[Cryptech Core] Speed grade differences in cost and performance

Joachim Strömbergson joachim at assured.se
Wed Oct 31 15:09:42 UTC 2018


Aloha!

On 2018-10-30 17:29, Peter Stuge wrote:
> Joachim Strömbergson wrote:
>>>> We are currently using an Artix-7 T200 device with -1 speed grade. Even
>>>> though the tool supports it, at least for commercial or industrial rated
>>>> devices, the only other speed grade is -2.
>>>
>>> Sorry, what about -3 ? Unsupported by no-cost Vivado, or?
>>
>> It is not a lack of support in the tool. It is lack of devices with the
>> "-3" speed grade. At least devices I'm able to find. Xilinx only lists
>> -1 and -2 for commercial an industrial devices.
> 
> https://www.digikey.com/products/en?keywords=122-1860-ND
> 
> "Extended" temp maybe? 0-100 vs. 0-85 Commercial and -40-100 Industrial.

I stand corrected. Yes. That is a potential option. I reran the sha256
implementation and got a minimum clock period of 6.673 ns.

> More speed more cost of course. How much faster would it in fact be?

From previous tests yesterday:
-2 and 6ns target: 7.600ns
-1 and 6ns target: 9.351ns

So from -1 to -3 we get 2.67 ns improvement. Or 28% better performance
margin. More importantly, we might be able to run the cores at 3 *
fmc_clock which would be a pretty significant improvement from today.
And 2 * fmc_clock would be possible with a lot of timimg headroom.

The price as indicated by digikey as you found is a bit of an increase.
But not a doubling in price compared to -1 devices.

Good find.

-- 
Med vänlig hälsning, Yours

Joachim Strömbergson
========================================================================
                               Assured AB
========================================================================

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL: <https://lists.cryptech.is/archives/core/attachments/20181031/9b9cd620/attachment.sig>


More information about the Core mailing list