Gadgetory

welcome back to harbor unbox for another benchmark video this time we're looking into the possible performance impacts caused when addressing the two now-famous security floors mount down and Spector in my last video I benchmark the windows update intended to address the amount down of under ability and I did this from a desktop users perspective or more precisely a content creator slash damer what I found was virtually no impact on gaming performance so good news for gamers and there was also no impact for content creators either what I did find though were a few troubling results for nvme storage devices and this mostly impacted 4k read performance since then a few major tech outlets have also published similar findings so those issues weren't just limited to my own testing overall though for a typical PC user or perhaps more importantly at least on this channel a gamer addressing the map down vulnerability wasn't a big deal there was however a second vulnerability that needed to be addressed but that one would require a firmware update or more precisely a motherboard BIOS update you've known that heard of Spectre by now and this one is a result of a fundamental flaw in the CPUs design as I understand it and keep in mind I'm far from a security expert we mostly test PC games and relating hardware at harbor and box but we're doing our best to cover this topic because it could affect our audience and benchmarking can help us understand what the performance impacts might be so as I understand it because Spectre is the result of a fundamental cpu design flaw it can't be fixed at least not entirely and not without replacing the CPU the firmware update or the BIOS update mitigates the problem but doesn't completely address the vulnerability this is also still primarily an Intel CPU floor it is reported by AMD that one of the two respective variants doesn't impact them at all while the other one does but is easily resolved by a software update that shouldn't impact performance in any meaningful way variant 3 which is now well known as meltdown doesn't impact AMD at all and this was covered in my previous video I'm yet to properly test any a DCP you so you'll have to sit tight for a moment while we focus on a few Intel CPUs since posting the meltdown benchmark video we now have access to BIOS updates and they do deliver the microcode updates necessary to mitigate the spectra of under ability and so far this has happened on Intel's latest said 370 platform this update changes the behavior of Intel's branch prediction to be less aggressive this will likely mean less effective branch prediction and that means reduced IPC as the execution pipelines wait for memory access more often of course we will get to the benchmarks in a moment but before we do I just have a few more notes I would like to cover right now at the time of shooting this video the only motherboard manufacturer to release an update is a soos and so far they've only addressed their Zed 370 series of motherboards unfortunately I don't actually own an a su said 370 motherboard believe it or not I have models from every other brand but a soos so that meant 48 hours ago I raced down to my local PC store and purchase the asou stuff Zen 370 plus gaming for this test I then went back and benchmarked the core i3 8100 without the windows update and then again with the windows update and then a third time with the windows update plus the latest BIOS update which included the micro code update I've included some of the updated core i7 87 arcade benchmarks as well once you've installed the windows update you can install a PowerShell module called speculation control and this will allow you to check if the update has worked by typing get speculation control settings with just the windows update and still which only addresses the meltdown vulnerability this is what you'll see all three requirements for meltdown aka rogue data cache load are greenlit and set to true for the branch target injection aka spectra vulnerability only OS support is present but not yet enabled as we still require a micro code update once the boss has been updated with the required version this is what you'll see ok so that's all the updates we have so far it's time to run some tests and see if there's any performance penalties let's start with the core i3 8100 results oh and please note all results are based on an average of it these three runs first up we have the core i3 Cinebench r15 results and very little has changed here from the pre update configuration we see less than a two percent reduction in the multi thread score and 1% for the single thread score so that's pretty well within the margin of error now this is a little more interesting the windows patch plus the BIOS update was consistently three percent slower than the previous test configurations please note that lower is better for this test as we're measuring the time it takes to complete a render so the BIOS update cost us nine seconds but as I said overall a very minor reduction in performance moving on we have Excel and once again the Excel work load goes unchanged we see the same six second completion time so really nothing to report here moving on we find the same with the blender render test all configurations took 58 seconds to complete the test we also find no real performance differences when testing with veracrypt the aes encryption and decryption results are all much the same then we have some 7-zip results and here we see no noticeable decline in performance with the windows and bars update applied the Geekbench for compute test also shows similar performance with the updates installed Geekbench also has a slew of CPU benchmarks and each category gets its own score as you can see the crypto and memory scores are all much the same we see a 3% reduction for the integer and floating-point scores with just a 2% reduction for the multi-core score