OVERALL
ALTERNATIVES
SPECS
VS AVERAGE
REVIEWS
Scored using Graphics cards Methodology v3.0
Graphics cards Methodology v3.0
Mathematical & Unbiased Scoring
Products in this category are scored under Frontum Graphics cards Methodology v3.0: the technical score from the specifications is the base, and independent expert and video reviews (up to ±0.5 points), owner opinions (up to ±0.3) and best-of lists (up to +0.3) move it up or down.
Explore Graphics cards Methodology v3.0

NVIDIA Tesla K40 review: 117 specs compared

Avg. price:~£150VRAM:12 GBMemory bus width:384 bitThermal Design Power (TDP):235 W

NVIDIA Tesla K40 Review: Compare 117 technical specifications and 1 expert and video reviews to see how this product ranks in the Graphics cards category and whether it is worth buying.

NVIDIA Tesla K40
5.8/10

Overall score

Score
5.8/ 10
✕
What it is
Our main score for this model, 1–10. It combines specs, expert reviews, best-of lists and owners.
How we count it
Parts: Technical score 40%, Reviews score 25%, List score 25%, Sentiment score 10%
Marks: Each part gets a mark out of 10. For reviews, lists and owners, a typical model in this category gets about 8 and the best one gets 10.
No data yet: A part with no data shows "?". It counts as the mark of a weaker model in this category (4 in 5 models with data score higher). This way, models with reviews, lists and owners usually come out ahead.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 91% of models
Score components

40%

4.8

Technical score

25%

4.8

Reviews score

25%

?

List scoreNot enough data yet, so we use 7.5, the mark of a weaker model in this category

10%

?

Sentiment scoreNot enough data yet, so we use 7.8, the mark of a weaker model in this category

â–¼Worse than 91% of graphics cards
7.0

Value for money

Value for money
7.0/ 10
✕
What it is
An indicator that combines the graphics card's overall rating with its cost.
When it matters
When you are looking for a graphics card that offers a strong balance of performance, features, and price.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 75% of models
Score components

70%

5.8

Overall score

30%

9.7

Price

â–¼Worse than 75%
4.8

Technical score

Technical score
4.8/ 10
✕
What it is
How good the specs are compared with the other models in this category, 1–10.
How we count it
Areas: We rate the specs in each area against the other models in this category.
Weighting: The areas that matter most to buyers count more. The shares are shown under Score components.
Share: The technical score is 40% of the overall score.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 86% of models
Score components

44%

2.6

Performance

24%

6.2

Memory

12%

7.1

Power

11%

6.4

Features

5%

8.2

Design

4%

5.9

Connectivity

â–¼Worse than 86%
4.8

Reviews score

Reviews score
4.8/ 10
✕
What it is
Average grade from independent test sites, magazines and video reviewers, 1–10.
How we count it
Grades: Each review's own score is converted to 1–10; without one, we grade the review ourselves. Big test sites count more, short reviews half as much as full tests. A video review counts half.
Score: The average grade is 80% of the review mark; the number of reviews is the other 20% (1 review = 5, the typical number in this category = 8, the most-reviewed model = 10).
Enough data: With few reviews, the average grade is pulled toward the typical grade of this category. No reviews yet: 7.6 for now.
Share: Reviews count 25% of the overall score.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 90% of models

4.8

Mark 7.3 is below the typical mark 7.9 of this category, so it cannot lift the score above the technical score

80%

7.9

Average grade

20%

5.0

Popularity

Grades by review (1)
1 video review, counting half as much as a written test. Reviews count 25% of the overall score.
User ratings collected online are not part of this score.
â–¼Worse than 90%
?

List score

List score
7.5/ 10
✕
What it is
How often independent test sites, magazines and video reviewers put this model in their best-of lists for this category.
How we count it
Points: The model in the most lists of this category gets 10; the others get 8 to 10 in proportion. In no list yet: 7.5. Lists count 25% of the overall score.
Enough data: Every list counts, even a single one. With fewer than 5 best-of lists in the category, this part is left out.
Lists with this model (0)
Not in any best-of list yet, so this part counts as 7.5. Models in a list get 8 to 10.
Not in any best-of list yet
  • nvidia-tesla-k40
nvidia-tesla-k40
Where to buy
*As an Amazon Associate, we earn from qualifying purchases.
Best rankings
Available: ranking among products currently available (including other versions of this product).All: ranking among all products in the database.
Loading category ranking...

