100% Original Wholesale Brand new TU104/400 TU104/450 TU104-410 ADA lovelace type
Here’s a detailed comparison table for the TU104-400, TU104-450, and TU104-410 GPU variants, following NVIDIA’s Turing-era naming conventions and specifications:
TU104 GPU Variants Comparison Table
(RTX 2080 Super / RTX 2070 Super / OEM Models)
| Specification | TU104-400Â (e.g., RTX 2080 Super) | TU104-450Â (e.g., RTX 2080 Super OEM) | TU104-410Â (e.g., RTX 2070 Super) | Notes |
|---|---|---|---|---|
| Die Variant | Full TU104 (400-A1) | TU104 with higher clocks (450-A1) | Cut-down TU104 (410-A1) | |
| CUDA Cores | 3,072 | 3,072 | 2,560 | -17% vs. 400 |
| SMs (Streaming Multiprocessors) | 48 | 48 | 40 | -8 SMs vs. 400 |
| Base Clock | 1,650 MHz | 1,650 MHz (OEM-adjusted) | 1,605 MHz | |
| Boost Clock | 1,815 MHz | 1,860 MHz (OEM-boosted) | 1,770 MHz | +2.5% (450 vs. 400) |
| Memory Size | 8GB GDDR6 | 8GB GDDR6 | 8GB GDDR6 | |
| Memory Bus Width | 256-bit | 256-bit | 256-bit | |
| Memory Bandwidth | 448 GB/s (15.5 Gbps) | 448 GB/s (15.5 Gbps) | 448 GB/s (15.5 Gbps) | |
| RT Cores | 48 (2nd Gen) | 48 | 40 | -17% vs. 400 |
| Tensor Cores | 384 (3rd Gen) | 384 | 320 | -17% vs. 400 |
| TDP | 250W | 250W (OEM-adjusted) | 215W | -14% vs. 400 |
| Typical SKU | RTX 2080 Super (Retail) | RTX 2080 Super (OEM) |
1. TU104-400 (Full Die – e.g., RTX 2080 Super)
The TU104-400 is the fully enabled version of NVIDIA’s TU104 Turing GPU, featuring 3,072 CUDA cores across 48 Streaming Multiprocessors (SMs). Designed for high-end gaming and creative workloads, it includes 48 RT cores (2nd-gen) and 384 Tensor cores (3rd-gen), enabling real-time ray tracing and DLSS acceleration. With a 256-bit memory bus and 8GB of GDDR6 (448 GB/s bandwidth), it delivers strong 1440p and 4K performance. The GPU operates at a base clock of 1,650 MHz and a boost clock of 1,815 MHz, with a 250W TDP.
Typical Use Cases:
- Flagship gaming (RTX 2080 Super)
- Entry-level professional rendering (Quadro RTX 5000 with 16GB VRAM)
2. TU104-450 (OEM-Boosted Variant – e.g., RTX 2080 Super OEM)
The TU104-450 is a specialized OEM variant of the TU104-400, retaining the same 3,072 CUDA cores and 48 SMs but with slightly higher clock speeds (e.g., 1,860 MHz boost). This adjustment allows system integrators (Dell, HP, etc.) to offer pre-overclocked configurations without custom PCBs. Memory specs remain identical to the TU104-400 (8GB GDDR6, 448 GB/s), and the TDP stays at 250W.
Key Differentiators:
- Factory-overclocked for OEM systems
- Minor performance uplift (~2-3%) over retail TU104-400
3. TU104-410 (Cut-Down Die – e.g., RTX 2070 Super)
The TU104-410 is a partially disabled TU104 die, featuring 2,560 CUDA cores (40 SMs) and reduced RT/Tensor core counts (40 RT cores, 320 Tensor cores). It powers the RTX 2070 Super, offering a balance between performance and efficiency. While it shares the same 8GB GDDR6 and 256-bit bus as the TU104-400, its clocks are lower (1,605 MHz base, 1,770 MHz boost), and the TDP drops to 215W.
Performance Trade-offs:
- ~15-20% slower than TU104-400 due to fewer cores
- Targets 1440p gaming at lower power consumption
Summary of Differences
| Aspect | TU104-400 | TU104-450 | TU104-410 |
|---|---|---|---|
| CUDA Cores | 3,072 | 3,072 | 2,560 |
| Boost Clock | 1,815 MHz | 1,860 MHz | 1,770 MHz |
| TDP | 250W | 250W | 215W |
| Market | Retail | OEM | Mid-range |
Why These Variants Exist
- TU104-400: Maximizes performance for enthusiasts.
- TU104-450: OEMs need flexibility for pre-built systems.
- TU104-410: Cost-reduced alternative for broader market reach.
Reviews
There are no reviews yet.