Gadgetory

welcome back to Haram box today we're gonna try and push the radio rx 5700 XT to its limits by strapping on EKG vector waterblock for thenar V based graphics cards ek were kind enough to quickly send one of these over to me so I can play around with it I've had it for a little while now but obviously we've been quite busy with a lot of those n2 content we've been doing but anyway I've got it now so a big THANK YOU to ek for sending that over if you missed it in my initial 5700 series review I noted that increasing the fan speed off the 5700 XT to an ear bleeding 100% I was able to boost the GPU up from around 1800 megahertz so that's basically what it runs that stock and I was able to push it up to 2050 megahertz so that was a 14% overclock then after all the day one reviews went live Eagles led published interesting pace based on some findings by a Radeon Tweaker group user helm I think if I got that wrong sorry anyway the user created some power play table mods for the 5700 series that made it possible to bypass AMD's overclocking limits imposed at the BIOS level these simple yet genius Windows registry modifications allowed users to circumvent AMD's limits and increase the GPU frequency as high as two thousand three hundred megahertz while also increasing the power limit of course the GPU silicon has to be good enough to run a 2.3 gigahertz and sadly neither of my retail cards were able to run and even 2.2 gigahertz rather 2.1 gigahertz was the limit which was a shame but it's still around at 18 percent overclock so I'm expecting a nice performance uplift before we get into the benchmarks though just a quick refresher on what we found so far after an hour-long f1 2019 loop the 5700 XT in its stock configuration with the reference blower cooler ran it around 1780 megahertz once the card reached its peak operating temperature of 84 degrees in a 21 degree ambient room under the same conditions with the e K vector block installed and attached to a 240 millimeter radiator with two super-quiet 120 millimeter fans the 5700 XT peaked at just 45 degrees so that's a mass 39 degree drop in temperature meanwhile the GDD r6 memory temperature dropped by 20 degrees and the vrm temperature dropped by as much as 30 degrees having tried all the registry modifications I found that 2.1 gigahertz was the limit of Maya Seuss reference card again I did try all of the registry options and just pushing past two thousand one hundred and ten megahertz just cause games to crash and I found the same limit with my power color reference 5700 X T as well so I guess I haven't done that well with this silicon lottery here but as I noted a moment ago that's still an 18% overclocked and given the 5700 XT was only 9 percent slower on average than the RT X 2080 this overclock should push the radio and GPU ahead of course you can also overclock the 2080 but that's not really the point we're trying to achieve 2080 light performance at a significant discount so let's fire up the car I $9.99 here okay GPU test rig and see how the liquid-cooled 5700 xt performs first up we have f1 2019 and here we see a modest 9% bump in performance for the average framerate and a slightly less impressive 6% boost for the 1% lower result this did allow the 5700 XT at 2.1 gigahertz to match the GTX 1080 TI though it was still 5% slower than the RT x 2080 not exactly a big margin but I was hoping the overclock would yield better results moving on to battlefield 5 and here we see a far less impressive 5 percent boost in performance for the average frame rate though the 1% low performance was improved by a 7% margin having said that the 1% low frame rate is still a bit weak at least relative to the RT X 2080 but overall an impressive result given that we're looking at our TX 2080 super light performance in this title here we see the most impressive gain yet a 10% boost in performance at least for the average frame rate when testing with Rainbow six siege well this does push the 5700 XT ahead of the stock 1080 Ti it's still slower than the RT X 2070 super and much slower than the RT X 2080 super next up we have Metro Exodus and here we see a very mild 7 percent improvement in performance going from 72 fps to 77 fps and that wasn't enough to catch even the 2070 super in this title another lackluster 7% gain can be seen when testing with Resident Evil 2 very this time that was enough to at least match the r-tx 2070 super shadow of the Tomb Raider disappointed with only a 6 percent improvement in performance but at least that was enough to match the gtx 980ti and r-tx 2070 super another 6 percent gain can be seen when testing with fortnight but here we're only talking in extra 6 FPS so in my opinion it's probably not worth the trouble then moving on to the division 2 we see another 7% performance uplift for the 2.1 gigahertz overclock and that was enough to match the r-tx 2070 super in this title dirt rally to seize just a 5% boost going from 102 FPS to just 107 fps so a very disappointing result here the Radeon rx 5700 XT already killed it in Forza horizon 4 and here we see just a week three percent improvement from the overclock but still that was enough to take out top spot and I suspect we're probably becoming CPU limited at this point interestingly Far Cry new dawn provides a really nice 12 percent performance boost and this put the 5700 XT right on the heels of the RT X 2080 and I really wish we saw more results like this second last game they're going to look at is World War Z and here we see a 6 percent improvement in performance from the overclock and that's just half of what we saw in Far Cry and we finish up on a very disappointing 4% performance boost in Assassin's Creed Odyssey so let's move on to check out some power consumption figures for what was a 7 percent boost on average you're looking at almost a 40% increase in power consumption as the GPU power draw increased from just 186 watts right up to a