Shogan.tech

Hashcat RTX 3090 Benchmarking and Performance

June 24, 2021 by Sean

I picked up an nVidia RTX 3090 toward the end of last year. In hindsight, I was lucky to have purchased it early on soon after release. The GPU shortage has caused a massive spike in prices and this card is now worth double what I originally paid for it! Anyway, acquisition story aside, I was curious how it would perform in a Hashcat benchmark (all) run. Here are my Hashcat RTX 3090 benchmark results.

For a quick and easy run I’m using the hashcat 6.2.2 (Windows) binary.

.\hashcat.exe -b --benchmark-all

The performance seems on-par if not slightly higher than some other RTX 3090 benchmarks I have seen around. An impressive set of results.

I am running the MSI GeForce RTX 3090 Ventus 3X OC 24GB model card. I upgraded from a GeForce 1080 Ti (12GB) model and the hashing speeds are way faster. The 3090 is a power hungry beast though. It gets hot and the fans are noisier than my 1080 Ti’s were. To ensure my system’s power delivery was up to the task, I also upgraded to a Seasonic Focus PX-850 850W 80+ Platinum at the same time.

Here is a shortened log of my benchmark –all run:

CUDA API (CUDA 11.2)
====================
* Device #1: GeForce RTX 3090, 23336/24576 MB, 82MCU

OpenCL API (OpenCL 1.2 CUDA 11.2.109) - Platform #1 [NVIDIA Corporation]
========================================================================
* Device #2: GeForce RTX 3090, skipped

Benchmark relevant options:
===========================
* --benchmark-all
* --optimized-kernel-enable

Hashmode: 0 - MD5
Speed.#1.........: 67033.9 MH/s (40.78ms) @ Accel:32 Loops:1024 Thr:1024 Vec:8

Hashmode: 10 - md5($pass.$salt)
Speed.#1.........: 66278.8 MH/s (41.26ms) @ Accel:32 Loops:1024 Thr:1024 Vec:8

Hashmode: 11 - Joomla < 2.5.18
Speed.#1.........: 64972.6 MH/s (42.10ms) @ Accel:32 Loops:1024 Thr:1024 Vec:8

Hashmode: 12 - PostgreSQL
Speed.#1.........: 64460.9 MH/s (42.44ms) @ Accel:32 Loops:1024 Thr:1024 Vec:8

Hashmode: 20 - md5($salt.$pass)
Speed.#1.........: 35775.1 MH/s (76.66ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 21 - osCommerce, xt:Commerce
Speed.#1.........: 36124.8 MH/s (75.92ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 22 - Juniper NetScreen/SSG (ScreenOS)
Speed.#1.........: 35747.7 MH/s (76.72ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 23 - Skype
Speed.#1.........: 35632.9 MH/s (76.96ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 24 - SolarWinds Serv-U
Speed.#1.........: 35107.4 MH/s (78.12ms) @ Accel:32 Loops:1024 Thr:1024 Vec:1

Hashmode: 30 - md5(utf16le($pass).$salt)
Speed.#1.........: 65511.3 MH/s (41.73ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 40 - md5($salt.utf16le($pass))
Speed.#1.........: 36398.3 MH/s (75.35ms) @ Accel:32 Loops:1024 Thr:1024 Vec:4

Hashmode: 50 - HMAC-MD5 (key = $pass)
Speed.#1.........: 10893.9 MH/s (62.90ms) @ Accel:8 Loops:1024 Thr:1024 Vec:1

Hashmode: 60 - HMAC-MD5 (key = $salt)
Speed.#1.........: 22468.1 MH/s (60.99ms) @ Accel:32 Loops:512 Thr:1024 Vec:1

Hashmode: 70 - md5(utf16le($pass))
Speed.#1.........: 64396.2 MH/s (42.49ms) @ Accel:32 Loops:1024 Thr:1024 Vec:1

Hashmode: 100 - SHA1
Speed.#1.........: 21045.1 MH/s (65.11ms) @ Accel:16 Loops:1024 Thr:1024 Vec:1

