Gadgetory

and these frontier editions seems to have confused about half the internet with its product positioning landing at a one thousand dollar price point and offering both gaming and professional level performance but without any of the professional level driver certification this makes the card most immediately analogous to a Titan XP and today we're reviewing the Vega frontier edition most heavily for thermal performance noise versus thermals power consumption noise stand alone and some smaller additional tests on production and gaming workloads before that this coverage is brought to you by the core g21 enclosure from thermal take a $70 case with two four millimeters thick tempered glass side panels mesh ventilation in the front for breathability a rarity in cases these days and a power supply shroud with top mounted SSD sleds learn more at the link in the description below again a large part of our review we focus on noise thermals and power because we think we can cover those the most adequately and with the greatest depth while hopefully offering some unique insight to the card but we do have some production and gaming tasks in there and before anyone starts screaming that this isn't the gaming card there are a few things to clear up about the Vega Frontier Edition first of all for thermal power and noise that's largely unchanging power does change a bit with workload but by and large when you look at the results that we present here it will apply pretty much everywhere so that gives a good understanding of how the card will perform in various workloads and also gives us some insight as to Vega as a whole as an architecture so that's number one the second part is with regard to product positioning we spoke with AMD in length about the Radeon at Vega Frontier edition cards unique position in the stack the device isn't quite a card targeted at the core gamer audience but also isn't exactly a professional class simulation card where you would need to pay for driver certification for assurance of accuracy the Radeon Pro line is still reserved for that side of the market which the Radeon Vega f/e does not fall into as it does not have those driver certifications and that means that this card is split marketed and again speaking with an v about this at gaming and developer applications this isn't for scientific or compute users who would be better served by fire pros but in our conversation with the companies they were big on the fact that it's a pro ready card with drivers but that it also has gaming is packaged in similar to a Titan XP except with more pro level optimization that makes Vega Fe a very oddity card indeed because it's positioned outside of the professional scientific space no certification no purchase but also outside of the core gamer space just like the Titan XT is that makes Vega Fe almost directly comparable to the Titan XP though again it does have some additional changes in the pro area so your typical buyer would be more likely to be someone like a game developer who creates content on a gaming class machine and also needs to be able to play to have stud content on the same machine while keeping cost relatively low certified drivers are expensive and that's not an object if you're designing an aircraft but it is if you're designing a game and although both pro applications are businesses there's a big difference between I'm a game developer and I design aircrafts that could kill people if the simulation is incorrect which is where you would want the driver certifications you would want the cards that are certified for usage with whatever CAD applications those might be and when you're spending twenty thirty thousand dollars on software anyway something like a Quadro or fire pro card doesn't seem so intimidating so the places you would buy something like a 1080i or maybe rx Vega in the future would be more for home freelance user someone who's making money doing freelance work on blender or whatever it may be and can't justify the $1,000 plus purchases but still want the performance of a decent card that's the distinction because even if you're just a game developer who doesn't need those pro certifications you can still make use out of the more expensive cards it's still a business expense buying something cheaper probably cost you more money in the long run anyway but there's a good cost savings over something like what Boeing might use for example so that's kind of where the card is targeted and that means that there are some confused comments online some of this I think can be drawn to the name of the card Vega it has adopted the name of the architecture which has been hyped and discussed so long now particularly in the gaming space because people want high-end competition for NVIDIA gaming cards that Vega F II has sort of become the de-facto flagship card for Vega as a whole the architecture and therefore is being seen as a Arta that could be a gaming targeted device and much like the Titan XP is not a good purchase for gaming when you could consider something like a 1080p I instead get very similar performance or not much of a difference in really anything except for price Vega F II is not really meant for gaming but could still be used to play games now the thin AMD told us is that it's basically a distinction of you develop things you're a content creator and you also want to be able to