NVIDIA GPU 运算能力列表

本文详细列出了 NVIDIA 各系列 GPU 的计算能力,包括 Tesla、Quadro、NVS、GeForce 和 Tegra/Jetson 系列,涵盖工作站、数据中心、专业可视化、桌面及笔记本电脑等应用领域的 GPU 产品,旨在帮助用户了解 NVIDIA GPU 在加速计算任务方面的强大性能。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

最新信息见:https://developer.nvidia.com/cuda-gpus

NVIDIA GPUs power millions of desktops, notebooks, workstations and supercomputers around the world, accelerating computationally-intensive tasks for consumers, professionals, scientists, and researchers.

Find out all about CUDA and GPU Computing by attending our GPU Computing Webinars and joining our free-to-joinCUDA Registered developer Program.

  • Learn about Tesla for technical and scientific computing
  • Learn about Quadro for professional visualization

If you have an older NVIDIA GPU you may find it listed on our legacy CUDA GPUs page
Click the sections below to expand

************************************************************************************************************

1) CUDA-Enabled Tesla Products

 

Tesla Workstation Products

GPUCompute Capability
Tesla K803.7
Tesla K403.5
Tesla K203.5
Tesla C20752.0
Tesla C2050/C20702.0

//////////////////////////////////////////////////////////////////

Tesla Data Center Products

GPUCompute Capability
Tesla P1006.0
Tesla P406.1
Tesla P46.1
Tesla M405.2
Tesla M405.2
Tesla K803.7
Tesla K403.5
Tesla K203.5
Tesla K103.0


************************************************************************************************************

 

2) CUDA-Enabled Quadro Products

 

 

 

Quadro Desktop Products

GPUCompute Capability
Quadro P60006.1
Quadro P50006.1
Quadro M6000 24GB5.2
Quadro M60005.2
Quadro K60003.5
Quadro M50005.2
Quadro K52003.5
Quadro K50003.0
Quadro M40005.2
Quadro K42003.0
Quadro K40003.0
Quadro M20005.2
Quadro K22005.0
Quadro K20003.0
Quadro K2000D3.0
Quadro K12005.0
Quadro K6205.0
Quadro K6003.0
Quadro K4203.0
Quadro 4103.0
Quadro Plex 70002.0

 

//////////////////////////////////////////////////////////////////

Quadro Mobile Products

GPUCompute Capability
Quadro K6000M3.0
Quadro M5500M5.0
Quadro K5200M3.0
Quadro K5100M3.0
Quadro M5000M5.0
Quadro K500M3.0
Quadro K4200M3.0
Quadro K4100M3.0
Quadro M4000M5.0
Quadro K3100M3.0
Quadro M3000M5.0
Quadro K2200M5.0
Quadro K2100M3.0
Quadro M2000M5.0
Quadro K1100M3.0
Quadro M1000M5.0
Quadro K620M5.0
Quadro K610M3.5
Quadro M600M5.0
Quadro K510M3.5
Quadro M500M5.0

************************************************************************************************************

 

3) CUDA-Enabled NVS Products

 

 

Desktop Products

GPUCompute Capability
NVIDIA NVS 8105.0
NVIDIA NVS 5103.0
NVIDIA NVS 3152.1
NVIDIA NVS 3102.1

 

//////////////////////////////////////////////////////////////////

Mobile Products

GPUCompute Capability
NVS 5400M2.1
NVS 5200M2.1
NVS 4200M2.1

************************************************************************************************************

 

4) CUDA-Enabled GeForce Products

 

 

GeForce Desktop Products

GPUCompute Capability
NVIDIA TITAN X6.1
GeForce GTX 10806.1
GeForce GTX 10706.1
GeForce GTX 10606.1
GeForce GTX TITAN X5.2
GeForce GTX TITAN Z3.5
GeForce GTX TITAN Black3.5
GeForce GTX TITAN3.5
GeForce GTX 980 Ti5.2
GeForce GTX 9805.2
GeForce GTX 9705.2
GeForce GTX 9605.2
GeForce GTX 9505.2
GeForce GTX 780 Ti3.5
GeForce GTX 7803.5
GeForce GTX 7703.0
GeForce GTX 7603.0
GeForce GTX 750 Ti5.0
GeForce GTX 7505.0
GeForce GTX 6903.0
GeForce GTX 6803.0
GeForce GTX 6703.0
GeForce GTX 660 Ti3.0
GeForce GTX 6603.0
GeForce GTX 650 Ti BOOST3.0
GeForce GTX 650 Ti3.0
GeForce GTX 6503.0
GeForce GTX 560 Ti2.1
GeForce GTX 550 Ti2.1
GeForce GTX 4602.1
GeForce GTS 4502.1
GeForce GTS 450*2.1
GeForce GTX 5902.0
GeForce GTX 5802.0
GeForce GTX 5702.0
GeForce GTX 4802.0
GeForce GTX 4702.0
GeForce GTX 4652.0
GeForce GT 7403.0
GeForce GT 7303.5
GeForce GT 730 DDR3,128bit2.1
GeForce GT 7203.5
GeForce GT 705*3.5
GeForce GT 640 (GDDR5)3.5
GeForce GT 640 (GDDR3)2.1
GeForce GT 6302.1
GeForce GT 6202.1
GeForce GT 6102.1
GeForce GT 5202.1
GeForce GT 4402.1
GeForce GT 440*2.1
GeForce GT 4302.1
GeForce GT 430*2.1