Hashmode: 101 - nsldap, SHA-1(Base64), Netscape LDAP SHA
Speed.#1.........: 20874.3 MH/s (65.66ms) @ Accel:16 Loops:1024 Thr:1024 Vec:1

Hashmode: 110 - sha1($pass.$salt)
Speed.#1.........: 21217.0 MH/s (64.60ms) @ Accel:32 Loops:512 Thr:1024 Vec:1

Hashmode: 111 - nsldaps, SSHA-1(Base64), Netscape LDAP SSHA
Speed.#1.........: 20608.3 MH/s (66.51ms) @ Accel:16 Loops:1024 Thr:1024 Vec:1

Full results can be downloaded here:

hashcat-rtx-3090-benchmark-all Download

As for the PC build around the RTX 3090, here are a few photos…

You might notice an AIO installed, but not connected – I was in the process of testing a dual 240mm radiator (AIO) versus a high performing Noctua air cooler, so had left it in the chassis during transition.

I’ll see if I can run the same benchmark suite on my uBuntu install and update the results here. I have not tested the RTX 3090 card under this OS yet so I’m not sure if I’ll run into any driver issues or not.

Funny story of how I ended up with an RTX 3090 (it is a bit overkill!)…

Back in September or October of 2020 (I forget when the 3xxx series launch was), I had pre-ordered the MSI RTX 3080 at MSRP. I was 180 or so in the pre-order queue after months of already waiting. My queue position was barely changing week over week and I got impatient. I saw a handful of RTX 3090 cards come in stock at a local retailer and purchased one.

These cards would generally remain in stock for a few days due to everyone holding out for the much cheaper (at the time) RTX 3080 pre-order promises.

It was a lucky break for me, as those 3080 cards never came for most in that queue. GPU mining and GPU shortages made sure of that. Prices sky rocketed. Looking up this card now I see it costs almost double what I originally paid last year (if you can even get stock that is).

Now I just hope the card lasts at least a few years or more so I don’t ever have to worry about RMA and stock levels…

The 486 era – My Computer Hardware Journey

August 15, 2021May 30, 2021 by Sean

My computer hardware journey started in the mid 1990s, when I was about 10 or 11 years old. It started out with me dismantling and reconfiguring various 486 machines. These were also known as i486 or Intel 80486 systems.

Electronics had always been a big curiosity for me. As a child I liked to open up devices and extract their internals. I would collect PCBs and join them in nonsensical ways, imagining I had invented new and wonderful devices, capable of many great things.

To illustrate my passion for tinkering with electronics, here is a quick story going back to my toddler years: My parents had noticed my love for tools as a toddler and had tried in vain to make them safe for me. According to my father, he once added a solid lump of pratley steel putty to the end of a sharp screwdriver for me. An attempt to make it safe for me to handle. As soon as I saw this, I threw it to the ground in protest. I only wanted real tools. One of the toys I had was an old 5+1⁄4-inch floppy disk drive!

My First 486 Build

Fast forwarding to the mid 1990s, I was far more capable with tools and had begun to scavenge surplus PC hardware. My uncle owned a PC repair shop where I grew up. As far as I recall I managed to scavenge old discarded parts from his collection, as well as other sources. Schools or universities that would throw out old machines as they cycled or upgraded them, cousins or family friends that had no need for old systems, etc…

The very first working system that I managed to modify was an Acer branded 486 desktop system. It had a horizontally aligned desktop chassis, one that allowed you to place a CRT monitor on top of.

This isn’t the exact one, but mine looked just like this, minus the CDROM drive.

I remember it had a 486 SX-33 CPU. A rather weakly clocked 486 processor, that didn’t even require a heatsink and fan assembly. The 486 SX processors had their FPU (floating point unit) disabled. I imagine this reduced cost and heat output.

This wasn’t exactly a scratch build, but it allowed me to get my feet wet in the world of computer hardware.

Upgrading the CPU

I swapped out the CPU for a 486 DX2-66, a much faster processor, adding a small aluminium heatsink and fan. The fan was powered by a molex power connector.