consume that content on the machine whether it's to play test or otherwise so that's the distinction now as for the testing as always full testing methodology will be in the description below and the article we have a lot of different test beds for this one for power thermals gaming and noise so those are all defined in the article drivers are specified in the article but they were the launch drivers made available publicly by AMD as this card was purchased and not a review sample so it wasn't tested prior to launch and that more or less goes over all the basics now there's one more note here overclocking we had some good success with overclocking the core originally but ran into some issues that were either bugs with wot man or just challenges that were not sure how to overcome just yet the two primary issues we ran into you were with HP m clock speeds when overclocking the core and with the fans beat so speak into these we were able to get the core overclock so something like 1670 megahertz we had 1682 fairly stable for a bit depending on the application it crashed in a few but survived and others ultimately settled around 16 67 megahertz and the problem was that the vram the HBM 2 seems to fall from 945 megahertz to about 500 megahertz when doing any kind of core clock change and wot man that includes under clock in it so as a test we dragged that bar down to negative 100 and the HBM readout I don't know if it was actually an accurate readout but through wot man the HBM readout said 500 megahertz when it was originally 945 even if you restored a default after attempting any kind of change in watt man negative or positive overclock on the core HBM would we'll be stuck at 500 and would not reset until rebooting the machine so for that reason a lot of the performance results with the overclocked version we're actually lower than the stock and that's because HBM was running lower than stock quite significantly actually so we only have one overclocked result in here pretty far n to show that change otherwise they've been eliminated until we understand either what we need to do to make it work or if AMD needs to roll out an update there's we're not clear on if it's an issue of this is a new card we don't know how to work with it or if it's an issue of wot man is buggy so that's something to note now as far as band speeds once you ramp them up they're kind of hard to ramp back down take several minutes and the best way to fix it is to basically go to auto fan speeds walk away from it come back and then try to only ramp up and not down because ramping up is not much overcome also there seems to be an offset of about 200 rpm on the wot man reading for the fan speed and the actual reported fan speed so keep that in mind as well but we accounted for all that in the testing all of that aside let's start with benchmarking for power versus thermals we're starting just with Vega and then we'll add the Titan XP for comparison under our thermal benchmark we measured at the wall the Vega FB stock configuration draws an average of 384 watts load with the overclock very in drawing five to the two watts average from the wall our peak power consumption is for 42 watts during the stock test and 569 watts during the overclock test thermally we're averaging 79 nine Celsius with the stock configuration and 7501 salties with the overclock configuration accounting for ambient which is lower because the fan is manually controlled to 3,700 rpm in that test adding the Titan XP stock card we see an average power consumption of about 357 watts so that's 27 watts lower than the Vega EFI system the same test at a maximum power consumption of about 370 dot 5 Watts about 72 watts lower than the Vega FPU configs max power consumption the Titan XP runs lower power consumption overall but also hires thermals and this has been a criticism of ours since we first reviewed the Titan XP and his resultant of the blower design one item of note is that the Vega EFI card has a much more sporadic clock behavior as the clock rises and falls frequently throughout testing we see this reflected in the therm and power lines while the Titan XP remains more stable we're not yet sure the connotation of this behavior and aren't sure if the root cause is related to Andy's execution of the software or of the architecture regardless we'll simply call it an observed behavior for now this next chart shows power draw and temperature trend as plotted against a 22-minute test run of spec view probe twelve which performs trace based testing of 3ds max Maya and other production applications the spiciness in the chart is from test starting and ending and they do so through an intervals because they're different speed cards with Vega Fe the total system power consumption sits reliably around the 425 watt mark pretty consistently actually the few dips down to 340 to 380 watts in the middle by and large though we're at around 425 for most of these tests adding the Titan XP to the chart reveals its total system power consumption at around 340 watts for the initial quarter of past in and around 396 to 4 or 4 watts for the second half of the test they're all performance for both devices is largely mirrored in this specific test pass as neither is permitted enough time to heat up to throttle points between the test execution for other power draw tests our Vega Fe air system draws 381 watts in a fire