Vega like 258 watts then looking at total system consumption we see a 25% increase in power draw for the overclocked 5700 XT making it slightly more power hungry than Vegas 64 and around 30 watts more than the RT x 20 atti so it seems the 2.1 gigahertz overclocked destroys our DNA's efficiency well that was super disappointing my I guess AMD can do it as well but seriously the Radeon r x57 and XT it is really a great product and I still have very high hopes for the upcoming AIB models unfortunately though your chances of pushing one of these things to RTX 20/80 super or even 2080 levels of performance are pretty slim even at 2.2 gigahertz I feel it's a bit of a stretch sure there are some games that well games that are more favorable to the red team let's say in those kind of games the 5700 XT will pull off the upset will will merge ahead of those 2080 super models but overall yeah it's it's not going to match them for for the vast majority of games spending a hundred and forty dollars u.s. on the e K vector navi block that does take the price off the 5700 XT up to five hundred and forty dollars u.s. and for that money you can buy an RT X 2070 super which overall will deliver a similar level of performance obviously increasing the cost of the 5700 XT by 35% for what we saw at best was a 12% performance increase at least for my particular retail card that could only do 2.1 gigahertz that's not really a good investment but would I rather in RTX 2070 superpartner card or a liquid-cooled Radeon rx 5700 XT it's a bit of a tough one I have to admit but I think I would rather have the liquid-cooled graphics card of course for that option to make any kind of sense you already need to have or at least be interested in getting a custom liquid cooling setup and there is a third option and that would be just to wait for the board partner cards to be released the air-cooled 5700 XTS and frankly that option makes the most sense now we should have those some point next month perhaps later next month so yeah at this point I just wait to see how that plays out with a decent air cooler my 5700 XTS will still run at 2.1 gigahertz while making very little noise and I expect the temperatures to be in the 70 degree range the main reason I made this video was to see what all the fuss was about and I also wanted to check out that really nice-looking vector block from ek for the Navi GPUs and I certainly wasn't disappointed there ek make excellent products just quickly as a side note I've seen heaps of youtubers raving about how amazing the power play table mods are jumping up and down about 15 even 20% overclocking Headroom and yeah it all sounds very exciting I get that I was very excited putting this thing together I couldn't wait to see what it could do but yeah unfortunately not as exciting as a 15 or 20% frequency increase seems to be because sadly on average I was only able to extract a very mild 7% boost so yeah 7% increase in frame rates on average and I just feel with so many people raving about the overclocking Headroom being as high as 20% but not actually doing any testing to see what that 20% boost in frequency gives you yeah I feel like the whole things just become massively overhyped and well that tends to be what happens with new AMD products these days it seems don't get me wrong the 5700 XT is an incredibly good value product and if we get quality ARB models for under $450 us they will be the graphics cards to get south of $700 us but they are overclocking monsters and if you care about value spending more money on upgraded cooling really isn't a wise choice just lastly in a lot of games tested I only saw gains as high as 6% from the overclock the 18 percent frequency overclocked and even sometimes the gains were just 5 for an even 3% so why does the 5700 XT seemed to scale so poorly when overclocked well at the moment I've really only got a theory to go on but that theory is it that the card is memory limited and you can't really overclock the GDD r6 memory stock it comes clocked at 875 megahertz and going beyond 900 megahertz isn't really possible for example I saw artifacting at just 920 megahertz something else I noticed is that when increasing the memory frequency by 3% to 900 megahertz that would often drop 1% low performance and at 920 mega Hertz where I was seeing graphical glitches on occasion the 1% low performance and even the average frame rate dropped further but of course memory errors will cause a performance regression what I really want to do is underclock the GDD r6 memory and I want to do that to look at two things firstly just how memory starved is the 5700 XT does dropping the frequency down to say 850 megahertz have a noticeable impact on performance I suspect it will also does performance tank at 800 megahertz again I suspect it will but sadly I can't test this as the card is locked at 875 megahertz and right now I haven't found a way to circumvent this I'd also like to see if the 1% low performance actually improves at slightly lower clock speeds for the memory and some titles narvi suffers from pretty weak 1% low performance and I wonder if that's a result of the memory being clocked so close to its limits out of the box of course this is just a theory don't go running with it and assuming that that is the case very little evidence to go on at this point so if anyone knows how to bypass the 875 megahertz frequency floor for the 5700 XT then please let me know in the comment section below and I'll happily go back and look at that anyway that is going to do it for this one if you enjoyed the video feel free to the like button that's always much appreciated and you can subscribe for more content if you haven't done so already also consider supporting us on patreon you will gain access to our exclusive discord chat and monthly live streams with Tim and myself anyway as always thank you for watching I'm your host Steve and I'll see you again next time