The November 22, 2024 leak was substantially accurate. It correctly identified the core counts, memory capacities and bus widths of the RTX 5090, RTX 5080 and RTX 5070 Ti, and got the RTX 5070’s memory configuration right. The notable numerical error was the RTX 5070’s CUDA-core count: NVIDIA launched it with 6,144 cores, not the leaked 6,400.
NVIDIA officially announced the Blackwell desktop family on January 6, 2025. The RTX 5090 and RTX 5080 launched on January 30, the RTX 5070 Ti was announced for February, and NVIDIA later set the RTX 5070’s retail launch for March 5, 2025. This is therefore a fact-check of the leak, not current launch news.
What the November 2024 leak claimed
The information originated with Taiwanese benchmarking publication Benchlife, according to HotHardware’s contemporary report. It listed four initial desktop GPUs, their Blackwell die designations, CUDA-core counts, GDDR7 configurations, memory interfaces, provisional power figures and an expected launch window. The report also mentioned an RTX 5090D and future RTX 5060-series models, but the four-card group was the main subject.
| Model | Leaked die | Leaked CUDA cores | Leaked memory | Leaked power estimate |
|---|---|---|---|---|
| RTX 5090 | GB202-300 | 21,760 | 32GB GDDR7 | 600W |
| RTX 5080 | GB203-400 | 10,752 | 16GB GDDR7 | 400W |
| RTX 5070 Ti | GB203-300 | 8,960 | 16GB GDDR7 | 300W |
| RTX 5070 | GB205-300 | 6,400 | 12GB GDDR7 | 250W |
These were rumor-era estimates. In particular, the power numbers should not be treated as universal specifications for every Founders Edition or partner board. Total graphics power, total board power and a board partner’s configured power limit are not interchangeable terms.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
- AI Performance: 772 AI TOPS
- OC mode: 2580 MHz Default mode: 2550 MHz(Boost clock)
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- SFF-Ready Enthusiast GeForce Card
- Axial-tech fans feature a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
Leak versus NVIDIA’s final specifications
NVIDIA’s official comparison data confirms that the leak captured the central configuration of three cards and was close on the fourth. The final lineup uses Blackwell architecture, fifth-generation Tensor Cores and fourth-generation ray-tracing cores throughout.
| Model | Leaked CUDA cores | Final CUDA cores | Final memory | Final bus | Official bandwidth | Launch MSRP | Availability |
|---|---|---|---|---|---|---|---|
| RTX 5090 | 21,760 | 21,760 | 32GB GDDR7 | 512-bit | Not stated in the cited announcement figures | $1,999 | January 30, 2025 |
| RTX 5080 | 10,752 | 10,752 | 16GB GDDR7 | 256-bit | 960GB/s | $999 | January 30, 2025 |
| RTX 5070 Ti | 8,960 | 8,960 | 16GB GDDR7 | 256-bit | 896GB/s | $749 | February 2025 announcement |
| RTX 5070 | 6,400 | 6,144 | 12GB GDDR7 | 192-bit | 672GB/s | $549 | March 5, 2025 |
Final CUDA counts, capacities and bus widths are listed in NVIDIA’s comparison tool. NVIDIA’s launch announcement supplies the bandwidth figures for the RTX 5080, 5070 Ti and 5070. Its Blackwell overview says GDDR7 can provide up to 1.8TB/s across the family, but each model has a different result.
Why GDDR7 and the bus width matter
All four cards use GDDR7. Memory capacity determines how much game, rendering or AI data can remain in VRAM without spilling into slower system memory. Bandwidth determines how quickly the GPU can move that data. A wide interface and fast memory are especially useful at high resolutions, with large textures, ray tracing and compute-heavy workloads.
The RTX 5090 is the clear outlier: 32GB paired with a 512-bit interface. The RTX 5080 and RTX 5070 Ti both provide 16GB on a 256-bit bus, while the RTX 5070 has 12GB on a 192-bit bus. Neither memory bandwidth nor CUDA-core count translates directly into a proportional frame-rate gain; architecture, clocks, cache, drivers, game engines and CPU limits also matter.
