it is worth noting that GDDR6 is also different. If in the case of the RTX 3060 chips with an effective 14,000 MHz were used, then the GeForce RTX 5050 is already equipped with chips with a standard 20,000 MHz. With an identical 128-bit bus, the memory bandwidth (MBB) differs very noticeably - 224 GB/s vs. 320 GB/s (+43%). GDDR7 would have increased this difference even more. But, we have what we have. A certain "tragedy" of the situation is added by the fact that the mobile version of the GeForce RTX 5050 Laptop received GDDR7 chips. The developers explain this by the fact that for this modification, improved efficiency is important, which is provided by GDDR7 with a lower supply voltage. For desktop models, this is not so important. Yes, but MBB…

The graphics card uses PCI-E 5.0 x8 to connect to the system. Eight PCI Express lanes are enough, especially when it comes to relatively new platforms, at least with PCI-E 4.0. On systems with PCI-E 3.0, you can expect some performance drops in some cases.

The GeForce RTX 5050 received a TGP (Total Graphics Power) rating of 130 W. Again, we have an indicator identical to that of the GeForce RTX 3050. But the GeForce RTX 4060, which can be used almost as a benchmark for efficiency in its class, had slightly lower consumption – 115 W.

As for performance, NVIDIA suggests comparing the GeForce RTX 5050 with the GeForce RTX 3050 using DLSS 4 and frame generation (FG/MFG). Of course, in this case, the predecessor will have no chance. The demonstrative "beating" can be observed on the presentation slide. Comments are superfluous here, so we will pay more attention to the practical results obtained by our colleagues from TechpowerUp.

In the theoretical part, we note that reference GeForce RTX 5050 adapters from NVIDIA are not provided. All proposed models will be original versions from partners. The recommended price of the video card for the American market is $249. The figure is identical to that for the GeForce RTX 3050 in early 2022. Recall that the GeForce RTX 5060 8 GB, like the RTX 4060, is offered $50 more expensive - from $299. Undoubtedly, when it comes to the actual cost in Ukraine, such indicators require appropriate adjustment.

1080p performance

Apparently, the GeForce RTX 5050 is positioned by the manufacturer as a solution for 1080p gaming. Even at this resolution, there are cases when 8 GB of memory is not enough, but for the most part, they are still rare.

According to the results of tests in 24 games, the GeForce RTX 5050 8 GB in Full HD mode is very close to those of the GeForce RTX 4060 8 GB. The model with GB207 turned out to be on average 3% slower than the conventionally older model of the previous generation. Among the closest competitors that offer similar performance in this mode, it is worth noting the Radeon RX 7600 8 GB and Intel ARC B580 12 GB.

As for absolute figures, in certain projects you will have to abandon the maximum quality settings if you want to get a stable 60 fps. The arithmetic average of 72 fps is somewhat misleading, because the contribution of Counter-Strike 2 and The Witcher 3 here filled the "gaps" of a whole list of games where the average fps is lower than the conditionally desired. But this can usually be fixed with the help of DLSS/FG (MFG).

Returning to the "confrontation" of RTX 5050 vs. RTX 4060, we note that in certain projects (Ghost of Tsushima, F1 24, Star Wars Outlaws) the older model of the previous generation may have an advantage of even 10–15%. However, there are also opposite examples when the RTX 5050 shows better results. But if we average all the indicators, we get +3% in favor of the RTX 4060. In general, we are undoubtedly talking about video cards of the same class.

The confrontation with the almost "cult" GeForce RTX 3060 turned out to be interesting. Here, the new product has a fairly confident advantage, in almost all projects. Which resulted in the final average +12.9% in favor of the RTX 5050. The exception, strangely enough, was F1 24, where the older model on Ampere received a 5% advantage.

Unfortunately, the review does not provide a detailed diagram of the situation "RTX 5050 vs. RTX 3050", but the average difference of 56% in favor of the RTX 5050 looks quite convincing. Let's recall the same number of GPU functional blocks of these models, as well as the identical 128-bit memory bus and GDDR6 standard chips. But, of course, there are a lot of nuances "in the details". Starting from architectural optimizations, ending with GUP/memory frequencies. The update turned out to be quite good, the results are relevant.

As for the comparison of the strengths of the GeForce RTX 5050 and GeForce RTX 5060, the older model has a noticeable advantage. A significantly larger number of processors and other functional blocks, as well as higher memory bandwidth, are reflected - the RTX 5060 8 GB is 23% faster on average. At the same time, it is 20% more expensive ($249 vs. $299).

