Page not found
Page not found
We couldn’t find such page: /en/video/quadro-p5000-vs-tesla-m60%23general-info
Popular graphics cards comparisons
GeForce RTX
3060 Ti
vs
GeForce RTX
3060
GeForce RTX
2060 Super
vs
GeForce RTX
3060
GeForce GTX
1060 6 GB
vs
Radeon RX
580
GeForce RTX
3060 Ti
vs
GeForce RTX
3070
GeForce GTX
1660 Super
vs
GeForce RTX
3050 8 GB
GeForce GTX
1660 Super
vs
Radeon RX
580
Popular graphics cards
GeForce RTX
4090
Radeon RX
580
Radeon RX
Vega 7
GeForce GTX
1050 Ti
GeForce GTX
1650
GeForce RTX
3060
Popular CPU comparisons
Ryzen 5
5600X
vs
Core i5
12400F
Ryzen 5
3600
vs
Ryzen 5
5500
Core i5
10400F
vs
Core i3
12100F
Ryzen 5
3600
vs
Core i5
10400F
Ryzen 5
3600
vs
Core i3
12100F
Core i5
12400F
vs
Core i5
13400F
Popular CPUs
EPYC
9654
Ryzen 5
5500U
Core i3
1115G4
Core i5
12400F
Core i5
1135G7
Ryzen 5
3600
Page not found
Page not found
We couldn’t find such page: /en/video/quadro-p5000-vs-tesla-m60%23characteristics
Popular graphics cards comparisons
GeForce RTX
3060 Ti
vs
GeForce RTX
3060
GeForce RTX
2060 Super
vs
GeForce RTX
3060
GeForce GTX
1060 6 GB
vs
Radeon RX
580
GeForce RTX
3060 Ti
vs
GeForce RTX
3070
GeForce GTX
1660 Super
vs
GeForce RTX
3050 8 GB
GeForce GTX
1660 Super
vs
Radeon RX
580
Popular graphics cards
GeForce RTX
4090
Radeon RX
580
Radeon RX
Vega 7
GeForce GTX
1050 Ti
GeForce GTX
1650
GeForce RTX
3060
Popular CPU comparisons
Ryzen 5
5600X
vs
Core i5
12400F
Ryzen 5
3600
vs
Ryzen 5
5500
Core i5
10400F
vs
Core i3
12100F
Ryzen 5
3600
vs
Core i5
10400F
Ryzen 5
3600
vs
Core i3
12100F
Core i5
12400F
vs
Core i5
13400F
Popular CPUs
EPYC
9654
Ryzen 5
5500U
Core i3
1115G4
Core i5
12400F
Core i5
1135G7
Ryzen 5
3600
Compare NVIDIA Quadro P5000 (Laptop) and NVIDIA Tesla M60
Comparative analysis of video cards NVIDIA Quadro P5000 (Laptop) and NVIDIA Tesla M60 according to all known characteristics in the categories: General information, Specifications, Video outputs and ports, Compatibility, dimensions, requirements, API support, Memory, Technology support.
Analysis of video card performance by benchmarks: PassMark — G3D Mark, PassMark — G2D Mark, Geekbench — OpenCL, CompuBench 1.5 Desktop — Face Detection (mPixels/s), CompuBench 1.5 Desktop — Ocean Surface Simulation (Frames/s), CompuBench 1.5 Desktop — T -Rex (Frames/s), CompuBench 1.5 Desktop — Video Composition (Frames/s), CompuBench 1.5 Desktop — Bitcoin Mining (mHash/s), GFXBench 4.0 — Car Chase Offscreen (Frames), GFXBench 4.0 — Manhattan (Frames), GFXBench 4.0 — T-Rex (Frames), GFXBench 4.0 — Car Chase Offscreen (Fps), GFXBench 4.0 — Manhattan (Fps), GFXBench 4.0 — T-Rex (Fps).
NVIDIA Quadro P5000 (Laptop)
versus
NVIDIA Tesla M60
Benefits
Reasons to choose NVIDIA Quadro P5000 (Laptop)
- Newer graphics card, release date difference 1 year(s) 4 month(s) 90 020
- Core frequency 2.3 times(a ) more: 1278 MHz vs 557 MHz
- 34% more core clock in Boost mode: 1582 MHz vs 1178 MHz
- A newer technological process for the production of the video card allows it to be more powerful, but with lower power consumption: 16 nm vs 28 nm
- 3 times lower power consumption: 100 Watt vs 300 Watt
- 20% more memory clock: 6008 MHz vs 5012 MHz
| Release date | 11 January 2017 vs 30 August 2015 |
| Core frequency | 1278 MHz vs 557 MHz |
| Boost core clock | 1582 MHz vs 1178 MHz |
| Process | 16 nm vs 28 nm |
| Power consumption (TDP) | 100 Watt vs 300 Watt |
| Memory frequency | 6008 MHz vs 5012 MHz |
Reasons to choose NVIDIA Tesla M60
- 11.
