Gadgetory

everyone talks big game about how they don't care about power consumption when it comes to Vega that's all we heard about in the reviews well here's Vega modified for people who don't care about power consumption in the live stream testing this card we were pushing something like 406 Watts through it it is a Vega 56 card for a reminder that was cap out at about 300 to 308 watts with the bios a lock and 50% offset from AMD so we've fixed it we can now push as much power as we want enough to kill the card really and we'll see how far it goes with the liquid cooled mod 360 millimeter thermal take flow radiator on it and a whole bunch of fans before that we partnered with LastPass to bring you this video LastPass is a password manager that helps generate random passwords for each account insuring unique passwords for every log in as fast strengthens account security so that you're no longer using the same password for multiple sites learn more at the link below for a quick recap on the mod what we have here is the flow cooler it's 360 millimeters we have to Corsair Maglev fans and one Noctua and f12 fan because those are the three best fans that I had easily available and the FETs most of them are cooled by well a direct airflow fan above them along with some heatsink so just aluminum heatsink sticking off the fins are from Arctic Cooling's kits that we've basically adapted to this and then for the actual mounting we bought some screws at a local hardware store and drilled holes in a mounting plate for our Vega frontier edition hybrid mod and then repurposed all of that here so it's working out pretty well now the thing is this is it's it's not really direct comparison anymore because we've also installed power play tables which is a registry hack that builds oyk provided to us with the help of I believe helm from the overclocking forums he was on so that allows us to do up to 242 percent offset which is enough to kill the card if the card actually pulls that much it's kind of hard to get it to pull more than 31 apps but actually we did 33 lets go anyway so it's not direct comparison anymore the amount of fans we have on this thing plus the overhead and cooling the FETs means that there's zero point to testing noise levels you could get it to be fairly quiet and actually cool well but then you're not gonna be doing the crazy overclocking stuff so it's a trade-off this isn't for people who care about noise or power because in the comment sections of most reviews you see people say I don't care about noise or power I just want something that performance well this is your unrestrained version of vagin 56 it should outperform Vegas 64 stock for sure we'll see how it does a overclocked in a future content piece but we'll look at all that today let's start with the power and thermals is the most important and then move into gaming performance some 3d mark stuff things like that given all the uncertainty surrounding the initial driver launch and limited voltage and broken clock reporting functionality we first needed to check whether the power play tables were actually doing anything fortunately the clock reporting was fixed in driver 17 8 1 onward something we showed in our last livestream with hybrid 56 so we now need to isolate the power target here's a look at the frequency versus the power target represented by PCIe power draw through the cables I was measured by a current clamp power consumption starts off in the 296 to 300 watt area when we have a 50% offset which was the max permitted by Andy originally that's a bit below what we saw on the reference Vega 56 card which was 308 watts so we've reduced maybe power leakage or something by about 4% giving us some more Headroom but not really enough to do anything with regardless that's the starting point the clock bounces around sporadically at this point jumping between 1474 and 1702 megahertz because the cards power consumption is limited to about 300 watts we can't achieve stability at higher clocks without pushing power we first go to 60% power off set permitted by the power play tables which gets us another 16 watts to work with the clock still aren't stable so we moved to 70% power around the 240 second mark on the chart that boosts us 33 watts over the original 50% offset and has a clear impact on helping that somewhat stabilize the frequency at 80% power we start pushing 350 watts through the PCIe cables which largely stabilizes the 1702 megahertz target the next attempt is 17 22 which proves unstable and requires a jump to 90% power we were not able to anywhere close to this with the original reference card this puts us at 362 365 watts 17:32 megahertz is roughly stable at the same power draw and finally for 1740 2 megahertz we boost a 95% power to fully stabilized and draw 370 to 380 watts down the PCIe cables plus whatever smaller amount is coming through the PCIe slot which largely powers the memory the result is 70 to 80 watts more than the reference card with the clock that stabilizes at 1742 megahertz rather than going between 1474 and 17:02 with our original target and we started crashing at 1762 megahertz so definitely up against stability there even in the live stream when we pushed 406 watts more than we did here because we were just trying to get the highest power and frequency as we could at the most stable setting we could even then we couldn't get really beyond 17 whatever 1742 17 22 so 17 42 is the final number for fire strike for the games later