Quite interesting is the confrontation of the GeForce RTX 5050 8 GB with the Intel ARC B580 12 GB. A certain advantage remains for the video card from Intel. Usually it offers more FPS, although in certain projects the new NVIDIA product has higher performance. The anomaly was the performance of Spider Man 2, where the RTX 5050 was 62.5% ahead of its rival. Obviously, the developers' cooperation with NVIDIA and the corresponding optimizations at the engine level made their mark.

Performance in 1440p and 4K modes

We can't expect any revelations regarding the GeForce RTX 5050 8GB's performance in 1440p/4K gaming. After all, this is a 1080p graphics card. However, overall the results aren't a complete disappointment. At least when it comes to 1440p.

The GeForce RTX 5050 confidently holds its own next to the GeForce RTX 4060 8 GB, even reducing the gap from 3% to 2%. Among potential rivals, only the Intel ARC A770 16 GB improved its position here, managing to slightly overtake NVIDIA's new product.

For owners of screens with a resolution of 2560 × 1440/1600, purchasing a GeForce RTX 5050 8 GB will probably be a forced step due to financial constraints. It is possible, but with certain reservations and a clear understanding of when you will have to give in. The average 50 frames/s and the geometric average of 44 fps are evidence that you will often have to resort to deep settings.

The transition to the "4K" league certainly does not improve the position of video cards with 8 GB of memory. In this case, even the GeForce RTX 3060 12 GB provided higher average performance than the RTX 5050. However, in any case, this is not the mode for video cards of this class. Of course, the conditional 25 fps is as much as 14% better than 22 fps, but when it comes to fundamentally important absolute performance, they are not enough in both the first and second cases.

4K at "maximum settings" for this category of video cards looks as "painful" as possible. Here it remains only to count on the miracle of DLSS and frame generation, or try to get by with minimal graphics quality settings. And even better, use video cards according to their initial positioning. Still, the GeForce RTX 5050 8 GB is a story about Full HD.

Ray tracing

You will have to actively play with ray tracing only with the support of the already mentioned "support" technologies. This is a resource-intensive process, which additionally increases the requirements for the video card's memory capacity.

According to the test results, with tracing enabled, the GeForce RTX 5050 is still slightly behind the RTX 4060 (9%), and is ahead of models with higher memory capacity – the GeForce RTX 3060 12 GB and the Intel ARC A770 16 GB. However, a lot depends on the specific project.

Overclocking

All RTX 50 series graphics cards have a pretty good hidden potential for additional overclocking. Despite using a "proven" GPU process, usually an additional +300 MHz is a typical value even in cases where you are not particularly lucky with a particular instance.

It looks like this will also apply to the GeForce RTX 5050. The tested model from GIGABYTE allowed us to increase the base frequency of the graphics processor by +310 MHz – from 2317 to 2627 MHz.

The memory chips operated at an effective 22,960 MHz (+15%).

In this case, you can expect a practical performance increase of 10–12%. Such a spurt will at least nominally allow it to outperform the RTX 4060 in most cases.

Power consumption

The actual power consumption of the GeForce RTX 5050 confirms the stated figure. The video card consumes up to ~130 W during resource-intensive games.

Without changing the GPU manufacturing process, we shouldn't expect any radical changes here. Perhaps the performance of using GDDR7 memory chips would improve somewhat, but the difference would hardly be fundamental for a desktop model.

Overall, the GeForce RTX 5050 definitely falls into the category of very economical models. Although, in terms of performance per watt of energy consumed, it is not a record holder.

Results

Summarizing the first impressions, we note that another entry-level graphics card has appeared on the market, which may interest buyers with a limited budget who use a 1080p screen and prefer not too demanding competitive projects. The GeForce RTX 5050 8 GB demonstrates performance almost at the level of the GeForce RTX 4060 8 GB, which it will probably replace later, and simply outperforms its predecessor, the GeForce RTX 3050 8 GB, which is three years old and has an identical recommended price ($249). And this is even without taking into account the additional "bonuses" of the Blackwell architecture in the form of an efficient video engine with new NVENC/NVDEC, tensor cores with FP4 and full support for DLSS 4 with Multi Frame Generation.

The success of the GeForce RTX 5050 will depend entirely on the actual retail price tags. They probably won't be attractive at the start of sales, but if we later get a video card of the RTX 4060 level, but $50 cheaper, then this will be quite an option for a starting platform that can already be called gaming. Although there will always be a nagging question, "why not add another $50, aiming immediately at the significantly more powerful GeForce RTX 5060 8 GB or Radeon RX 9600 XT 8 GB?".