Memory and Upgrades

Another upgrade on this particular system that I remember was going from 4MB of RAM (SIMM, or single in-line memory module) to 8MB.

I don’t recall if the memory was the 30 or 72 pin type.

Hard Drive Configuration

Among other upgrades, the hard disk drive got swapped out for a slightly larger one. I think the capacity was somewhere around 400MB.

This upgrade had me reading manuals and learning about the importance of jumper placement. The drives connected with an ATA ribbon cable. It was important that the jumper on the back of the hard drive was configured according to the other devices sharing the same cable.

The jumper could be configured to make a drive device 0 or 1 (also known back then as the master or slave device) when sharing the same ATA cable.

The cables themselves were also rated for different transfer speeds. For example 16 or 33 MB/s, and later cables allows Ultra DMA to be used for speeds above 66 MB/s.

Future 486 Systems and Nostalgia

That Acer machine was a good learning experience for me. Soon after that time, I started building and configuring more 486 systems. I had all sorts from 486 DX2-66 to DX4-100 machines.

I started a little side hustle, where I would pick up old components and build fully configured systems, selling them in the local paper’s classifieds section.

Thinking back, this was seriously impressive for a kid my age. By the late 90s (at age 12 or so) I was earning chunks of cash from selling systems that I had built for next to nothing.

Fixing Old CRT Monitors

One of the big wins for me was when I learned from my uncle how to ‘fix’ blurry CRT monitors.

If opened up, all monitors had analog adjustment dials (potentiometers) that could be turned with a screwdriver to control the flyback. One of these would control focus.

I would gather up old CRT monitors that nobody wanted anymore because they had gone ‘blurry over time’. The fix was simply to adjust the focus potentiometer inside while the monitor was on. Of course this was very dangerous as these old monitors contained high voltage capacitors that remained charged up for a while, even after power was disconnected. However, having learned from an adult, I was very careful about this, and the payoff was being able to make cash from selling full systems with monitor included.

This was an amazing time for me as a kid. I learned how to operate DOS, and further operating systems such as Windows 3.1 and further on, Windows 95.

Computer Knowledge Acquired from my ‘486 era’

Here are some random things I learned in the 1990s during my time tinkering with 486 systems. I guess some of the time was also spent with older 386 machines too.

DOS configuration using the CONFIG.SYS and AUTOEXEC.BAT files
Expanded memory manager configuration with EMM386.
Configuring CDROM device drivers in DOS and setting them up on boot.
Writing simple batch (.BAT) scripts.
Simple QBASIC programming.
A little later on, Turbo Pascal programming basics.
Compression and decompression with tools like ARJ.
Manual sound card configuration by selecting specific addresses and channels.
Networking two machines using serial or parallel cables (COM or LPT), and using laplink software to copy files between.

Gaming and PC use in the 1990s

Not all of my computer time as a kid was spent working with computer hardware. I enjoyed playing games just as much.

Some of my most memorable and favourite DOS games that I enjoyed playing on the first 486 system my parents owned, (and later on the 486 systems I built were):

X-Wing and Tie Fighter
UFO – Enemy Unknown
Ultima VIII: Pagan
Master of Orion (MOO)
Wacky Wheels
Death Rally
Screamer
Cannon Fodder
Strike Commander
Wing Commander

I also fondly remember writing science fiction (short stories) in word processing software in DOS and then printing them out on a dot matrix printer. I would take these to school and share with friends (who would also do the same thing).

Follow my Retro Hardware blog post series:

Notes:

The featured image of the 486 SX CPU is sourced from: https://en.m.wikipedia.org/wiki/File:Intel_i486_sx_25mhz_2007_03_27b.jpg and comes with this creative commons license.

Using Node.js Streams to Create a Toy Version of jq

April 25, 2021 by Sean

To play around with Node.js streams, I made a simple ‘toy’ version of jq, a handy command line JSON processor.

