Category: Computer Hardware

Using a Samsung T7 USB-C external SSD to run lab VMs

My use case

When I started more actively playing around on different hacking CTF platforms I felt the need to carry my “CTF pwn box” with me. Syncing the Kali VM constantly between my workstation and my laptop was time consuming and well I mostly just forgot to do it.

My doubts

Previously only having owned USB3 HDD’s for back up purposes and having seen their latency issues/slowness I was quite skeptical.. I was wondering about the latency penalties that would come from having the NVME drive over USB instead the m2 slot. And as reviewers somehow didn’t want to cover the latency aspects I was left wondering.

How it actually is with the Samsung T7 portable drive

After having done some research I opted to buy the 1TB Samsung T7 Portable SSD and long story short my use case works fine. VM-s run like a dream on it.

When comparing latency on a machine running on my internally installed m2 NVME drive and the USB-C connected Samsung drive there is no difference that I can feel when doing labs.

But as feeling isn’t actually accurate I thought I’d actually measure the IO performance. In order to measure latency/IO performance I used DD with a small block size. That little idea I got from a nixCraft article.

The internal M2 NVME I use in the test is the Western Digital 1TB WD Blue SN550. Although it doesn’t change much the system itself is a Ryzen 5800x on a X570 chipset motherboard.

The results show in my case that the internal M2 drive is about 10% faster. Yet it is something that would make VMs run sluggish.

Results for the internal 1TB WD Blue:

Results for the USB-C 1TB Samsung T7:

One more thing that I feel must be said about the portable M2 drive is.. That although synthetic benchmarks like CrystalDiskMark show performance “like its written on the box” then real life differs from it. Namely Samsung promises “up to 1000MB/s” writes and having tried it on multiple computers actually copying files onto it lands somewhere around 350MB/s for me. And I was copying large files from internal M2 drives capable of more than that. And not so large that they should overwhelm a M2 drives cache functionality/slow it down. I used different files with sizes ranging from 1GB (like crystal disk mark) and up to 20GB.

It doesn’t necessarily make the T7 a bad drive, but it’s just something to take into account. For me I’m still pleased with the drive, as it has served me well for syncing & running my LAB VM-s between my different computers.

HDD’s still surprise me, the Seagate IronWolf Pro 4TB

Intro/Random Rant

As with many people I had moved to SSD drives a long time ago for main active storage and almost forgot about HDD’s. I kept some of my old HDD-s active for dumping data that I didn’t need to access so quickly/often. But now that my 9 years old 2TB Hitachi disk is showing signs of its death approaching (rattling sounds/SMART errors) it has finally come time to replace it.

Price check HDD vs SSD

I initially thought that SSD’s have been around long enough so their prices might be reasonable.. Oh how wrong was I. If you need ample space to store all your “password lists, strange old Linux version compilation machines and test hack test boxes”, then SSD-s are still not reasonable at all. What actually surprised me was that the “slow” SATA SSD-s are priced the same as M2 NVME drives.

A 4TB SATA SSD will set you back 300-700 EUR. So if you don’t care about spending money on data that just mostly sits around, be my guest, but at least for me it seemed a waste. So I started looking at HDD prices and they were much more reasonable. Starting from as low as 75 EUR and ranging up to ~250EUR for 4TB.

After a bit of looking around on the market for disks that seem to have “long term survivability”, a reasonable price and availability, I ended up choosing the 4TB Seagate IronWolf Pro. It cost me ~120 EUR, so not bad at all.

Performance

Yes I know it’s a NAS oriented drive and I’m currently putting it inside my “regular computer”. But as my PC is “like a server” and it’s always on doing something.. Then, the fact that the disk has a long MFTB (1,2M hours) and is rated for 24/7 work is important to me.

After having installed the drive, it actually surprised me. When Seagate’s tech spec PDF states that it’s read/write speed should be 220MB/s, it out performs it. CrystalDiskMark results say that it’s actually 246MB/s read and 242MB/s write.