on we have to drop down to 17 32 because 1740 Caesar is not stable as for HB m 2 it stays at 980 megahertz because that was the most stable there was some test like fire strike we were able to complete with 990 megahertz and here's the temperature response to all of that core temperature paintin's 28 C quickly maxing out at 31 C this is not a delta T over ambient reading it's just really low ambient is about 24 Celsius here but we know now that our Vega 56 under reports at the temperature by at least a few degrees so it's either just inaccurate like a lot of thermal sensors are but more than normally or it's confused about the temperature we can't be sure by how much unfortunately either way we're about 40 to 45 C below the reported temperature on the air-cooled version the MOSFET temperatures max out at 46 C for the middle right fat 32 C for the top fat both of which were cooled by direct airflow from a fan for perspective the reference card had fed temperatures of about 63 C for the right FET and 73 C for the top FET we've managed to reduce MOSFET temperatures by blasting them with air despite these 70 to 80 watts of extra power to the core so they're actually cooler than previously but dealing with a lot more power getting it's a comparative thermals with another test we see the RX Vega 56 reference card operates close to the 75 Celsius target with our hybrid moderating a 25 Celsius GP diode temperature overclocked into 17 42 megahertz core and 9 80 megahertz HBM 2 and using a 105 percent power target we pushed 30 Celsius T V diode temperature but note here that ambient temperature again is 23 to 24 so we do still have that same inaccuracy as previously mentioned under idle workloads we're just to kind of illustrate this sitting between 2 and 4 Celsius below ambient which obviously isn't possible with this kind of cooling setup you need something chilled to actually achieve that so the temperature reporting is definitely incorrect looking at power consumption at the wall our RX Vega 56 hybrid system pushes for 65 watts when overclocked 1742 my card score and nine 80 megahertz HP m2 that's an increase of 165 watts over the reference vega card this chart will be quick this is what we use when doing our initial undervolt in testing that sporadic orange line is from fighting with the software back then and shows our range of PCIe consumption being roughly 180 to 355 watts we were able to wrangle the hybrid down to 355 in this particular task by dropping to a 95% power target which proved stable for a fire strike though ultimately increased it a bit for gaming keep in mind that power consumption varies based on scene and benchmark use so you'll see different results based on which tests were actively showing for some quick power numbers from fire strike as measured at the wall and again we're switching back to wall draw here we see total system power at 447 watts for the hybrid ocv 56 nearly tie-in it with the crossfire rx 580 and 480 cards the stock Vega 64 system draws 372 watts so we've got a 75 watt increase here and the stock v56 system drops 303 watts a gtx 980ti system draws 347 watts or 100 watts at less than our b56 hybrid OC and about 40 more than the reference v56 card there'll be more of these power charts in the article along with the fire strike benchmark scores as for gaming results we'll start with the more exciting ones then work our way toward the disappointing outcomes this also means starting with games that tend to be favored toward AMD Sniper Elite 4 has a few relevant the stock rx Vega 56 card operated an average FPS of 53 with a 9% overclock and 9 50 megahertz HBM to clock and in us at 58 FPS average that's an increase of nine and a half percent and is largely from the boosted power limit and HBM - as we learned in previous content the core overclock and is a lot less relevant than HP m2 and power offsets a hybrid card clocked to 17 32 megahertz and 980 for stability and at 66 FPS average with its 100% power target ensure we're pushing 30 amps down the PCIe cables for 80 Watts more power consumption but we're not concerned about power consumption today because that's what people keep saying the gain is 13.8% over the overclock sat Vega 56 card or 25% over the stock making 56 card compared to the stock Vega 64 card where 14.8% ahead in performance note also that vega 64 is basically tied technically behind the overclocked reference at vega 56 which is sort of further reinstating vega 64 spore value just get vega 56 and overclock it is the lesson here even with a and these BIOS locks and other limitations we're doing better than the stock 64 we have an overclock to Vega 64 yet but that'll come sometime after pax west looking to Nvidia cards the v56 hybrid OC is now a bit ahead of the GT X 1080 reference card without an overclock we haven't yet gotten around to overclock in the 1080 for this new round of test so keep that in mind but in the very least we're ahead of the reference card which is pretty damn good for a Vega 56 granted we had the bypass restrictions put in place by AMD but it worked out after the singularity testing lands our stock rx Vega 56 blower card at 67 FPS average with the overclocked reference card as howdy to FPS a gain of 7.5% Vega 64 runs 8.5% faster and our Vega 56 hybrid runs 7% faster than that 83 FPS average the hybrid model is now 16% ahead of the overclocked Vega at 56 reference card and 24 percent of the full stock card this lands the overclocked hybrid 56 ahead of the