To be clear, the real jq utility is very lightweight and has a ton of functionality. In this post I’ll be mimicking some of it’s functionality and using Node.js to do so.

This will of course result in a bloated tool with way more bundled in than what we actually need.

That being said, it’ll be a useful exercise to go through to learn a little bit about Node.js streams.

In the github repository, you’ll see how easy it is to hack together a very simple command-line tool to process data streamed in from stdin.

node.js stream transform example with PowerShell pipeline processing. — JSON going in to the tool, being filtered, and then converted to an object in PowerShell through the pipeline.

Node.js Streams

The stream documentation describes them as as:

A stream is an abstract interface for working with streaming data in Node.js. The stream module provides an API for implementing the stream interface.

For this example I’ll be jumping straight into the Transform Stream – stream.Transform. This is a duplex stream, where the input would usually be related to the output in some way.

Transforming Input from stdin

The basic use of a Transform stream in Node.js (to process input from stdin) looks like this:

const {Transform} = require('stream')
const TransformStream = new Transform;
TransformStream._transform = (chunk, encoding, callback) => {
    // do something with the chunk
    console.log(chunk.toString().toUpperCase());
    callback();
}

process.stdin.pipe(TransformStream);

Toy jq Utility With Transform Streams

I created a ‘toy’ version of jq using a Node.js Transform stream. It’s a very quickly hacked together example, so don’t expect it to do everything that jq can do. I’m also fully aware that the real jq utility is a very lightweight tool and that doing this in the Node.js runtime adds a lot of unecessary bloat!

This is purely for demonstration purposes.

Packaging up the Node.js app with pkg, we get a platform specific binary called toyjq.

Examples

Pretty print input JSON

cat ./example.json | toyjq-linux

{
  name: 'directoryobject',
  path: '/path/to/directoryobject',
  type: 'Directory',
  children: [ { foo: 'bar' }, { foo: 'bar1' } ]
}

Output the `type` field only from input JSON:

cat ./example.json | toyjq-linux

cat ./example.json | toyjq-linux '.type'

"Directory"

Output just the `name` and `children` fields in the input JSON:

cat ./example.json | toyjq-linux '{name, children}'

{
  name: 'directoryobject',
  children: [ { foo: 'bar' }, { foo: 'bar1' } ]
}

Assuming now the Windows platform version of toyjq, and using a PowerShell cmdlet for this example…

Select the children array, convert it to an object in PowerShell and then select the last item in that object array:

(cat .\example.json | .\toyjq-win.exe '.children' | ConvertFrom-Json).foo | Select -Last 1

bar1

The above examples show how you can easily process data from one input pipeline (stdin in this case) and send it along through the pipeline using Node.js streams.

You can find the example toyjq app on my GitHub repository.

SSM and socat Port Forwarding to Access Private VPC Resources

April 6, 2021April 2, 2021 by Sean

AWS System Manager Session Manager added the port forwarding feature, announced in this blog post back in 2019. In this post I’ll show you how to leverage SSM and socat port forwarding to access systems in a private subnet that don’t have the SSM agent installed.

You’ll use an SSM agent enabled EC2 instance as an initial target for the ssm port forward session. On this instance, you’ll run socat as a relay for the incoming TCP session to the other instance that does not have the SSM agent.

What is socat?

To quote the official man page, socat (SOcket CAT) is a multipurpose relay. It is a command line tool that establishes two bidirectional byte streams and transfers data between them.

You can use it to connect all sorts of channels. For example:

files
pipes
devices
sockets, such as TCP, UDP, IPv4, etc
SSL sockets
programs

SSM and socat Port Forwarding Example

In my example I have an AWS EMR (Elastic Map Reduce) master node running a web dashboard for ganglia in a private VPC subnet.

I don’t want to add a bastion host / jump box or provide SSH access from the public net.

SSM would provide a nice way for me to connect a remote session, or port forward using IAM authentication and negating the need for any ingress security group rules, but only if I had the SSM agent available on this instance.