okay well it looks like the applications and synthetic tests didn't really give us much let's move on to the games to see if they have any scary numbers so here we are our first game benchmark result and while things don't look that dramatic we do see about a 4% reduction in performance in the core i3 8100 was consistently slower in this test by a three to four percent margin also for those wondering the geforce gtx 1080i an insanely overpowered GPU for the core i3 cpu is used to try and remove the GPU as a potential bottleneck and this gives us a better idea of the impact the cpu has on gaming performance for a slightly more realistic test though we have boosted the quality preset from medium teletraan and again we see the same 4% reduction in performance with the windows and bass updates install moving on we have ashes of the singularity escalation and again we see a three to four percent reduction in performance with these updates installed not a massive decline but a reduction in frame rate all the same and again the update configuration did consistently come in lower even with the crazy preset enabled we still see the same three to four percent drop in frame rate assassin's creed origins isn't GPU bound heavy 1% low results with the 87 okay for example reach around 90 FPS so all these results certainly look GPU limited I can assure you they aren't so that being the case the BIOS update does deliver similar performance to that of the pre updated configuration certainly within the margin of error even with the ultra high quality settings enable was still at GPU bound despite all three configurations delivering the exact same results I had honestly thought we might see some larger margins in Rainbow six siege but that wasn't the case in fact here we see almost no difference at all with the medium quality preset shifting to the ultra quality settings doesn't really change anything and again we see that the BIOS update has no negative impact on performance for this title the results in f1 2017 also go largely unchanged though this time we do see a dip in the 1% lower results that said it's nothing more than a 3% reduction the same 3% drop for the 1% minimum can be seen with the ultra high quality settings though again the average frame rates are all much the same the last game that I've tested with the core i3 8100 is total Warhammer 2 and here we see a 4% drop in frame rate for the 1% lower result while the average remains much the same than with the ultra quality settings we see almost no change in performance as all three configurations provide similar results but what about the core i7 87 ok how does it get on with the BIOS update well all the application and productivity tests were all much the same so I'm not going to throw a heap o graphs at you showing absolutely nothing all over again so we'll just move on and have a look at a couple of games here we see that the battle for one performance using the medium quality preset allows for similar 1% low results while the average dips by a 3% margin with the ultra quality preset we see virtually the same results the Windows Update actually boosted the average frame rate by 2% which is of course still within the margin of error so I wouldn't read too much into that result using the higher quality preset we see virtually no change with ashes of the singularity this is also true with the crazy quality preset enabled then finally we have assassin's creed origins and again the margins are very slim here with the bios update enabled we saw the 1% low figure reduced by 2% well the average frame rate remained much the same then with the ultra high quality preset name we'll be saying no difference in performance at all ok moving on what about storage performance this was the only area where we saw any real impact last time let's start with the Samsung SSD 950 pro nvme drive on the core i7 87 ok system here we see a 5 percent reduction in throughput for the sequential write test an 8 percent reduction for the 4k 64 thread write test and a 20 percent reduction for the 4k write test this is also seen when comparing the pre update configuration to the Windows Update with BIOS configuration read performance though that's all pretty much similar across the board it's just the writes that look to have taken a hit moving on we find some even more extreme results when testing with crystal disk mark the sequential read and write performance all looks ok then we get to the 512 K write test and well what's gone wrong here a 41 percent reduction in performance can be seen and this wasn't a once-off deal I ran this test a dozen or so times after multiple resets to try and work out if it was just some kind of glitch but every time this is the result I got interestingly though the 512 K read performance isn't nearly as heavily impacted just 8% reduction here though that's still certainly very noteworthy then when we move to the 4k queue depth of 32 we did find a 10% drop for both the read and write results the single 4k wrap performance is also reduced by 19% with the BIOS update while the read throughput goes unchanged I then decided to do some testing with a toe disk benchmark and hautboy what has gone wrong here both the sequential read and write tests took a massive hit and again throughput was reduced by as much as 40% with the BIOS update enabled so if this has happened to nvme SSDs what does this mean for your more run-of-the-mill SATA SSDs then well here's the Samsung SSD 850 Evo 2 terabyte with the core i3 8100 and here we do see some declines again the sequential read and writes for the crystal mark test they do look ok in fact they look quite good the 512 kilobytes read and write performance is also very good this time the 4k queued at 432 results though they do see a small decline of around 3% so nothing too alarming that said though we do still see some big hits for the 4k transfers which is interesting