************************************************************************************************************

 

5) CUDA-Enabled TEGRA /Jetson Products

 

 

GeForce Notebook Products

GPUCompute Capability
GeForce GTX 10806.1
GeForce GTX 10706.1
GeForce GTX 10606.1
GeForce GTX 9805.2
GeForce GTX 980M5.2
GeForce GTX 970M5.2
GeForce GTX 965M5.2
GeForce GTX 960M5.0
GeForce GTX 950M5.0
GeForce 940M5.0
GeForce 930M5.0
GeForce 920M3.5
GeForce 910M5.2
GeForce GTX 880M3.0
GeForce GTX 870M3.0
GeForce GTX 860M3.0/5.0(**)
GeForce GTX 850M5.0
GeForce 840M5.0
GeForce 830M5.0
GeForce 820M2.1
GeForce 800M2.1
GeForce GTX 780M3.0
GeForce GTX 770M3.0
GeForce GTX 765M3.0
GeForce GTX 760M3.0
GeForce GTX 680MX3.0
GeForce GTX 680M3.0
GeForce GTX 675MX3.0
GeForce GTX 675M2.1
GeForce GTX 670MX3.0
GeForce GTX 670M2.1
GeForce GTX 660M3.0
GeForce GT 750M3.0
GeForce GT 650M3.0
GeForce GT 745M3.0
GeForce GT 645M3.0
GeForce GT 740M3.0
GeForce GT 730M3.0
GeForce GT 640M3.0
GeForce GT 640M LE3.0
GeForce GT 735M3.0
GeForce GT 635M2.1
GeForce GT 730M3.0
GeForce GT 630M2.1
GeForce GT 625M2.1
GeForce GT 720M2.1
GeForce GT 620M2.1
GeForce 710M2.1
GeForce 705M2.1
GeForce 610M2.1
GeForce GTX 580M2.1
GeForce GTX 570M2.1
GeForce GTX 560M2.1
GeForce GT 555M2.1
GeForce GT 550M2.1
GeForce GT 540M2.1
GeForce GT 525M2.1
GeForce GT 520MX2.1
GeForce GT 520M2.1
GeForce GTX 485M2.1
GeForce GTX 470M2.1
GeForce GTX 460M2.1
GeForce GT 445M2.1
GeForce GT 435M2.1
GeForce GT 420M2.1
GeForce GT 415M2.1
GeForce GTX 480M2.0
GeForce 710M2.1
GeForce 410M2.1

************************************************************************************************************

 

6) Tegra Mobile & Jetson Products

 

 

Tegra Mobile & Jetson Products

GPUCompute Capability
Jetson TX15.3
Jetson TK13.2
Tegra X15.3
Tegra K13.2

 

 


Notes

(*) OEM-only products

(**) The GeForce GTX860 and GTX870 come in two versions depending on the SKU, please check with your OEM to determine which one is in your system.

  1. 1152 Kepler Cores with Compute Capability 3.0
  2. 640 Maxwell Cores with higher clocks and Compute Capability 5.0 or 5.2
     

Frequently Asked Questions

1) How can I find out which GPU is in my computer?

Answer:

On Windows computers:

  1. Right-click on desktop
  2. If you see "NVIDIA Control Panel" or "NVIDIA Display" in the pop-up window, you have an NVIDIA GPU
  3. Click on "NVIDIA Control Panel" or "NVIDIA Display" in the pop-up window
  4. Look at "Graphics Card Information"
  5. You will see the name of your NVIDIA GPU

On Apple computers:

  1. Click on "Apple Menu"
  2. Click on "About this Mac"
  3. Click on "More Info"
  4. Select "Graphics/Displays" under Contents list

2) Do I have a CUDA-enabled GPU in my computer?

Answer: Check the list above to see if your GPU is on it. If it is, it means your computer has a modern GPU that can take advantage of CUDA-accelerated applications.

3) How do I know if I have the latest drivers?

Answer: Go to www.nvidia.com/drivers

4) How can I obtain a CUDA-enabled GPU or system?

Answer:
For Tesla for HPC and supercomputing applications, go to www.nvidia.com/object/tesla_wtb.html
For GeForce for entertainment, go to www.nvidia.com/object/geforce_family.html
For Quadro for professional visualization, go to www.nvidia.com/object/workstation_wheretobuy.html

5) How can I download the CUDA software development kit?

Answer: Go to CUDA Development Tools.