Also tried measuring disk performance with the build in Windows System Assessment Tool. It showed a bit slower results on the read. Did 10 runs and this was ~the average:

Besides that, I was surprised that copying stuff there from my NVME drives was “almost instant”.. I guess Windows is doing some magic in the background with newer drives. When I copied a few VM-s from my NVME to the IronWolf 10GB “was done in a few seconds”. Tested it a couple of times “over and over again” with always the same results:

And yes I know that the data wasn’t actually written that fast to the disk that fast. But it still surprised me and was interesting to see this behavior. When looking at task manager steady data transfer could be seen even after explorer had stated that it had finished copying the files:

That did not happen when I copied the same things to one of my older WD Black HDD’s that isn’t yet dying.

Conclusion

I was sad to see that the SSD prices are still high when you need higher capacity. But I was pleasantly surprised at the read/write performance of todays HDD’s. They actually come really close to cheaper SSD drives. When you add in the price per GB/TB then I completely understand why people still build machines which have a relatively small SSD for the boot drive and a HDD larger amounts of data.

AMD Ryzen 5800X seems warm when compared to others in the same series

After a few years it was time again to upgrade my computer. The “itch” started during the summer already. When I started looking at nearest release dates to see if it would be worth while to wait a bit. As Intel hadn’t made too much progress over the years over my current system, just added some cores and that’s it and AMD’s coming Zen3 was rumored to be really good. So I decided to wait and I’m so glad I did. After seeing AMD-s presentation I thought I might as well give it a try.

On the launch day I managed to grab myself a Ryzen 5800X. So that meant that after many more than a decade I was switching back to the Red Team from the Blue Team.

I paired the 5800X with the NZXT X63 AIO cooler and currently I think it was a good choice that I didn’t try and save on the cooler. Initially thought I botched the cooler installation somehow when I saw the full load temperatures go well over 87 degrees Celsius, under CineBench R23 multicore load . Even went as far as swapped the original paste that was pre-applied at the factory for Thermal Grizzly Hydronaut paste. Temp dropped only ~1-2 degrees.

Although while gaming or other lighter multicore loads the temperature is a bit more reasonable, the CineBench temperature still bothered me. So I started googling a bit about 5800X temperature. As it turns out some of the 5800X CPU-s seem to be running a bit warm, as other people had reported similar higher temperatures. Also it is interesting that the 5800X seems to be running on higher voltage out of the box when compared to other Ryzen 5000 series CPU’s.

To get the temperature to a more comfortable zone I finally turned to under volting. By lowering the voltage by 0,05 the temperature while gaming is ~63 degrees and running Cinebench R23 it is ~81 degrees Celsius. Still a bit high, but much better. Lowering the voltage didn’t luckily cost me any performance. At least when looking at the boost clock and benchmark numbers. I haven’t actually had time to find the lowest stable under volt yet, maybe I could get it lower and cooler.

Long story short. Although CineBench R23 load shouldn’t be the normal load for most of the people, but as I’m getting 67 degrees Celsius under normal gaming load and ~81 degrees Celsius on Cinebench with the NZXT X63. It makes me wonder if only a high end cooler is a must for these “warmer” CPU-s. Also I’m wondering what would the temperatures be with a regular cheaper (~30$ or so) AM4 air cooler. Even if the temperatures are higher than I’m used to, even compared to my old overclocked I7. I’m still pleased with my purchase and the performance boost was phenomenal.

What was interesting is, that even after under volting the CPU I was able to set boost clocks 200Mhz higher and its stable. So single core boost is over 5019Mhz, lighter multi core load 6 cores or so goes over 4900Mhz, yet all core full load is 4541 Mhz.

On the infinity fabric side, getting it to run on 1900Mhz didn’t require much messing about, just set it at that and that’s it. So 3800Mhz memory is easy to get working with 1:1 settings.