Technical Specifications of graphics card NVIDIA Tesla K40

Technical score
Technical score
4.8/ 10
✕
What it is
How good the specs are compared with the other models in this category, 1–10.
How we count it
Areas: We rate the specs in each area against the other models in this category.
Weighting: The areas that matter most to buyers count more. The shares are shown under Score components.
Share: The technical score is 40% of the overall score.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 86% of models
Score components:

44%

2.6

Performance

24%

6.2

Memory

12%

7.1

Power

11%

6.4

Features

5%

8.2

Design

4%

5.9

Connectivity

4.8
Reviews score
4.8
List score
?
Value for money
Value for money
7.0/ 10
✕
What it is
An indicator that combines the graphics card's overall rating with its cost.
When it matters
When you are looking for a graphics card that offers a strong balance of performance, features, and price.
Score distribution
â–¼
<55-66-77-88-99-10
â–¼ Worse than 75% of models
Score components:

70%

5.8

Overall score

30%

9.7

Price

7.0
3DMark Time Spy benchmark score
?
3DMark Port Royal score
0 points
Market range: <0 points – >22.4k points
This graphics card scores 0 points in 3DMark Port Royal, which is lower than 83.1% of graphics cards and equal to 16.9% of graphics cards.
PassMark (G3D) result
4,495 points
Market range: <7.7k points – >36k points
This graphics card scores 4,495 points in PassMark G3D, which is lower than 94.2% of graphics cards.
PassMark (DirectCompute) result
1,429 points
Market range: <3.1k points – >26.3k points
This graphics card scores 1,429 points in PassMark DirectCompute, which is lower than 96.1% of graphics cards.
FP32 performance
4.29 TFLOPS
Market range: <3.94 TFLOPS – >56.3 TFLOPS
This graphics card delivers 4.29 TFLOPS FP32 performance, which is lower than that of 89.2% of graphics cards.
Show more
VRAM
12 GB
Market range: <4 GB – >16 GB
This graphics card has 12 GB of VRAM, which is more than 46% of graphics cards and equal to 16% of graphics cards.
Memory type
GDDR5
This graphics card uses GDDR5 memory, which is older than on 85.4% of graphics cards and equal to 13.2% of graphics cards.
Memory bus width
384 bit
Market range: <128 bit – >384 bit
This graphics card uses a 384 bit memory bus, which is wider than that of 89.7% of graphics cards and equal to that of 7% of graphics cards.
Maximum memory bandwidth
288 GB/s
Market range: <168 GB/s – >960 GB/s
This graphics card reaches 288 GB/s memory bandwidth, which is lower than that of 69.9% of graphics cards and equal to that of 6.5% of graphics cards.
Effective memory speed
3,000 MHz
Market range: <7k MHz – >28k MHz
This graphics card reaches an effective memory speed of 3,000 MHz, which is slower than that of 98.1% of graphics cards.
Show more
PCI Express (PCIe) version
3
This graphics card supports PCIe 3, which is older than on 77.6% of graphics cards and equal to 20.1% of graphics cards.
PCIe lanes
x16
This graphics card uses x16 PCIe lanes, which is more than 24.1% of graphics cards and equal to 75.9% of graphics cards.
DirectX version
12
This graphics card supports DirectX 12, which is older than on 81.8% of graphics cards and equal to 17% of graphics cards.
Vulkan version
1.2
This graphics card supports Vulkan 1.2, which is older than on 94.9% of graphics cards and equal to 3.9% of graphics cards.
OpenGL version
4.6
This graphics card supports OpenGL 4.6, which is more advanced than on 2% of graphics cards and equal to 98% of graphics cards.
Show more
Max displays supported
?
Max digital resolution
?
DisplayPort outputs
1
Market range: <1 – >3
This graphics card offers 1 DisplayPort outputs, which is fewer than 88.1% of graphics cards and equal to 8.7% of graphics cards.
DisplayPort version
?
DisplayPort link rates
?
Show more
Thermal Design Power (TDP)
235 W
Market range: <86 W – >360 W
This graphics card has a TDP of 235 W, which is higher than that of 57.3% of graphics cards.
Power consumption while under peak load
235 W
Market range: <106 W – >365 W
This graphics card draws 235 W under peak load, which is higher than 55% of graphics cards and equal to 0.1% of graphics cards.
Recommended PSU wattage
235 W
Market range: <450 W – >850 W
This graphics card recommends a 235 W PSU, which is lower than that of 99.6% of graphics cards.
Board power limit
235 W
Market range: <107 W – >400 W
This graphics card has a board power limit of 235 W, which is higher than that of 52.7% of graphics cards and equal to that of 0.2% of graphics cards.
PCIe power spec
235 W
Show more
Size
267x 111x 38 mm
Length
267 mm
Market range: <200 mm – >337 mm
This graphics card is 267 mm long, which is shorter than 59.3% of graphics cards and equal in length to 3.5% of graphics cards.
Height
110 mm
Market range: <109 mm – >146 mm
This graphics card is 110 mm tall, which is shorter than 88.8% of graphics cards and equal in height to 1.1% of graphics cards.
Slot width
2 slot/s
Market range: <2 slot/s – >3.25 slot/s
This graphics card occupies 2 slot/s, which is slimmer than 52.4% of graphics cards and equal in width to 43.4% of graphics cards.
Weight
826 g
Market range: <541 g – >2k g
This graphics card weighs 826 g, which is lighter than 68.8% of Graphics cards.
Show more