<think>好的,我需要帮助用户了解如何测试NVIDIA GPU的计算性能。首先,我应该整理用户提到的引用内容,看看有哪些工具和方法已经被提到。引用[1]提到了Cinebench、PassMark PerformanceTest和Geekbench,不过这些主要是CPU测试工具,可能不直接适用于GPU,但用户的问题是关于GPU的,所以需要更相关的工具。引用[2]提到AMD的Radeon GPU Profiler,但用户询问的是NVIDIA GPU,所以可能需要转向NVIDIA自家的工具。引用[3]强调基准测试工具和实际项目测试的重要性,关注训练时间、推理速度等指标。引用[4]列出了多个NVIDIA GPU型号,说明测试可能需要针对不同型号进行调整。 接下来,我应该考虑NVIDIA GPU常用的测试工具。例如,NVIDIA提供了Nsight Systems和Nsight Compute用于性能分析,这些工具可以监控GPU使用率、内存带宽等。此外,基准测试工具如CUDA-Z和GPU-Burn可能有用。还有第三方工具比如TensorFlow的基准测试脚本和PyTorch的Profiler,适用于深度学习场景。 用户需要具体的测试方法,可能包括理论计算能力测试和实际应用测试。理论测试方面,使用CUDA的deviceQuery示例可以获取硬件规格,而实际测试可能涉及运行深度学习模型或HPC应用。同时,基准测试工具如NVIDIA的官方工具和第三方工具应该分别介绍。 需要确保回答结构清晰,分步骤说明,并引用用户提供的参考资料。例如,在讨论基准测试时,引用[3]提到的关注训练时间、推理速度等指标,可以强调这些方面的重要性。同时,引用[4]中的不同GPU型号可能需要不同的测试方法,可以建议用户根据具体型号选择合适的工具和测试场景。 最后,生成相关问题,帮助用户进一步探索。例如,如何选择适合的基准测试工具,或者如何分析测试结果中的性能瓶颈。这些问题需要基于回答内容,引导用户深入了解。 需要检查LaTeX格式是否正确,所有数学表达式用$...$,独立公式用$$...$$,但在这个回答中可能不需要太多数学公式,除非涉及到计算指标。此外,引用标识要正确添加,比如在提到基准测试时引用[^3],在提到不同GPU型号时引用[^4]。 总结回答结构:首先介绍测试工具,然后分理论和应用测试方法,接着分析性能指标,最后注意事项和相关问题。确保每个部分都有对应的引用支持,并正确标注。</think>### NVIDIA GPU计算性能测试方法 #### 一、测试工具与基准评测工具 1. **NVIDIA官方工具** - **Nsight Systems/Compute** 用于性能分析,支持跟踪GPU内核执行、内存使用、指令级优化等。例如,使用Nsight Systems可生成时间轴视图,定位计算瓶颈[^2]。 ```bash nsys profile --stats=true ./your_cuda_app ``` - **CUDA Toolkit内置工具** `nvprof`(旧版)和`ncu`(NVIDIA Compute Sanitizer)提供内核执行时间、寄存器占用等指标。例如: ```bash ncu --metrics sm_efficiency ./your_kernel ``` 2. **第三方基准测试工具** - **Rodinia Benchmark** 包含多种并行计算模式(如K-means、BFS),用于测试GPU的通用计算能力。 - **TensorFlow/PyTorch基准测试** 运行标准模型(如ResNet-50)测试训练/推理吞吐量。例如,PyTorch的`torch.utils.benchmark`模块支持逐层性能分析[^3]。 3. **理论峰值计算** GPU的**理论浮点性能**可通过公式计算: $$ \text{FP32 TFLOPs} = \text{流处理器数} \times \text{核心频率(GHz)} \times 2 \times 10^{-3} $$ 例如,NVIDIA A100的FP32性能为$6912 \text{ cores} \times 1.41 \text{ GHz} \times 2 = 19.5 \text{ TFLOPs}$。 #### 二、测试方法 1. **理论性能测试** - 使用`deviceQuery`(CUDA Samples)获取硬件参数: ```bash ./deviceQuery | grep "CUDA Capability" ``` - 运行**HPL(High Performance Linpack)**测试双精度浮点性能。 2. **实际应用场景测试** - **深度学习训练**:通过`nvprof`监控显存带宽利用率(`dram_throughput`)和SM活跃率(`sm_efficiency`)[^3]。 - **科学计算**:测试CUDA核函数在流体动力学(如LAMMPS)中的加速比。 #### 三、关键性能指标 | 指标 | 工具/方法 | 参考标准 | |--------------|----------------------------|--------------------| | FP32/FP64 | HPL、Nsight Compute | 接近理论峰值 | | 显存带宽 | `bandwidthTest`(CUDA) | 80%以上为优 | | 延迟敏感性 | STREAM Benchmark | 低延迟高吞吐 | #### 四、注意事项 1. **环境配置**:关闭无关进程,固定GPU频率(`nvidia-smi -lgc`),避免动态Boost干扰[^3]。 2. **结果对比**:参考NVIDIA官方白皮书(如A100架构指南)中的基准数据[^4]。 3. **瓶颈分析**:若计算利用率低于70%,需检查内存访问模式或线程调度策略。
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值