1x faster texture speed: 2x 151.0 GTexel / s billion / sec vs 193.7 GTexel / s 2x 2048vs 2048 - 56% better floating point performance: 2x 4.833 gflops vs 6.197 gflops
| Texturing speed | 2x 151.0 GTexel / s billion / sec vs 193.7 GTexel/s |
| Number of shaders | 2x 2048 vs 2048 |
| Floating point performance | 2x 4.833 gflops vs 6.197 gflops |
Benchmark comparison
GPU 1: NVIDIA Quadro P5000 (Laptop)
GPU 2: NVIDIA Tesla M60
| Name | NVIDIA Quadro P5000 (Laptop) | NVIDIA Tesla M60 |
|---|---|---|
| PassMark — G3D Mark | 7771 | |
| PassMark — G2D Mark | 563 | |
| Geekbench — OpenCL | 24595 | |
CompuBench 1. 5 Desktop — Face Detection (mPixels/s) |
114.389 | |
| CompuBench 1.5 Desktop — Ocean Surface Simulation (Frames/s) | 1451.124 | |
| CompuBench 1.5 Desktop — T-Rex (Frames/s) | 8.266 | |
| CompuBench 1.5 Desktop — Video Composition (Frames/s) | 70.038 | |
| CompuBench 1.5 Desktop — Bitcoin Mining (mHash/s) | 314.404 | |
| GFXBench 4.0 — Car Chase Offscreen (Frames) | 8663 | |
GFXBench 4. 0 — Manhattan (Frames) |
886 | |
| GFXBench 4.0 — T-Rex (Frames) | 1791 | |
| GFXBench 4.0 — Car Chase Offscreen (Fps) | 8663 | |
| GFXBench 4.0 — Manhattan (Fps) | 886 | |
| GFXBench 4.0 — T-Rex (Fps) | 1791 |
Performance comparison
| NVIDIA Quadro P5000 (Laptop) | NVIDIA Tesla M60 | |
|---|---|---|
| Architecture | Pascal | Maxwell 2. 0 |
| Codename | GP104 | GM204 |
| Production date | 11 January 2017 | 30 August 2015 |
| Price at first issue date | $1,885 | |
| Ranking | not rated | 427 |
| Price now | $1,699.99 | |
| Type | Mobile workstation | Workstation |
| Price/performance ratio (0-100) | 8.05 | |
| Boost core clock | 1582MHz | 1178 MHz |
| Core frequency | 1278 MHz | 557 MHz |
| Floating point performance | 6. 197 gflops |
2x 4.833 gflops |
| Process | 16nm | 28nm |
| Number of shaders | 2048 | 2x 2048 |
| Texturing speed | 193.7 GTexel/s | 2x 151.0 GTexel / s billion / sec |
| Power consumption (TDP) | 100 Watt | 300 Watt |
| Number of transistors | 7,200 million | 5,200 million |
| Video connectors | No outputs | No outputs |
| DisplayPort | 1. 4 |
|
| Interface | MXM-B (3.0) | PCIe 3.0 x16 |
| Notebook size | large | |
| Length | 267mm | |
| Additional power connectors | 1x 8-pin | |
| DirectX | 12 | 12.0 (12_1) |
| OpenGL | 4.5 | 4.6 |
| Shader Model | 5.1 | |
| Maximum memory size | 16GB | 2x 8GB |
| Memory bandwidth | 192GB/s | 2x 160. 4 GB/s |
| Memory bus width | 256 Bit | 2x 256 Bit |
| Memory frequency | 6008MHz | 5012MHz |
| Memory type | GDDR5 | GDDR5 |
| Shared memory | 0 | |
| 3D Stereo | ||
| 3D Vision Pro | ||
| Mosaic | ||
| nView | ||
| nView Display Management | ||
| Optimus |
choosing the best of nine new GPUs / Habr
In the spring of 2021, NVIDIA introduced a new line of RTX Ax000 and Ax0 graphics cards based on the Ampere architecture, with third-generation tensor cores.