NVIDIA Tesla K40 vs the average graphics card

246810TotalPerformanceMemoryFeaturesConnectivityPowerDesign
  • 192 bit wider memory bus
    This graphics card has a wider memory bus than the average graphics card (384 bit vs 192 bit). The average graphics card has a memory bus width of 192 bit.
    384 bit vs 192 bit
  • 63.8% lower PSU requirement
    This graphics card has a lower PSU requirement than the average graphics card (235 W vs 650 W). The average graphics card has a PSU requirement of 650 W.
    235 W vs 650 W
  • 71.4% faster VRAM clock
    This graphics card has a higher VRAM clock than the average graphics card (3,000 MHz vs 1,750 MHz). The average graphics card runs its VRAM at 1,750 MHz.
    3,000 MHz vs 1,750 MHz
  • Supports ECC memory
    This graphics card supports ECC memory, the average graphics card does not.
  • 56 more TMUs
    This graphics card has more TMUs than the average graphics card (240 vs 184). The average graphics card has 184 TMUs.
    240 vs 184
  • 15 mm lower card height
    This graphics card is shorter than the average graphics card (110 mm vs 125 mm). The average graphics card has a height of 125 mm.
    110 mm vs 125 mm
  • 0.5 slot/s slimmer design
    This graphics card occupies fewer slots than the average graphics card (2 slot/s vs 2.5 slot/s). The average graphics card occupies 2.5 slot/s.
    2 slot/s vs 2.5 slot/s
  • 8 more GPUs supported
    This graphics card supports more GPUs in multi-GPU mode than the average graphics card (8 vs 0). The average graphics card supports 0 GPUs in multi-GPU mode.
    8 vs 0
  • 3.47x cheaper
    NVIDIA Tesla K40 is cheaper than the average graphics card (£150 vs £520).
    £150 vs £520
  • 56 more TMUs
    This graphics card has more TMUs than the average graphics card (240 vs 184). The average graphics card has 184 TMUs.
    240 vs 184
  • More FP64 throughput
    This graphics card has a better FP64 ratio than the average graphics card (1:3 vs 1:64).
    1:3 vs 1:64
  • 2.98x higher FP64 performance
    This graphics card has a higher FP64 performance than the average graphics card (1.4 TFLOPS vs 0.48 TFLOPS). The average graphics card has FP64 performance of 0.48 TFLOPS.
    1.43 TFLOPS vs 0.48 TFLOPS
  • 192 bit wider memory bus
    This graphics card has a wider memory bus than the average graphics card (384 bit vs 192 bit). The average graphics card has a memory bus width of 192 bit.
    384 bit vs 192 bit
  • 71.4% faster VRAM clock
    This graphics card has a higher VRAM clock than the average graphics card (3,000 MHz vs 1,750 MHz). The average graphics card runs its VRAM at 1,750 MHz.
    3,000 MHz vs 1,750 MHz
  • Supports ECC memory
    This graphics card supports ECC memory, the average graphics card does not.
  • 8 more GPUs supported
    This graphics card supports more GPUs in multi-GPU mode than the average graphics card (8 vs 0). The average graphics card supports 0 GPUs in multi-GPU mode.
    8 vs 0
  • More NVENC sessions
