dobry den, totok tato karta mi rozbehne hunters dream 2 na fullscreene ?
HP GFX NVIDIA A800 40GB
Graphics Card - 40GB GDDR6, NVIDIA Ampere (, 765 MHz), Boost 1410 MHz, PCI Express x16 4.0 , 5120Bit, width 266,7 mm
Show full description
20,979.99 €
Warranty 24 months
Code:
EN064p7e
Producer's Website:
HP GFX NVIDIA A800 40GB
20,979.99 €
Ordered on request
Buy
HP GFX NVIDIA A800 40GB
HP GFX NVIDIA A800 40GB graphics card
NVIDIA A800 40GB Active brings data centre computing power directly to your desk. This card is not designed for graphics rendering in the traditional sense, but for the most demanding computational tasks: training and inference of large AI models, scientific simulations, HPC analysis and data science development. The NVIDIA Ampere architecture with 6912 CUDA cores and 432 third-generation Tensor cores delivers 19.5 TFLOPS of FP32 performance, 9.7 TFLOPS of FP64, and a class-leading 623.8 TFLOPS in tensor operations. Forty gigabytes of GDDR6 memory with an extremely wide 5120-bit bus ensures that even the largest models and datasets will not hit a memory bottleneck. For developers and researchers who need supercomputer performance on a local workstation, the A800 40GB Active defines what is possible today.
GDDR6 / 40GB
Operating memory
PCIe 4.0 x16
Bus interface
2× 8-pin
Auxiliary power
NVIDIA A800
Chip architecture
850W
Recommended source power
Key features of the HP GFX NVIDIA A800 40GB graphics card
- Features 40GB of GDDR6 memory with an extra-wide 5120-bit bus, enabling it to handle the most demanding AI models and scientific simulations without memory bottlenecks
- The NVIDIA Ampere architecture with 6912 CUDA cores and 432 third-generation Tensor cores delivers a boost frequency of up to 1410MHz and peak computing performance of 19.5 TFLOPS in FP32, 9.7 TFLOPS in FP64 and 623.8 TFLOPS in tensor operations
- Third-generation NVIDIA NVLink technology lets you connect two A800 40GB Active cards with 400GB/s of bidirectional throughput to get 80GB of shared memory for extremely large models and datasets
- The Multi-Instance GPU (MIG) function enables you to partition the card into up to 7 independent 5GB GPU instances, allowing a single card to serve multiple developers or tasks simultaneously
- A800 40GB Active is a 'headless' card with no display outputs, designed exclusively for computational tasks; it requires a supplementary graphics card such as an NVIDIA T1000 or RTX 4000 Ada Lovelace for display
- A dual-slot active cooler maintains stable performance at 240W TBP, even during long-running AI training cycles and HPC computations
- The PCI Express 4.0 x16 interface and 40MB L2 cache (almost 7× larger than the previous generation) ensure a smooth flow of data between the memory, GPU and the rest of the system
A comprehensive platform for AI development and deployment
Training an AI model is only half the battle – the other half begins when you need to deploy it reliably. NVIDIA A800 40GB Active therefore comes with a complete software platform for developing and deploying generative AI, image and video recognition, speech models and predictive analytics in a production environment. The card natively supports PyTorch, TensorFlow, CUDA-X libraries and containerised workflows via NVIDIA NGC, so the transition from experiment to production happens without rewriting code. In addition, the Multi-Instance GPU function lets you partition a single card into up to seven independent 5GB GPU instances, making the A800 40GB Active an ideal platform for developer teams sharing a single workstation or for parallel A/B testing of models.
Scientific computing and HPC tasks locally
Where conventional graphics cards hit their limits, NVIDIA A800 40GB Active is just getting started. The card delivers a full 9.7 TFLOPS of FP64 double precision, a performance previously reserved for data centres. It is therefore well-suited for scientific simulations, CFD analysis, quantum chemistry, seismic modelling and financial Monte Carlo calculations, where every decimal place matters. For tasks with extreme memory requirements, you can connect two cards using third-generation NVIDIA NVLink technology with a bidirectional throughput of 400GB/s to get 80GB of fully coherent shared memory. For engineers, scientists, and analysts, this means that simulations which previously required access to a cluster can now run right under their desk – with full control over data and task prioritisation.
Specifications
Chip manufacturer
Model designation
Chip architecture
Professional
Card profile
Chip
Chip manufacturer
Model designation
Chip architecture
Performance & usage
Memory
Frequency
765 MHz
Core clock speed (Boost clock)
Interface
Technology
Special graphics card features
Cooling system
Active
number of fans
Dimensions
Width
Depth
Consumption
240 W
AI
Total number of TOPS
INT8 TOPS
Max. number of MIG instances
Supported precision
Tops
Total number of TOPS
AI specifications
Primary AI use
More Specifications
HP Inc Czech Republic s.r.o Za Brumlovkou 5/1559 140 00 Praha 4 VAT ID: CZ 03208427 Radek Mattas, [email protected]"
GO Europe GmbH, Zum Kraftwerk 1, Hattingen 45527, Germany, +49232468010, [email protected]
Hewlett-Packard was founded in the difficult times of the Great Depression. The founders were a pair of friends whose name the company still proudly bears. They started their business in an unobtrusive garage near the city of Palo Alto. It is now a national monument. HP’s success lay not in copying existing products, but in the ability and courage to come up with something new.
The first commercial triumph was an oscillator that surpassed all competition in quality, yet sold at a quarter of the price. In 1968, HP released their first desktop computer - a desktop calculator. The company currently manufactures products primarily related to computer technology - computers and laptops, printers, scanners, digital cameras, servers, and last but not least, calculators.
Note:
This product is not an independently functional unit and may require professional installation.
Specifications can be changed without notice. Images are for illustrative purposes only.