At that time, it was already possible to rent dedicated and cloud servers with Tesla M60, T4, V100 GPUs and even top-end NVIDIA A100s from Selectel.
Since we are trying to provide customers with only up-to-date hardware with modern technologies, we decided that it was time to update the line of video cards. Offering all the video cards announced by NVIDIA is irrational both for us and for our customers. Under the cut, I’ll tell you how we chose the best of the best and share the results of our benchmark on a test build.
The approach by which we at Selectel choose hardware — video cards, processors and other components — is quite simple. We assume that the client wants to solve their business problems efficiently and at minimal cost. Accordingly, we start from the following formula:
According to it, we chose leaders among new video cards.
Which video cards were considered
We compared nine GPUs: RTX video cards from A2000 to A6000, A10, A16, A30, A40 and A100 PCIe.
A2000 was only released this summer, but that didn’t stop us from reviewing the chip’s specifications and testing the sample.
Here we have the RTX Ax000 “family” — from the older A5000 to the younger A2000.
All test participants are server video cards, desktop GeForce RTX 3080 and 3090 are not on the list. These cards (and to be precise, the installation of NVIDIA drivers) are prohibited for use in servers in data centers. The manufacturer strictly monitors compliance with the restrictions: sanctions for violation are applied not only to the provider, but also to the client who rents a server with desktop hardware or installs NVIDIA software on it.
To evaluate video cards, we started from several characteristics that are important for solving problems that customers often have. That is, we looked at what, in general, these GPUs are taken for. The purpose of the cores is presented in a simplified form, each type affects the performance of the video card.
Among them:
- The number of CUDA cores (for those who do not know, this is a symbol for scalar computing units in NVIDIA video chips). The more cores, the better the card can handle graphics and computing in general.
- The number of tensor cores that dynamically optimize calculations and are great at handling workloads typical of working with AI, matrix multiplication for training neural networks and data analysis.
- The number of RT (Ray Tracing) cores that provide high rendering accuracy.
By the way, NVIDIA does not always indicate the exact number of CUDA, RT and tensor cores. For comparison, we used data from third-party sources.
- Memory size.
- Memory bandwidth. These two points logically affect the performance of the video card.
- VDI virtual GPU support. This point is important because VDI is often used by our customers.
- Energy consumption. This is rather a fad for us: for a data center, this indicator is important when choosing a case, power for the server and rack.
Here’s what happened in numbers:
| GPU | RTX A2000 | RTX A4000 | RTX A5000 | RTX A6000 | A10 | A16 | A30 | A40 | A100 PCIe |
| CUDA core | 3328 | 6144 | 8192 | 10752 | 9216* | 1280×4 | 3804* | 10752 | 6912* |
| Tensor cores | 104 | 192 | 256 | 336 | 288* | 40×4 | 224* | 336 | 432* |
| RT core | 26 | 48 | 64 | 84 | 72 | 10×4 | — | 84 | — |
| Memory size (GB) | 6 GDDR6 | 16 GDDR6 | 24 GDDR6 | 48 GDDR6 | 24 GDDR6 | 16×4 GDDR6 | 24 HBM2 | 48 GDDR6 ECC | 40 HBM2 |
| Memory bandwidth (Gb/s) | 288 | 448 | 768 | 768 | 600 | 200×4 | 933 | 696 | 1555 |
| vGPU VDI support | — | — | + | + | + | + | + | + | + |
| Maximum power consumption (W) | 70 | 140 | 230 | 300 | 150 | 250 | 165 | 300 | 250 |
Data not provided by NVIDIA, taken from open third-party sources (pny.
eu, techpowerup.com).
What can you conclude from this plate
For the RTX Ax000 line, the performance grows almost linearly with the model index.
A16 is four graphics cards in one. NVIDIA is positioning the device as a dedicated VDI solution.
A30, at first glance, less productive than A10, but the HBM2 memory type has a higher bandwidth. NVIDIA is positioning the A30 as an AI solution. For both devices, the company does not publish data on the number of tensor and other cores (characteristics obtained from third-party sources).
Compared to other video cards in the table, the top PCIe A100 solution has the maximum memory bandwidth and the maximum number of tensor cores, which is expected. Obviously, the main purpose of this GPU is to work with artificial intelligence and complex calculations. This is the highest performing graphics card in NVIDIA’s lineup to date, especially the 80GB version in the SXM form factor. But the latter is soldered on the board, and for reasons of unification, we considered only the option in the PCIe form factor.