reference again stock clocked GTX 1084 honor shows surprisingly positive scaling for this overclocked at 4k we moved from a 38 FPS average to 44 FPS average over clocks on reference that ties up Vega 56 with the 64 showing once again that see use matter a lot less than clock speed on and these architectures the hybrid OC gets up to 51 FPS average again of 16.5% over the overclocked reference 56 and now surpasses a stock clocked reference 1080 again with how close these cards are note that a partner model 1080 would out match the hybrid OC 56 but still it's an admirable jump if we're roughly tied with partner models that's not a bad place to be for a 56 granted it's at two times the power consumption but that's the cost to have the power play table mods for folks who don't care about power at 1440p we're at 98 FPS average over the hybrid OC which is a gain of 11 percent over the overclocked reference card and the Vegas 64 add stock clocks the 31 percent gain scene is over the Vega 56 reference card with no overclocked and is again impressive we're at way higher power consumption once again but have boosted performance greatly 1080p post similar results we've gained 8 percent over the overclocked reference card which is getting much less exciting so we'll move on brace yourselves to the next one Ghost Recon wildlands at 4k positions us at 32 FPS average for the stock card 39 erage for the 50 percent offset and 41 FPS for the overclocked reference card by 9 percent 9 50 megahertz HP m2 with the 50 percent offset compared to the power offset that overclocked does almost nothing the hybrid OC doesn't do much either as we gain just over 5 FPS versus the 50 percent offset with a stock card so it's not at all worth the power consumption or the time in this instance 1440p is next to Vega 56 reference around 67 FPS average and the 50 percent offset is at 65 FPS average have 14% improvement just from the offset but the overclock and the 50 percent offset get us only an extra 3% disappointingly a hybrid OSI model entities which is another couple percentage points gain so not all exciting 1080p follows this trend and scaling nearly exactly so we'll skip it he'll play does last 10 is our least populated chart we only just started testing this game so excuse the lack of data let's just focus on Vega 56 is scaling for now at 4k the reference 56 operates at 27 fps boost 34 with the reference OC a hybrid OC gets us to 37 FPS average both for massive gains even with the reference overclock and over power we're doing pretty well we were clearly starved for power and clocks in the reference testing we're not sure what specific piece of hardware in the GPU caused the choke point at 4k but the overclock and over power helps significantly in addressing it 1440p scales a lot less impressively but there's also less ground to gain the hybrid OC card puts a 71 FPS average outperforming the overclocked reference card by 8% and the stock card by 13% again an 8% gain is hard to get excited about particularly given all the effort and power that went into the card so that's it that's the Vega hybrid card the 56 if you takeaways here one compute units are a lot less important than just frequency if your option is vacant 56 overclocked or Vega 64 stock you might as well just go with Vega 56 even if you're considering overclock in 64 these cards are gonna cap out at roughly the same area anyway so end of the day 56 remains the way better buy from Andy 64 just it doesn't really have a place in the market 156 does so damn well by comparison just a 50% offset and 5 minutes of overclocking without any of this stuff all the cables and mods and registry edits without any of that you can still achieve Vega 64 performance if not slightly outperform it so I guess the takeaway is Vega 56 looks pretty good once you start doing all this stuff to it especially pricing notwithstanding we're ignoring the pricing right now altogether because it's it's too hard to follow frankly so what we've got at the end of the day is a unit that does now 17 32 to 17 42 megahertz on the core 980 on the HBM to stable anyway with power that's 70 to 80 watts higher than it was with the 50% offset and performance that's at times completely not worth it at all as in like a couple percentage points to actually really good at 16% in a couple of the cases of the games that tend to favor Andy and show those responses to the clock changes so just what you're doing I don't know this might not be something I'd recommend the powerplay tables are not something you should mess with you shouldn't mess with the anything that takes the card out of a stock state in terms of power delivery unless you know how much power the vrm can take and we have it in theory build Zoids vrm analysis of the vega frontier edition kind of addresses all that stuff if you'd like to watch it and learn more about how much power can go through the thing but you still take risk when you're pushing a hundred percent offset power through the card so keep that in mind we did pretty well though it's just for 24/7 use or for actual user placement and gaming rather than benchmarking it's probably less advisable but that's totally up to you either way we've learned a bit here and it was fun so thank you for watching as always you can help us out with this stuff directly by going to patreon.com/scishow to gain root access squarespace.com to make them a shirt like this one subscribe for more thank you for watching I'll see you all next time you