45score
#53 of 131
Overall rank
NVIDIAMid-range
TITAN Xp
As an Amazon Associate we may earn from qualifying purchases.
VRAM
12 GB
CUDA
3,840
FP32
12.15 TF
Bandwidth
547.6 GB/s
TDP
250W
Boost
1582 MHz
Strengths at a Glance
How it stacks up to the flagship
Each metric is shown as a percentage of the GeForce RTX 5090 D, the strongest card we track.
FP32 compute12.15 TFLOPS12%
vs RTX 5090 D: 104.8 TFLOPS
Memory bandwidth547.6 GB/s31%
vs RTX 5090 D: 1790 GB/s
VRAM capacity12 GB38%
vs RTX 5090 D: 32 GB
Shading units3,84018%
vs RTX 5090 D: 21,760
Power efficiency16/10028%
vs RTX 5090 D: 58/100
Synthetic Benchmarks
Estimated benchmark results
Each result is shown as a share of the RTX 5090 D's score in the same test.
3DMark Time Spy16,200pts
vs RTX 5090 D: 36,000 pts
3DMark Port Royal (RT)8,280pts
vs RTX 5090 D: 18,000 pts
Blender (samples/min)2,340spm
vs RTX 5090 D: 5,200 spm
Geekbench Compute112,800pts
vs RTX 5090 D: 240,000 pts
Performance breakdown
Gaming45
Ray tracing46
AI / Compute47
Creator / 3D45
Power efficiency16
Real-World Gaming
FPS Across Resolutions
1080p · Ultra preset1440p · Ultra preset4K · Ultra preset
Cyberpunk 2077avg 39 fps
1080p
53
1440p
39
4K
24
Call of Duty: MW IIIavg 54 fps
1080p
74
1440p
55
4K
34
Alan Wake 2avg 32 fps
1080p
43
1440p
32
4K
20
Forza Horizon 5avg 57 fps
1080p
77
1440p
57
4K
36
Baldur's Gate 3avg 46 fps
1080p
63
1440p
47
4K
29
Community Feedback
What Owners Say
Owners love it as a beast for 4K gaming and VR. The main gripe is the loud fan noise under load and the high power draw.
Pros
- Handles huge textures without stuttering
- Runs hot but needs good airflow
- Crushed 4K gaming for its time
- Still fast for older AAA titles
Cons
- No modern gaming features support
- Runs hot under sustained load
- Huge price for its age
Supported technologies
NVENCAV1 Encode
Full specifications
Graphics processor
Core configuration
- CUDA Cores
- 3,840
- TMUs
- 240
- ROPs
- 96
- L2 cache
- 3 MB
Memory
- Size
- 12 GB
- Type
- GDDR5X
- Bus width
- 384-bit
- Bandwidth
- 547.6 GB/s
- Memory clock
- 1426 MHz
Clocks & throughput
- Base clock
- 1405 MHz
- Boost clock
- 1582 MHz
- FP32 (float)
- 12.15 TFLOPS
- FP16 (half)
- 12.15 TFLOPS
- Pixel rate
- 152 GPixel/s
- Texture rate
- 379.7 GTexel/s
Board & power
- TDP
- 250W
- Suggested PSU
- 500W
- Power connectors
- 1x 6-pin + 1x 8-pin
- Bus interface
- PCIe 3.0 x16
- Length
- 267 mm
- Slot width
- 2-slot
Display & outputs
- Max resolution
- 7680×4320
- Outputs
- 1x HDMI 2.0, 3x DisplayPort 1.4a
API and SDK support
- DirectX
- 12 (12_1)
- Shader Model
- 6.8
- OpenGL
- 4.6
- OpenCL
- 3.0
- Vulkan
- 1.3
- CUDA
- 6.1
Verdict
Our verdict on the TITAN Xp
This is NVIDIA's 2017 flagship single-GPU card, packing immense raw compute power but lacking dedicated tensor or ray tracing hardware.
Get it if you need a legacy card for compute workloads that rely on older CUDA versions or specific driver support. Skip it if you want modern gaming performance or ray tracing, as newer cards beat it handily.
Buy it if…
- You are a pro 3D artist who needs massive VRAM for complex scenes.
- You want the best single-GPU performance in a legacy workstation build.
- You are a deep learning researcher on a tight budget for older hardware.
Leaderboard
Its place in the overall top
2.3
372 votes
Rate this GPU
Add your verdict
Keep exploring