    This graphics card has more concurrent NVENC sessions than the average graphics card (Unlimited vs 8). The average graphics card has 8 concurrent NVENC sessions.
    Unlimited vs 8
  • 63.8% lower PSU requirement
    This graphics card has a lower PSU requirement than the average graphics card (235 W vs 650 W). The average graphics card has a PSU requirement of 650 W.
    235 W vs 650 W
  • 15 mm lower card height
    This graphics card is shorter than the average graphics card (110 mm vs 125 mm). The average graphics card has a height of 125 mm.
    110 mm vs 125 mm
  • 0.5 slot/s slimmer design
    This graphics card occupies fewer slots than the average graphics card (2 slot/s vs 2.5 slot/s). The average graphics card occupies 2.5 slot/s.
    2 slot/s vs 2.5 slot/s
  • 65.1% lower boost clock speed
    This graphics card has a lower boost GPU clock than the average graphics card (875 MHz vs 2,505 MHz). The average graphics card has a boost GPU clock of 2,505 MHz.
    875 MHz vs 2,505 MHz
  • 40 fewer ray tracing cores
    This graphics card has fewer ray tracing cores than the average graphics card (0 vs 40). The average graphics card has 40 ray tracing cores.
    0 vs 40
  • 5.6x larger process node
    This graphics card uses a larger manufacturing process than the average graphics card (28 nm vs 5 nm). The average graphics card uses a manufacturing process of 5 nm.
    28 nm vs 5 nm
  • 2 fewer fans
    This graphics card has fewer fans than the average graphics card (1 vs 3).
    1 vs 3
  • 25 fewer compute units
    This graphics card has fewer compute units than the average graphics card (15 vs 40). The average graphics card has 40 compute units.
    15 vs 40
  • 61.2% lower base clock speed
    This graphics card has a lower base GPU clock than the average graphics card (745 MHz vs 1920 MHz). The average graphics card has a base GPU clock of 1920 MHz.
    745 MHz vs 1,920 MHz
  • Less advanced frame generation
    This graphics card has less advanced frame generation than the average graphics card (no vs multi frame generation).
  • 33.3% lower review score
    Reviews score
    4.8/ 10
    ✕
    What it is
    Average grade from independent test sites, magazines and video reviewers, 1–10.
    How we count it
    Grades: Each review's own score is converted to 1–10; without one, we grade the review ourselves. Big test sites count more, short reviews half as much as full tests. A video review counts half.
    Score: The average grade is 80% of the review mark; the number of reviews is the other 20% (1 review = 5, the typical number in this category = 8, the most-reviewed model = 10).
    Enough data: With few reviews, the average grade is pulled toward the typical grade of this category. No reviews yet: 7.6 for now.
    Share: Reviews count 25% of the overall score.
    Score distribution
    â–¼
    <55-66-77-88-99-10
    â–¼ Worse than 90% of models

    4.8

    Mark 7.3 is below the typical mark 7.9 of this category, so it cannot lift the score above the technical score