Seeing as though the EMR master node is not SSM agent enabled, and I can’t use SSM port forwarding directly to this instance, we could use an interim machine with SSM as a jump box.

Example Configuration

Here is how I configured port forwarding in my use case to access ganglia on a private instance EMR node.

The EC2 instance with SSM agent must have an IAM policy attached that allows the relevant ssm access. The blog post linked above has instructions. In a nutshell though, most standard Amazon AMIs include the ssm agent. Your EC2 instance profile should include the required actions too. The AmazonSSMManagedInstanceCore managed policy includes these.
Install socat on the SSM agent enabled interim machine the private subnet. For this I connected an SSM session to get shell access and ran sudo yum install -y socat
Now I needed to open a source channel for the SSM port forward aws cli command to connect, and connect that source to the destination of the EMR master node running ganglia.

socat TCP4-LISTEN:8080,fork,reuseaddr TCP4:10.0.4.149:80

The command listens on port 8080, and forwards TCP to the EMR node, 10.0.4.149 on port 80. Importantly, the command uses fork and reuseaddr to allow multiple connections.

Next is to use the AWS CLI ssm start-session command to start a port forwarding session to the interim instance with the SSM agent running. Grab the Instance ID for the EC2 machine and:

aws ssm start-session --target {your-instance-id-here} --document-name AWS-StartPortForwardingSession --parameters '{"portNumber":["8080"],"localPortnumber":["8089"]}'

If you setup socat correctly to listen on port 8080, then the connection should be opened and accepted.

Now you can simply open a web browser locally and direct it to http://localhost:8089/ganglia to access ganglia on the remote EMR master node.

Accessing EMR cluster memory stats via the remote port forwarded session.

Closing

AWS SSM is a useful tool to get access to instances in a secure, audited fashion without needing to open up risky SSH access or other remote ports to the public internet.

When constrained and needing a jump across to an instance without the SSM agent you can leverage tools to help. Socat is one such tool that can facilitate this within the private network.

SpotiPod – Spotify Streaming Device from an iPod Classic

March 28, 2021 by Sean

I recently came across the sPot: Spotify in a 4th-gen iPod (2004) project on hackaday.io by Guy Dupont. This post is my go at building Guy’s project from the ground up – the SpotiPod.

Here is my version, up and running, albeit with some hardware leaking out the side for now…

The main components required in terms of hardware are:

iPod Classic (4th gen) – at least the device case and clickwheel components
A Raspberry Pi Zero W
2″ LCD display (see note further on about an alternative)
3.7V 1000mah LiPo battery
Adafruit boost module (to boost 3.7V to 5V required by the Pi, LCD, etc…
Adafruit USB charge controller module

Once it’s all connected and configured, you’ll be able to load up your own Spotify playlists and libraries, browse them, and play them all from your iPod Classic device over a bluetooth speaker or remote system.

The SpotiPod build almost complete. — Some wire reduction still required before I can completely close it up.

SpotiPod High-level Build

I won’t be getting into the low-level parts of the build, so for specific details I would recommend viewing Guy’s project log and branching off the posts there.

Initial Raspberry Pi Zero W setup with 2″ LCD display

I started off loading Raspbian Lite OS onto an 8GB microSD card and booting up a barebones Pi Zero W. The W version is important because it has WiFi and Bluetooth on the board.

My first task was to configure SSH to start automatically on boot. Useful when I have no display to start with. I also added some of the required software components to begin testing and messing around with, including redis server, compiling the click c++ application that Guy wrote to interface with the iPod’s clickwheel, and setting up some of the X11 components for a basic UI.

Soldering up the connection for the 2″ LCD display was my first hardware connection task. My soldering iron might come out once in a year, so I’m a bit of a novice in this area, but I managed the task on my first go.

Connecting the LCD to Raspberry Pi Zero via Composite Connector

testing the LCD — First LCD test successful

I posted more details here on the 2″ LCD connection and configuration in software.

Connecting the Charge Controller and Voltage Boost Modules

Next up I focused on getting the power delivery working. The goal was to be able to charge the LiPo battery via USB and have the circuit supply 5V to all SpotiPod components where required.