writes dropped by 27% and reads by 19% then again we have some auto gist benchmark results and again we do see some performance drop-offs here oddly though this time the read results don't actually seem to suffer but the write throughput certainly does here we see up to a 17% reduction in write performance with the BIOS update applied and that's certainly a note where they drop off for a SATA device ok so we've got a bit to talk about here I want to start by saying please take those storage results with a grain of salt at least for now until they can be confirmed as at least one other reliable source I wouldn't go too crazy over the potential performance impact here I'm not sure if this is something we're gonna see others reporting or if this is just an issue I ran into with my particular hardware configuration I have though come across a test done by guru 3d which does show up to a 33% reduction in performance with the Samsung SSD 960 pro 2 terabyte nvme drive and this was for the 4 kilobyte Q depth of 32 test so that's somewhat in line with what I saw they didn't test with outer unfortunately though they did also test with a sous motherboard so perhaps it's an issue with your sous bias update and they'll battle address that or possibly this is just what the band-aid looks like as always though time will tell I should also note that I was able to revert the bass back to the previous version and doing so did restore the original SSD performance horrible storage performance aside we constantly saw less than a 5% reduction in gaming performance you're probably looking at more around a 3 to 4 percent drop and that's when you are CPU limited less of course when you are GPU limited SSD performance that doesn't really impact framerate so we've seen this in the past when comparing slow hard drives to Oh snappy SSDs is really nothing again there we're a drop in storage performance can hurt is with the game load times and this is something we'll dive into in the future should our storage figures prove accurate of course then we have the more commonly used applications like Excel and 7-zip for example which seem unaffected using either the core i3 8100 or Core i7 8700 K processors and we found a little too note impact for those doing content creation type work as well but what about older CPUs you know your wells and bridges and older 45 nanometer processors well as it stands right now we can't actually test those it's possible to test the impact of the Windows 10 meltdown patch but it's not yet possible to test the impact of the big one Spectre for this we need BIOS updates for older motherboards supporting those CPUs when or if we get them at all is a bit of an unknown at this stage we haven't found a motherboard maker that really wants to give us the inside scoop one thing is for sure though this is a mess of epic proportions and it will still require further OS and past updates this is really just the start to close these security flaws it's also just the start of Intel Spain already there are three class action complaints that have been filed against Intel over these CPUs security flaws their potential impact and also Intel's response time to address them Intel can no doubt expect to face further lawsuits as well and not only are these security risks a serious issue but the impact addressing them has on businesses it's also very serious as another example before I switched my attention to focusing on this BIOS update I was waist-deep in benchmarking fortnight I've got another one of those 40-plus GPUs showdowns in the works anyway I bring this up because on the last day of testing I was running into a few login problems which was quite unusual for fortnight the issue was I often had to wait 15 or even 20 minutes before I could get into a game and that is just a load the game which point I would run my tests shut it down install a new GPU fire it up and Sharaf had to wait another 15 or 20 minutes to get back into the game so I quickly gave up on that and thought I'd give it a shot later when they addressed whatever the problem was and I'm here under testing something else at the time though I do remember thinking to myself this is just a day or so after the meltdown exploit started to get addressed wouldn't it be funny if that was the reason for the busy servers well turns out not that funny because that was the exact reason or at least this is what epic games are saying for the login issues soon after the login problems began that the fortnight team made a post on the epic games forums as they wanted to bring the issue to the community's attention explain what was going on turns out the updates required to mitigate the meltdown vulnerability we're slowing the server's down by causing massive spikes in CPU utilization and at that point they had just patched a single service and that one cost CP utilization to triple they also know that there's more painting to come but they'll do their best to manage the situation no doubt we're gonna keep hearing stories like this as more companies update their servers over the coming days and weeks anyway I'm gonna wrap this up as it has no doubt been a very long video so thank you for watching the whole way through if you did there's certainly more testing that still needs to be done and are you gonna see what others find when it comes to the storage performance that's obviously an area to focus on for now though your frame rates look pretty safe and applications that aren't storage sensitive will also perform much the same as I've said much more testing needs to be done but I've been chipping away at this for two days straight now with almost no rest so yeah it's time to get some rest and see what others dig up if you liked the video please leave a like and if you'd like to support our work directly then please check out our patreon account the link is in the video description I'm your host Steve see you next time