Reprinted with permission as originally written by Mark Vogel of F.A.S.T Forensics.
Kind of a book here but there’s a LOT going on in the processor market right now between Intel & AMD, so there’s a ton of information and considerations between the two now.
I have done a couple Ryzen builds since the release of the Ryzen 7 CPUs earlier this year to test out. The chipsets for this platform seem to have finally matured as the most recent BIOS updates resolved a long ongoing issue with properly clocking faster RAM models to their factory spec for speeds of 2666MHz and higher. This is actually important for Ryzen as the CPU does benefit more from faster memory than most Intel platforms, so running DDR4 3200 memory does help these builds.
As far as CPU shootouts go, for multithreading applications like what’s used in the forensic world for the most part, the 8-core Ryzen CPUs do hold their own against the $1000 Intel 8-core CPUs like the 5960X and 6900K pretty well. I built on a 1700X and manually overclocked it on all cores to 3.95GHz to make it perform more like the 1800X while saving $100, and it benchmarked (synthetically) actually higher than a 5960X. That’s quite a feat considering the CPU at the time was only $399, and is now even cheaper. It still was not quite as powerful as an i7-6900K 8-core, but it was close, and considering that’s also a $1000 CPU its still a huge bang for the buck. At the time also, my DDR4 2800 RAM I was using wouldn’t clock to 2800MHz due to early problems with the platform. I could only reach 2400 on all 64GB or RAM. Like I eluded to earlier, the most recent BIOS release from July finally resolved this and now we’re hearing about people hitting 3200 all around and my memory is hitting 2800 no problem.
Now, Ryzen has proven to put Intel on its heels a bit as we expected and Intel’s new Skylake-X CPUs on X299 are cheaper. The 8-core Intel i7-7820X is a very powerful CPU, even more so than the previous 6900K, and the 6900K was still more powerful than my overclocked 1700X from AMD. So, Intel is still king when we’re talking instructions per clock, and the Intel chips have MUCH more room for overclocking and now the Ryzen 1800X is only about $140 cheaper than the new Intel 7820X. This cost difference much closer justifies the extra $$ for the Intel, BUT the Intel motherboards are more expensive so there’s still a bit of a price gap (figure maybe $300 overall). That said, you could grab a Ryzen 1700X and put a mild overclock on it and save another $130 on your Ryzen build if you’re open to overclocking which is very easy and straightforward on this platform. You’d just want to grab a good quality CPU cooler (I prefer liquid).
This is a great resource to compare real-world stock CPUs to each other, and even overclocked CPUs (though the numbers for OC’d CPUs are typically a touch low) and single core performance. There’s different charts for each. Start at the “high end CPUs” to get a feel for real world performance difference. https://www.cpubenchmark.net/
. You’ll see a stock 1700X scores 14,677 for instance. My overclocked 1700X scored 16,216 @ 3.95GHz on all cores which is higher than the i7-5960X but can’t reach a stock i7-6900K. In the real world though, you’re not talking about much difference in application performance.
In regards to motherboards and chipsets, this is where things get even more interesting. My concern with trying to make the Ryzen build have close to identical spec as my Intel X99 builds was the big difference in PCIe bandwidth and lanes between the two setups. On X99 with a 6900K we have 40 PCIe lanes from the CPU. The Ryzen 7 CPUs with X370 chipset only provides 20 PCIe lanes from the CPU. If you get creative with your build, or don’t require all those lanes for a fully loaded build, it can suffice just fine. Depends on your build. I could dive in to this complex subject with much more detail but it makes this Email even longer! Long story short, the Intel platform provides more expandability to fit more components but that only matters if you’re loading out the build.
In the end, this machine was used daily in an agency for about 5-6 weeks and they noticed no discernible difference in performance between it and the FAST Ultra machine they also bought from me which is based on Intel’s X99 chipset and an 8-core 6900K. We never did actually process an identical case on both to see what the real world difference was however. The thing is, as long as the CPUs are close, it typically comes down to storage performance which makes a HUGE difference, especially when using tools like FTK, and I set up the disk I/O on both builds identically with multiple NVMe SSDs, same RAID array on the same RAID card, and what not.
Getting back to comparing the new Intel X299 platform with the Intel 7820X ($599) vs the Ryzen 1800X ($460): Intel has scaled back on PCIe lanes with the new 8-core 7820X (28 lanes) vs the last gen 6900K (40 lanes). This is presumably to get you to pony up for the new 10-core $1000 7900X which provides 44 PCIe lanes for machines (like ours) that can take advantage of those lanes. However, the Intel X299 chipset provides a bunch of PCIe lanes to the motherboard which are typically used to juice 3 x m.2 NVMe connectors. So we can still load out a system with tons of storage I/O, a dedicated RAID card, a single GPU and a 10Gbps adapter. I’ve done a couple builds thus far on this board (http://www.asrock.com/mb/Intel/X299%20Taichi/index.asp#Specification)
and this will be my new board when I cut my standard builds over to the new platform.
So, a current Ryzen build is a great bang for the buck if built properly and can save a few bucks for a budget build over Intel, but the gap isn’t as huge as it was a few months ago. Figure you can build a more powerful Intel-based build (8-core CPU) for about $300 more.
BUT…..AMD just put their new Threadripper platform with their X399 chipset on pre-sale. This is kind of Ryzen 2.0. It uses the same CPU core so its not more powerful than Ryzen 7/5/3 on a per-core basis, but these are AMD’s answer to Intel’s X299 platform and higher core-count CPUs. They’re releasing the Threadripper 1920X 12-core (24 thread) at $799 and the 1950X 16-core (32 thread!) for $999. On top of this, the new CPU/chipset combo provides a TON of PCIe bandwidth with 64 lanes! Now this platform, in my and many others’ opinions, is finally the Intel killer in the high end workstation and professional market (not in gaming for a few reasons). Now consider that you can have a 16-core AMD CPU for $1000 vs Intel’s 10-core processor for the same price. This is a HUGE difference for applications that can take advantage of those cores/threads. On top of this, all those PCIe lanes that were limited on Ryzen 7/5/3 are no longer limited and in fact beat Intel. Yes, the 12-core is still more expensive ($799) than the 8-core Intel 7820X ($599) but for good reason and AMD will be releasing an 8-core Threadripper CPU for this platform which is reported to land (in another month I believe) at $549. To be completely honest with you, between the 8-core AMD Threadripper offering on this new platform and the Intel 8-core offering at just $50 more for the CPU (and the motherboard could actually be a few bucks cheaper), I’d go Intel. However, if you wanted to jump to a 12-core you’re getting a huge bang for your buck increase.
Note that Threadripper won’t even ship until the 10th of this month. I’ve got a 1920X on pre-order and will be testing this platform literally right next to an Intel 10-core 7900X build to compare. There will no doubt be bugs and challenges for at least the first 30 days of this brand new AMD platform so I wouldn’t recommend building on it quite yet if you wanted to go that route.
In regards to password cracking, that’s all about GPUs. This is another conversation really and it won’t matter which AMD/Intel CPU platform you go with. I documented the build process through a password cracking machine on my forum (https://www.getfastforensics.com/forum/)
if you want to check that out. Even in this market though, AMD is stirring things up as their new Vega RX series GPUs will be releasing soon and are providing a huge amount of processing units (over 4000 vs nvidia’s top end offering of 3500) so this may flip the password cracking market back to AMD. The only current issue with using AMD GPUs vs Intel is you can’t use their GPUs to crack specific types of files IF using passware. I’m not sure about hashcat as that tool may work on all file types regardless of GPU make/model/drive