    80%

    7.9

    Average grade

    20%

    5.0

    Popularity

    Grades by review (1)
    1 video review, counting half as much as a written test. Reviews count 25% of the overall score.
    User ratings collected online are not part of this score.
    4.8 vs 7.2
  • 33.3% lower review score
    Reviews score
    4.8/ 10
    ✕
    What it is
    Average grade from independent test sites, magazines and video reviewers, 1–10.
    How we count it
    Grades: Each review's own score is converted to 1–10; without one, we grade the review ourselves. Big test sites count more, short reviews half as much as full tests. A video review counts half.
    Score: The average grade is 80% of the review mark; the number of reviews is the other 20% (1 review = 5, the typical number in this category = 8, the most-reviewed model = 10).
    Enough data: With few reviews, the average grade is pulled toward the typical grade of this category. No reviews yet: 7.6 for now.
    Share: Reviews count 25% of the overall score.
    Score distribution
    â–¼
    <55-66-77-88-99-10
    â–¼ Worse than 90% of models

    4.8

    Mark 7.3 is below the typical mark 7.9 of this category, so it cannot lift the score above the technical score

    80%

    7.9

    Average grade

    20%

    5.0

    Popularity

    Grades by review (1)
    1 video review, counting half as much as a written test. Reviews count 25% of the overall score.
    User ratings collected online are not part of this score.
    4.8 vs 7.2
  • 6.3% worse value for money
    NVIDIA Tesla K40 has worse value for money than the average graphics card (6.98 vs 7.4).
    6.98 vs 7.41
  • 65.1% lower boost clock speed
    This graphics card has a lower boost GPU clock than the average graphics card (875 MHz vs 2,505 MHz). The average graphics card has a boost GPU clock of 2,505 MHz.
    875 MHz vs 2,505 MHz
  • 40 fewer ray tracing cores
    This graphics card has fewer ray tracing cores than the average graphics card (0 vs 40). The average graphics card has 40 ray tracing cores.
    0 vs 40
  • 25 fewer compute units
    This graphics card has fewer compute units than the average graphics card (15 vs 40). The average graphics card has 40 compute units.
    15 vs 40
  • 61.2% lower base clock speed
    This graphics card has a lower base GPU clock than the average graphics card (745 MHz vs 1920 MHz). The average graphics card has a base GPU clock of 1920 MHz.
    745 MHz vs 1,920 MHz
  • 77.8% lower pixel rate
    This graphics card has a lower pixel rate than the average graphics card (35.8 GPixel/s vs 161.3 GPixel/s). The average graphics card has a pixel rate of 161.3 GPixel/s.
    35.8 GPixel/s vs 161.3 GPixel/s
  • 144 fewer AI cores
    This graphics card has fewer AI cores than the average graphics card (0 vs 144). The average graphics card has 144 AI cores.
    0 vs 144
  • 52.7% lower texture rate
    This graphics card has a lower texture rate than the average graphics card (178.8 GTexel/s vs 378 GTexel/s). The average graphics card has a texture rate of 378 GTexel/s.
    178.8 GTexel/s vs 378 GTexel/s
  • 81% lower FP32 performance
    This graphics card has a lower FP32 performance than the average graphics card (4.3 TFLOPS vs 22.6 TFLOPS). The average graphics card has FP32 performance of 22.6 TFLOPS.
    4.29 TFLOPS vs 22.6 TFLOPS
  • 10 years older
    This graphics card was released earlier than the average graphics card (2013 vs 2023).
    October 2013
    2013 vs 2023
  • 16 fewer ROPs
    This graphics card has fewer ROPs than the average graphics card (48 vs 64). The average graphics card has 64 ROPs.
    48 vs 64
  • 80.1% lower PassMark score
    This graphics card has a lower PassMark score than the average graphics card (4,495 points vs 22,638 points). The average graphics card has a PassMark score of 22,638 points.
    4,495 points vs 22,638 points
  • 88.2% lower compute score
    This graphics card has a lower compute score than the average graphics card (1,429 points vs 12,70.5 points). The average graphics card has a compute score of 12,070.5 points.
    1,429 points vs 12,070.5 points
  • 85.5% lower FP16 performance
    This graphics card has a lower FP16 performance than the average graphics card (4.3 TFLOPS vs 29.5 TFLOPS). The average graphics card has FP16 performance of 29.5 TFLOPS.
    4.29 TFLOPS vs 29.5 TFLOPS
  • 84.2% slower memory speed
    This graphics card has a lower effective memory speed than the average graphics card (3,000 MHz vs 19,000 MHz). The average graphics card reaches an effective memory speed of 19,000 MHz.
    3,000 MHz vs 19,000 MHz
  • 35.7% lower memory bandwidth
    This graphics card has a lower memory bandwidth than the average graphics card (288 GB/s vs 448 GB/s). The average graphics card has a memory bandwidth of 448 GB/s.
    288 GB/s vs 448 GB/s
  • 50% smaller L1 cache
    This graphics card has fewer L1 cache than the average graphics card (64 KB vs 128 KB). The average graphics card has 128 KB L1 cache.