RTX 5090: the halo Blackwell card
The RTX 5090 combines 21,760 CUDA cores, 32GB of GDDR7 and a 512-bit memory interface. NVIDIA priced it at $1,999 and launched it on January 30, 2025. Its specification is aimed at maximum-quality 4K gaming, high-refresh 4K, heavy ray tracing and path tracing, 3D rendering, local AI and large creative projects.
Rank #2
- [3352 AI TOPS, 5th Gen Tensor Cores, AI Content Creation] Built for AI-assisted photo and video workflows including upscaling, denoise, background removal, masking, and generative AI creation for faster creator productivity.
- [32GB GDDR7 VRAM, Local LLM Inference, Larger Models] Run local LLM inference and on-device AI tools with massive VRAM headroom for larger models, longer context, and heavier multitasking across AI and creator apps.
- [28 Gbps, 512-bit, 21760 CUDA Cores] High-throughput next-gen memory and core resources for demanding creator projects, complex timelines, large assets, and GPU-accelerated ML experimentation and inference pipelines.
- [Quad-Fan Force, Vapor Chamber, Phase-Change Thermal Pad] Designed for sustained performance under heavy loads with quad-fan cooling, a patented vapor chamber, and a phase-change GPU thermal pad to help lower temps and reduce hotspots.
- [DP 2.1b x3, HDMI 2.1b x2, Bundle GPU Holder] Multi-display ready with up to 4 displays and up to 7680 x 4320 max digital resolution, plus an included GPU Holder to help reduce GPU sag and improve long-term build stability.
NVIDIA’s RTX 5090 product page describes performance of up to twice that of the RTX 4090 in selected games and configurations. That is an NVIDIA claim tied to its test conditions and Blackwell features, including DLSS 4 and Multi Frame Generation; it is not a promise of a universal native-rendering doubling.
The card also demands careful system planning. Check the exact board’s PSU recommendation, 12V-2×6 or adapter arrangement, connector seating, case length, thickness, airflow and cooler clearance. A partner RTX 5090 can be materially larger or configured for a different power limit than a Founders Edition.
RTX 5080: high-end 4K with a major step down from 5090
The RTX 5080 has 10,752 CUDA cores, 16GB of GDDR7, a 256-bit bus and 960GB/s of bandwidth. Its $999 launch MSRP and January 30 launch made it the less expensive high-end option, but it is not a slightly trimmed RTX 5090: the official configuration has half the 5090’s CUDA cores and half its VRAM.
That unusually large gap is central to the buying decision. The 5080 is suited to premium 4K gaming and demanding ray-traced workloads, while the 5090 buys substantially more memory and throughput for users who can justify flagship pricing, power and cooling.
NVIDIA promoted the RTX 5080 as up to twice as fast as the RTX 4080 in selected tests. Interpret that claim through the same DLSS 4 qualification: generated frames and upscaling can materially change the result compared with native raster rendering.
Rank #3
- Equipped with GeForce RTX 5090 chipset
- Integrated with 32GB of 512-bit GDDR7 memory interface
- WATERFORCE All-in-One Cooling System
- RGB Lighting
- Protective metal back plate
RTX 5070 Ti: the 16GB enthusiast option
The RTX 5070 Ti uses 8,960 CUDA cores, 16GB of GDDR7, a 256-bit interface and 896GB/s of bandwidth. NVIDIA announced a $749 MSRP and February 2025 availability. Its 16GB capacity gives it more asset headroom than the RTX 5070, making it an attractive choice for high-refresh 1440p and entry-level-to-capable 4K gaming, as well as creative applications that benefit from additional VRAM.
It is not automatically a 4K flagship replacement. Resolution, ray-tracing mode, texture quality, game engine and the selected DLSS mode still determine whether its memory and rendering throughput are sufficient.
Free tools Windows power users keep installed
One-click scans. No signup required.
RTX 5070: the corrected specification and 1440p focus
The leak’s 12GB GDDR7 and 192-bit bus were correct, but its 6,400 CUDA-core figure was not. NVIDIA’s final specification is 6,144 CUDA cores, with 672GB/s of bandwidth. The launch MSRP was $549. Although NVIDIA initially described February availability for the 5070 family, its later retail announcement set the RTX 5070 launch for March 5, 2025.