NVIDIA A4000.
How much is
Following the already voiced formula for the choice of components, consider the prices. It is difficult to write about them in 2021, which was remembered for the chip crisis and constant supply disruptions.
There will be no exact figures for two reasons. First, it is a trade secret. Secondly, and most importantly, since the cards were announced in the spring, prices have changed (and I’m sure they will continue to change in the future).
Let’s use the following approach: let’s take the GPU A5000 as a standard — its price in the comparative table will be equal to 1 «parrot». I will present the prices for the rest of the cards through the ratio to the price of the A5000. A10 and A16 are in a close price range, so they «cost» the same.
| GPU | A2000 | A4000 | A5000 | A6000 | A10 | A16 | A30 | A40 | A100 |
| Price | 0. 2 |
0.5 | 1 | 2 | 1 | 1 | 1.6 | 1.7 | 3.3 |
At this stage, the ratio of prices and declared characteristics is expected. The first candidate to be added to the line of Selectel video cards, for the role of the junior model, is the A2000. Also of interest is the parity between the A5000, A10 and A16.
The underside of our GPUs.
Let’s move on to testing the performance of the applicants.
Video card testing
Conducting equipment tests is a common practice for Selectel. We use a large amount of hardware in various company products, so we test it both for compatibility with each other and software, and for performance.
For this we have our own «laboratory» — Selectel Lab. We even provide some equipment to customers for free testing in their projects. From fresh examples: we are giving a real monster DGX A100 with 8 video cards of the same name for testing.
You can read more about its benchmark at the link.
To test the new video cards, we have assembled test servers with two powerful processors from Intel and enough RAM.
Here is one of the servers to test.
The specifications are as follows:
- 2 × Intel® Xeon® Gold 6240: 18 cores @ 2.6 GHz
- 192-384 GB DDR4;
- 240-480 GB SSD SATA;
- 1 × selected GPU
The benchmarks we chose are:
GeekBench 5 is a general benchmark that simulates task execution and determines GPU performance.
AI-benchmark is a performance test that measures the speed of learning and applying various neural networks to recognition and classification tasks.
V-Ray Benchmark is a test to check rendering speed.
ffmpeg NVENC is a video transcoding performance test.
Test results are presented in the table. We identified leaders for each item.
| Model | RTX A2000 | RTX A4000 | RTX A5000 | A10 | A30 | A40 | A100 | |
| GeekBench 5 | OpenCL Compute Score | 81 638 | 137 850 | 182 930 | 167 215 | 122 106 | N/A | 170 137 |
| CUDA Compute Score | 87 283 | 144 283 | 197 025 | 172 765 | 134 492 | 221 139 | 213 899 | |
| AI benchmark | Inference Score | 8 611 | 13 707 | 18 947 | 15 860 | 18 016 | 18 489 | 25 177 |
| Training Score | 9 127 | 14 123 | 19 183 | 16 279 | 19385 | 19 265 | 23775 | |
| AI-Score | 17 738 | 27 830 | 38 130 | 32 139 | 37 401 | 37 754 | 48 952 | |
| V-Ray | V-Ray Benchmark, vpaths | 721 | 1 317 | 1742 | 1 193 | 897 | 1738 | 1539 |
| ffmpeg NVENC benchmark | fps | 172 | 173 | 175 | N/A | N/A | 157 | N/A |
| Time, s | 110. 98 |
110.38 | 108.81 | N/A | N/A | 121.85 | N/A |
At the time of writing, we did not have A16 and RTX A6000 video cards on hand, so they were not included in the table. Their benchmark is planned later.
Benchmark leaders
According to the test results, the A5000 wins in terms of price-quality ratio. The best result in OpenCL Compute Score is slightly inferior to the more expensive A40 and A100 in CUDA Compute Score and is suitable for working with graphics. Second place in AI-benchmark after A100. Leader in the V-Ray rendering speed test, leader in the transcoding test. Supports VDI. Unconditionally our option, if compared with the price table.
A2000 is five times cheaper than the A5000, while still showing acceptable benchmark results for the base model. Does not support VDI, but is suitable for graphics and AI tasks.
The A4000 is mid-range between the A2000 and A5000 in terms of performance, does not support VDI, but otherwise withstands criticism in terms of price and benchmark results.