According to reputable hardware leaker, @Zed__Wang, Nvidia intends to release its inexpensive GeForce RTX 4050 and GeForce RTX 4060 Ti graphics cards for desktops in May – June timeframe, whereas the launch window for GeForce RTX 4060 is not yet decided. The leaker, has a track record of revealing accurate information about Nvidia's forthcoming hardware but as always take the news with the necessary amount of sodium chloride.
Release date：Desktop 4060Ti - end of MayDesktop 4050- JuneDesktop 4060- not decided yetMarch 29, 2023
Nvidia allegedly plans to release its performance-mainstream GeForce RTX 4060 Ti sometime in late May and then launch its inexpensive GeForce RTX 4050 graphics board in June, the leaker claims. Keeping in mind that Computex Taipei — the world's largest PC hardware trade show — is set to be held on May 30 – June 2, it is possible that Nvidia will use the exhibition to launch new its new desktop products (which have a chance to join the ranks of best graphics cards), though we are speculating.
The green company still has not decided when to release its mainstream GeForce RTX 4060, but it is logical to release it after the RTX 4060 Ti variant in a bid to motivate mainstream buyers to get the more expensive Ada Lovelace product.
In fact, @Zed__Waang claims that Nvidia has already supplied its partners among graphics cards suppliers with templates for GeForce RTX 4060 Ti packages, which is an indicator that launch preparations for this product have started. Unfortunately, the package does not reveal any particular information about the upcoming graphics adapter, such as the number of CUDA cores or the amount of memory it carries onboard.
Based on what we know from previous leaks, Nvidia's GeForce RTX 4060 Ti is set to be based on the AD106 graphics processor with 4352 CUDA cores mated with 8GB of GDDR6 memory using a 128-bit interface. By contrast, the GeForce RTX 4060 is rumored to feature a cut-down version of the AD106 GPU (presumably with 3072 CUDA cores) and 8GB of memory on a 128-bit bus.
|Row 0 - Cell 0||GPU||FP32 CUDA Cores||Memory Configuration||TBP||MSRP|
|GeForce RTX 4090 Ti||AD102||18176 (?)||24GB 384-bit 24 GT/s GDDR6X (?)||600W (?)||?|
|GeForce RTX 4090||AD102||16384||24GB 384-bit 21 GT/s GDDR6X||450W||$1,599|
|GeForce RTX 4080||AD103||9728||16GB 256-bit 22.4 GT/s GDDR6X||320W||$1,199|
|GeForce RTX 4070 Ti||AD104||7680||12GB 192-bit 21 GT/s GDDR6X||285W||$799|
|GeForce RTX 4070||AD104||5888 (?)||12GB 192-bit 21 GT/s GDDR6X (?)||250W (?)||?|
|GeForce RTX 4060 Ti||AD106||4352 (?)||8GB 128-bit 18 GT/s GDDR6 (?)||160W (?)||<$500?|
|GeForce RTX 4060||AD106||3072 (?)||8GB 128-bit (?)||?||?|
|GeForce RTX 4050||AD107 (?) AD106 (?)||?||?||?||?|
Meanwhile the GeForce RTX 4050 is largely a mysterious product. Nvidia's Ampere-based GeForce RTX 3050 used both GA106 and GA107 graphics processors and it is reasonable to expect Ada Lovelace-based GeForce RTX 4050 to use both AD106 and AD107 as well as to give the GPU developer and its partners a lot of flexibility with these boards that are meant to be priced reasonable. Then again, any assumptions about specifications of upcoming products that are over two months away are pure speculations.
But the chances are high this GPU might sport the AD107-400-A1 GPU instead, the fully enabled AD107 Ada die.
Do note that the same configuration is used by the GeForce RTX 4060 Laptop GPU, so it makes some sense to use the same die for the desktop variant as well, BUT we can't completely rule out the AD106 die just yet.
The RTX 4060 Laptop card also offers 3072 shaders and uses 8 GB GDDR6 dedicated graphics memory with a clock speed of 16 Gbps (effective) and a 128 Bit memory bus. So the desktop variant should sport similar specs, unless NVIDIA has some other plans for this mainstream desktop SKU.
The other question is if that 4050 will be sub 75 watt or will it require an extra 8 pin. The 4060 is apparently 115 watt, so it is possible the 4050 is sub 75 watt.
I hope Intel sorts out its drivers soon so I can get an A750 without worrying too much about weird frame time jitters.
More likely, I could see Nvidia's RTX 4060 Ti design being released as an RTX 4070 variant in May and their RTX 4050 design being sold as an RTX 4060 in June.
The same leaker has also confirmed the memory specs of the RTX 4050. He claims the card will come out in a 6GB VRAM flavor, so that means a 96-bit memory bus width interface, and most likely sporting 2560 CUDA cores ?
A HUGE downgrade than the RTX 3050 which had at least 128-bit memory bus and 8GB VRAM. But I'm just assuming the 4050 sports a 96-bit memory bus. Maybe Nvidia has other plans for this SKU.
But jeez, can a 96-bit wide memory interface GPU be worth the purchase in 2023 ? Though, the RTX 4050 might still end up being faster than the RTX 3050 despite this narrower/smaller memory, since it is based on the new ADA architecture. But this is just a speculation for now.
But a 96-bit bus will still be a huge limiting factor for most of the modern games, if not all. Main question, is how much Nvidia will charge for this RTX 4050-class GPU. My guess is even more than the RTX 3050 given the latest pricing trend of ADA lineup.
With 8GB quickly becoming the minimum amount of memory that is advisable for 1080p gaming, Nvidia's RTX 4050 will be poorly positioned within today's PC market. This is especially true knowing that Nvidia's RTX 3050 has 8GB of graphics memory.
As far back as Nvidia's RTX 20 series, 8GB of graphics memory was not enough for modern games, and since then memory requirements for modern PC games have only gotten higher. Like the RTX 2060 6GB, we can expect a 6GB RTX 4050 to age poorly, as GPU memory requirements aren't going to get any lower.
If Nvidia releases an RTX 4050 with 6GB of memory, Nvidia will be releasing a compromised product. But I digress.
Now, we have to cross our fingers that AMD's lacklustre sales on RX7xxx' inflated prices combined with far less impressive share of HPC, AI, etc. sales are enough to convince it to make more reasonable pricing decisions on its lower-end GPUs to remain competitive with what Intel may be coming up next. Recent rumors say BMG may actually launch with something more than twice as fast as the A770. Hopefully Intel will have learned its lessons from the A7xx and aim for market share and brand recognition in the GPU space instead of trying to be greedy and get ignored by most of the market.
If there is a 6GB/96b 4050, then there will almost certainly be a 8GB/128b 4050Ti later.
I certainly hope the 6GB/96b version, if that is really Nvidia's plan, won't cost more than $250. I wouldn't buy one for more than $180.
As for a 6GB GPU being "compromised", nobody buys this tier of graphics without expecting significant compromises in modern-day titles and almost certainly even more compromises in the near future. I'm still using a GTX1050 and getting most of the newer stuff to run remotely acceptably means turning everything down as low as it goes.
Exactly this. Nvidia is transitioning from a "gaming" GPU company to an AI/ML general compute company. Gaming as a share of revenue and profit will continue to shrink so they won't be as interested in keeping that market happy as the will be for the ML training market.