    64 KB vs 128 KB
  • 5.6x larger process node
    This graphics card uses a larger manufacturing process than the average graphics card (28 nm vs 5 nm). The average graphics card uses a manufacturing process of 5 nm.
    28 nm vs 5 nm
  • Less advanced frame generation
    This graphics card has less advanced frame generation than the average graphics card (no vs multi frame generation).
  • No DLSS support
    This graphics card does not support DLSS, the average graphics card does.
  • No ray tracing
    This graphics card does not support ray tracing, the average graphics card does.
  • Older PCIe version
    This graphics card supports an older PCIe version than the average graphics card (3 vs 4).
    3 vs 4
  • No AI upscalers
    This graphics card does not support AI upscalers, the average graphics card does.
  • No XeSS support
    This graphics card does not support XeSS, the average graphics card does.
  • No mesh shaders
    This graphics card does not support mesh shaders, the average graphics card does.
  • Older Vulkan version
    This graphics card supports an older Vulkan version than the average graphics card (1.2 vs 1.4).
    1.2 vs 1.4
  • No DirectStorage support
    This graphics card does not support DirectStorage, the average graphics card does.
  • No Resizable BAR
    This graphics card does not support Resizable BAR, while the average graphics card does. 76.4% of Graphics cards support it.
  • Older DirectX version
    This graphics card supports an older DirectX version than the average graphics card (12 vs 12 Ultimate).
    12 vs 12 Ultimate
  • Older shader model
    This graphics card supports an older shader model than the average graphics card (5.1 vs 6.8).
    5.1 vs 6.8
  • 90.8% larger GPU die
    This graphics card has a higher GPU die size than the average graphics card (561 mm² vs 294 mm²). The average graphics card has a GPU die size of 294 mm².
    561 mm² vs 294 mm²
  • No sampler feedback
    This graphics card does not support sampler feedback, the average graphics card does.
  • Older FSR version
    This graphics card supports an older FSR version than the average graphics card (1 vs 3.1).
    1 vs 3.1
  • 69.1% fewer transistors
    This graphics card has fewer transistors than the average graphics card (7,080 million vs 22,900 million). The average graphics card has 22,900 million transistors.
    7,080 million vs 22,900 million
  • 2 fewer DisplayPort outputs
    This graphics card has fewer DisplayPort outputs than the average graphics card (1 vs 3). The average graphics card has 3 DisplayPort outputs.
    1 vs 3
  • No AV1 encoding
    This graphics card does not support AV1 encoding, the average graphics card does.
  • No AV1 decoding
    This graphics card does not support AV1 decoding, the average graphics card does.
  • No DSC support
    This graphics card does not support DSC, the average graphics card does.
  • No VRR support
    This graphics card does not support VRR, the average graphics card does.
  • More basic codec support
    This graphics card encodes more basic formats than the average graphics card (H.264 vs H.264/H.265/AV1/VP9).
    H.264 vs H.264/H.265/AV1/VP9
  • Older NVDEC generation
    This graphics card has a lower NVDEC generation than the average graphics card (1 vs 6). The average graphics card offers NVDEC generation 6.
    1 vs 6
  • No HDMI output
    This graphics card does not include HDMI output, the average graphics card does.
  • Not VR ready
    This graphics card is not VR ready, while the average graphics card is.
  • No 3D output
    This graphics card does not support 3D output, the average graphics card does.
  • 2 fewer fans
    This graphics card has fewer fans than the average graphics card (1 vs 3).
    1 vs 3
  • No dual BIOS
    This graphics card does not include dual BIOS, the average graphics card does.
  • No fan stop
    This graphics card does not support fan stop, the average graphics card does.
  • 28 °C higher idle temperature
    This graphics card has a higher idle temperature than the average graphics card (65 °C vs 37 °C). The average graphics card has an idle temperature of 37 °C.
    65 °C vs 37 °C
  • 9.3% higher TDP
    This graphics card has a higher TDP than the average graphics card (235 W vs 215 W). The average graphics card has a TDP of 215 W.
    235 W vs 215 W
  • 8 °C higher load temperature
    This graphics card has a higher load temperature than the average graphics card (75 °C vs 67 °C). The average graphics card has a load temperature of 67 °C.
    75 °C vs 67 °C
  • 66.7% higher idle power draw
    This graphics card has a higher idle power draw than the average graphics card (20 W vs 12 W). The average graphics card has an idle power draw of 12 W.
    20 W vs 12 W
  • No backplate
    This graphics card does not include a backplate, the average graphics card does.
  • No RGB lighting
    This graphics card does not include RGB lighting, the average graphics card does.