The 5070 is most naturally a high-refresh 1440p card. It can run 4K in suitable games and settings, particularly with DLSS, but its 12GB capacity leaves less margin for future ultra-quality textures and intensive path tracing than the 16GB cards. That does not make it unusable at 4K; it makes the settings and workload more consequential.
What the leak got right—and what it did not
- Core counts: exact for the RTX 5090, 5080 and 5070 Ti; overstated for the RTX 5070.
- Memory capacities: correct for all four cards.
- Bus widths: correct for all four cards.
- Power figures: useful as leak-era estimates, not confirmed universal launch specifications.
- Pricing: unavailable in the leak; NVIDIA later announced launch MSRPs of $1,999, $999, $749 and $549.
- Performance: cannot be inferred from CUDA counts alone and should not be equated with NVIDIA’s feature-dependent DLSS 4 comparisons.
How DLSS 4 changes performance comparisons
Separate five different results when evaluating these GPUs:
Rank #4
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5080
- Integrated with 16GB GDDR7 256bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
- Native rasterized rendering.
- Native ray-traced rendering.
- DLSS Super Resolution, which reconstructs a higher-resolution image from a lower-resolution render.
- Conventional Frame Generation.
- Multi Frame Generation, which creates multiple frames between traditionally rendered frames.
A generated frame is not equivalent to a traditionally rendered frame. Latency, image quality, CPU limits and game support remain important. NVIDIA’s “twice as fast” statements should therefore be attributed to NVIDIA and read as test-condition-dependent rather than as universal native performance claims. See NVIDIA’s launch coverage and RTX 50-series overview for the feature and test context.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Power, case and compatibility checks
Before buying any model, verify the exact card rather than relying on the leaked wattage table.
- Use the manufacturer’s PSU recommendation for the specific Founders Edition or add-in-board model.
- Confirm whether the card uses a native 12V-2×6 cable or an adapter, and seat the connector fully.
- Follow the cable’s bend-radius guidance near the plug.
- Measure card length, thickness and radiator or front-fan clearance; large cards can block adjacent PCIe slots.
- Check case airflow and the cooling design, since partner cards may have substantially larger heatsinks.
- Account for factory-overclocked models that can use more power than reference designs.
Which card fits which buyer?
RTX 5090
Choose it for maximum 4K quality, large VRAM requirements, professional rendering or local AI workloads. It is poor value for ordinary 1440p gaming or a compact, low-power build.
RTX 5080
Choose it for premium 4K performance without the 5090’s $1,999 launch price. Its 16GB capacity is the principal long-term limitation relative to the flagship.
RTX 5070 Ti
Choose it when 16GB matters and your target is high-refresh 1440p or capable 4K, but the 5080 is outside budget.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsBest Value
- Powered by GeForce RTX 5090
- Integrated with 32GB GDDR7 512bit memory interface
- WINDFORCE HYPERBURST cooling system
- RGB Halo
- Die-casting Metal Backplate
RTX 5070
Choose it as a practical Blackwell upgrade for high-refresh 1440p, especially from older RTX 20-series, RTX 30-series or GTX hardware. Buyers prioritizing uncompromised long-term 4K or large local-AI workloads should look higher in the stack.
Existing RTX 4090 owners
Upgrade only for a specific need such as more VRAM, stronger ray tracing, creator throughput or Blackwell-only features. NVIDIA’s feature-enabled comparisons do not establish a universal native advantage in every game.
Small-form-factor builders
Do not select by GPU name alone. Exact dimensions, connector routing, slot occupancy, heat output and partner-card cooler design can determine whether a model fits.
MSRP is not a promise of retail price
The announced figures are launch MSRPs, not verified current prices. Founders Edition availability, partner premiums, regional taxes, retailer markups and supply can change what buyers actually pay. Use NVIDIA’s comparison tool to verify specifications, then check region-specific listings for the exact board.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Verdict
The leak was informative rather than infallible. It correctly previewed the defining memory and bus configurations and three of the four final CUDA counts. Its clearest miss was the RTX 5070’s 6,400-core figure, which became 6,144; its power numbers remained provisional. The final lineup’s defining strategic fact is the enormous divide between the 32GB, 21,760-core RTX 5090 and the 16GB, 10,752-core RTX 5080, while the RTX 5070 Ti offers the first-wave range’s most obvious 16GB compromise below flagship pricing.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
