NVIDIA Ampere A100 GPU Gets Benchmark and Takes the Crown of the Fastest GPU in the World

[AleksandarK]

When NVIDIA introduced its Ampere A100 GPU, it was said to be the company's fastest creation yet. However, we didn't know how fast the GPU exactly is. With the whopping 6912 CUDA cores, the GPU can pack all that on a 7 nm die with 54 billion transistors. Paired with 40 GB of super-fast HBM2E memory with a bandwidth of 1555 GB/s, the GPU is set to be a good performer. And how fast it exactly is you might wonder? Well, thanks to the Jules Urbach, the CEO of OTOY, a software developer and maker of OctaneRender software, we have the first benchmark of the Ampere A100 GPU.

Scoring 446 points in OctaneBench, a benchmark for OctaneRender, the Ampere GPU takes the crown of the world's fastest GPU. The GeForce RTX 2080 Ti GPU scores 302 points, which makes the A100 GPU up to 47.7% faster than Turing. However, the fastest Turing card found in the benchmark database is the Quadro RTX 8000, which scored 328 points, showing that Turing is still holding well. The result of Ampere A100 was running with RTX turned off, which could yield additional performance if RTX was turned on and that part of the silicon started working.


View at TechPowerUp Main Site

 

[SamuelL]

Maybe low clocks on this early model? Obviously apples to oranges comparison between server/workstation and a consumer gaming card, though not really a great look based on specs alone: 7nm, new memory, and roughly a 33% performance improvement for about 37% more cuda cores?

 

[Lionheart]

I'm salivating at the mouth while my wallet is starting to cry.

 

[Anymal]

I'm salivating at the mouth while my wallet is starting to cry.

Yes, would 1500eur for top geforce ampere be enough? How much faster than 2080ti will be?

 

[r.h.p]

if this is true awesome ... then why isn't there a picture of the finished card ? That picture looks like a cinebench render

 

[FocFireMK]

2080 Ti fastest GPU? you forgot Titan RTX and Titan V

 

[Deleted member 185088]

Not very impressive given the specs unless like suggested in the first comment it runs very slow.

 

[maxitaxi96]

if this is true awesome ... then why isn't there a picture of the finished card ? That picture looks like a cinebench render

This "card" is a data-center GPU only available in the DGX A100. 8 Of those things in the Render in one machine. This benchmark result only shows the maximum potential of the Ampere Silicon (yet). There are yet no quadro or geforce cards out. Think a better Titan Ampere.

 

[Caring1]

So rendering is a basis for assessing GPU's now?

 

[dyonoctis]

So rendering is a basis for assessing GPU's now?

For the people who aren't buying those gpus just for gaming it is.

 

[Bruno Vieira]

Dont forget, the AX100 is 400w and the 2080ti is 280w, and Nvidia never listed RTX support in ANY material so far. It is a computing card.

 

[EarthDog]

Look at all the chubbies on the NV peeps so far...

So, solid rendering performance of this compute card. Cool.

 

[Hexa]

I think this says nothing about the top end consumer ampere cards, they only maybe are 7nm and for sure not hbm2. But it goes into the other direction as well, the A100 was build for a server environment with multiple of them side by side, its clocks are optimized not just so power will be saved, but also so they don't overheat. Besides, do we even know the A100 is what will be powering the card? in the last 2 generations they committed to a naming scheme, where the names of the 80 ti´s end in a 2(GP102 for Pascal and TU102 for Truing)

 

[Caring1]

So someone compares apples and oranges and we're meant to be impressed by a 47% difference.
Let me know when he wakes up and compares it to compute cards only.

 

[theonek]

this is not a consumer chip/card at all, so don't be mistaken by any of these bench numbers....

 

[CoD511]

Impressive by any means. I'm curious how the 250w PCIe version of the card performs compared, considering Nvidia stated 90% of the performance of the SXM4 card.. it'd be nice to know what that 90% is, in different metrics for performance.

Also, extremely curious once perhaps yields are more mature (I feel uncertain in how long that might take, considering the gigantic size of the each die), how a fully enabled (potentially higher clocked) die of an A100 would perform. This only seems like good news though for any corporation interesting in it, considering how fully integrated its compute capabilities are into that single die/chip.

 

[Steevo]

I wonder the pricetag considering yield loss on that big of a chip, and the cost of the other components.

 

[aQi]

Ready player one

 

[Minus Infinity]

Not very impressive given the specs unless like suggested in the first comment it runs very slow.

I agree, very ordinary score unless it’s running very low clocks. Based on all the leaks Big Navi and 3080 Ti will do at least as well.