Video reviews

Hands-on testing, unboxing, and real-world video evaluations

7.0
Garry Patricio
≈ estimated
08/06/2021
Hands-on test · gaming in a water-cooled setup
Verdict
NVIDIA Tesla K40 works as a gaming card with a water-cooling mod, but only suits a small group willing to hack it.
Not graded

Third-party reviews

What customers like about NVIDIA Tesla K40

  • Extremely affordable high-capacity VRAM (12GB) for budget machine learning or AI experiments
  • Support for Error Correcting Code (ECC) memory, making it reliable for scientific computations
  • Good double-precision (FP64) performance (1.43 TFLOPS) compared to consumer cards of its era
  • Passive cooling design makes it suitable for server rack environments with high-airflow
  • Standard PCIe 3.0 interface allows it to be installed in many older workstation motherboards

What customers don't like about NVIDIA Tesla K40

  • Abysmally slow performance by modern standards; often outperformed by cheap, low-power consumer cards like the GTX 1650
  • High power consumption (235W TDP) results in poor performance-per-watt
  • Lacks integrated active cooling, requiring DIY fan modifications for use in standard desktop cases
  • No native video output, necessitating a second GPU to drive a display
  • Outdated 'Kepler' architecture lacks modern AI instructions and has limited software support in newer frameworks
  • Fidgety driver installation, especially when trying to use it in virtualised or non-headless environments

Verdict

The Nvidia Tesla K40 is a high-performance GPU accelerator based on the 28nm Kepler architecture (GK110B), featuring 2,880 CUDA cores, a base clock of 745 MHz (boostable to 875 MHz), and 12 GB of GDDR5 memory with a 384-bit interface providing 288 GB/s of bandwidth. It excels in scientific computing and big data analytics by delivering up to 1.43 TFLOPS of double-precision and 4.29 TFLOPS of single-precision performance, while offering critical enterprise features like ECC memory for reliability and Dynamic Parallelism for increased efficiency. Its main advantages include a large VRAM capacity for its era and superior double-precision capabilities compared to consumer cards; however, it lacks physical display outputs, carries a high 235W–245W TDP, and its aging architecture now faces compatibility issues with modern software and significantly lower performance compared to newer Pascal or Ampere-based alternatives.

Compare NVIDIA Tesla K40 with other products in the Graphics cards category

1Product to compare no. 1
VS
2Product to compare no. 2
VS
3Product to compare no. 3 (optional)

Graphic comparison of NVIDIA Tesla K40 and other products in the Graphics cards category

No results found