I familiarised myself with the Adafruit module documentation and pinout diagrams before connecting the circuit up.

Adafruit Powerboost 1000 module
Adafruit Charge Controller

Here is a useful diagram to follow, posted by Kakoub on the hackaday.io project page, re-hosted here in case the imgur upload ever disappears:

So after getting power delivery and connections soldered in place, and precariously placing all the components away from eachother to prevent shorts, here is where I was at:

SpotiPod Powerboost and Charge modules connected — Powerboost and Charge modules connected

Testing the Click wheel and Software

The click wheel connectivity is made easier with an 8 Pin FPC cable breakout board. I hooked this up next, soldering the 4 wires required for power and data.

The click wheel ribbon connector could then snap into the breakout board ribbon connector. This is by far the most delicate part of the build in my opinion. The ribbon cable is super thin and delicate.

Connecting the 8 pin FPC breakout board. My soldering skills slowly coming back with a bit of practice…

Testing the interface was a case of compiling the clickwheel program with gcc over an SSH connection then executing it with everything connected.

If wired up correctly, the program will output touch and click data to stdout.

iPod Classic click wheel testing over SSH

Gutting the old iPod Classic

I managed to snag an old iPod Classic 4th gen off eBay for about £20. It wasn’t working, but had the two bits I needed – the chassis, and the clickwheel.

Breaking out my trusty iFixit essentials toolkit, I set about opening it up to remove the unecessary components.

The method I found to work was to pry it open on the left and right edges using the pry tool. Once you can get into one of the edges, slide the tool around, and things get easier.

Insulating the case and components

With most of the electronics connected, I began insulating things with polyimide tape. Most importantly, the metal iPod case. I put down about three layers of tape and tried to cover all parts as best I could.

Insulating the iPod classic shell for the SpotiPod build.

Installing into the iPod Classic Case

This is the tricky part. It’s quite difficult to squeeze all the SpotiPod hardware in.

I started out by strengthening all my soldering connections with a bit of hot glue.

adding hot glue to the soldered connections for the SpotiPod — Adding hot glue to the soldered connections

After a bit of arranging, squeezing, and coercing, everything fits… Mostly!

Things are still not perfect though. I need to reduce my wire lengths before I can get the case to fully close.

For now though, everything works and I have a fully functional SpotiPod!

Tips and Tricks

I’ve put together a list of things that might help if you do this yourself. These are bits that I recall getting snagged on:

Make sure you setup all the X11 and software dependencies correctly. Getting Openbox and the frontend application to launch on start up can be tricky and this is crucial. Pay attention to your /etc/X11/xinit/xinitrc and /etc/xdg/openbox/autostart configurations.
You don’t have to use the more expensive Adafruit composite LCD display. Ricardo’s build at RSFlightronics uses a much cheaper LCD and some creative approaches to get display output working.
Watch out for the click wheel ribbon orientation when you connect it to the breakout board!
Use thin and short length wires for connections where possible. Not too short though as it is useful to be able to open the device up and put the two halves side-by-side.
Make sure you have a Spotify Premium subscription. I can’t remember exactly, but I’m sure that creating your own app to get your client and secret keys, or some of the scopes required for the app will only work on Premium. (It might have even been spotify connect).
You’ll need to configure your own Spotify App using the Spotify Developer portal. Keep your client and secret keys safe to yourself. Remember to setup environment variables with these that the openbox session can access.
The frontend/UI application has a hardcoded reference to the Spotify Connect device as “Spotifypod”. Keep things simple by setting your raspotify configuation to use this name too, otherwise you need to update the code too.
If you’re struggling to get the software side working at first, it can really help to setup VNC while you debug things. This allows you to get a desktop environment on the Pi Zero and execute scripts or programs in an x session as openbox would.

Thanks again to Guy Dupont for his excellent SpotiPod project and idea. Putting this all together really makes for a fun and rewarding hardware/software hacking experience.