strike workload which is a delta of 40 watts more than the stock Titan XP at the wall and a 1080 ISC to stock card we also see a power draw of 401 watts and Ghost Recon wildlands which is about 31 watts more than the Titan XP and 1080i and for honor we're at 393 watts or about 37 watts at more than the Titan XP testing thermal performance comparatively next with a steady-state chart plotting noise normalized performance at 40 DBA we see AMD's at Vega EFI card operating at 59 9 DBA ranking it alongside the worst gtx 1080i partner card we've tested the armor while still retaining a lead over the reference to Titan XP card which is saying something these numbers are delta T over ambient so if we account for ambient the Titan XP was operating around 96 Celsius with this 40 DBA output and clock throttling as a result to be fair and II was also clocked throttling on the Vega F II card dropping about 9% off of its clock by the end of the test which is certainly significant but the point is that while both of these devices can avoid clock throttling by running higher p.m. neither of them is quiet while doing so this is precisely why AIB partner models exist our liquid called hybrid model the Vega app ecard will further emphasize this shortly but to provide perspective versus the AIB partner 1080i cards which is when we started using this testing methodology and to their abundance we see a prime example of noise efficient cooling in the asus strix ROG card the gaming X isn't far behind and blower fans in general have just never really been good at this particular task and that remains true with Vega Fe that said big F II does manage to retain lower temperature at its noise level than the Titan XP neither should really be run at 40 DBA though and instead should be left to run higher fan RPMs reduce clock throttling still using our 40 DBA tests we're not looking at MOSFET measurements on each board with the hot spots measured with thermocouples we see Angie Vega EFI hot spot set around 50 4.2 Celsius delta T over ambient switch although a few degrees warmer than the armor is still completely within operating spec for a MOSFET remember that power components can take far more heat than a GPU with many supporting 125 or 150 C before any major problems developed in this instance we remain below a DC prior to our Delta P calculation and that is acceptable speaking of board components let's take a look at some of Vegas's component temperatures under three different states otto OC and a 3700 RPM fan to prevent throttling and our 40 DB a noise normalized config the GPU certainly runs warm but is nothing out of the ordinary for a blower card you see this performance almost identically on both end video and AMD blowers hence why we normally recommend buying the AV partner models for the frontier Edition though it's between air and liquid units from AMD right now we'd recommend a DIY option as possible which is what we plan to find out soon regardless about temperatures in both our chores and hot spots are well within operating range the backplate itself which was measured on the inside of the aluminum plate opposite the top fat measurement posted a temperature of around 61 to 65 Celsius which means that the card gets physically hot to the touch when it's running but just don't touch it when it's on I guess this next chart shows fan noise at each rpm measured in 10% increments at a significant milestone like lowest duty cycles and average duty cycle for auto of note the Vega EFI card tends to auto spin at around 40 to 41 percent in our bench setup with a Titan XP spending out around fifty to fifty-five percent of its total fan speed in our set up keep in mind that the fan sizes and speeds are different so these percentages mean different things for each device the Vega F II card can spend up to forty nine hundred rpm for instance so it's forty to forty one percent Auto is around two thousand rpm this place is Vega Fe at around forty seven DBA output in its average Auto configuration with a Titan XP at around forty seven to forty 7.9 DBA output in its average auto configuration these two are pretty darn close the noise output when configured to their auto stocks and curb matching them and noise output is the best way to see a side-by-side but we kind of already did that in the previous noise normalize charts before getting the fire strike and gaming let's start with a few production tasks we're using a spec view per for 3ds max my Acadia and creo testing between the Titan XP and Vega EFI card primarily these tests were conducted in Pro Mode because Vega is targeted at content creators who game these choices make the most sense and are adjacent or inclusive of our core audience without going off the deep end in production tasks we'll leave that to our friends at Tech gauge when they get their review up 3ds max positions the Titan XP at 170 5.9 weighted mean fps during its bench pass with the Vega EFI card 17% behind at 140 9.3 FPS with the CAD modeling software cRIO AMD trades places with Nvidia and leads with 90 4.4 FPS while the Titan XP runs at around 60 or about thirty six point eight percent behind Vega F II the tide turns again with maya where nvidia now operates a commanding lead at 145 f7 FPS and Vega at 139 for deficit of 22% versus nvidia finally Cadia another cad solution has Vega in the lead at 150 2.3 vs one 13.5 for the Titan XP a commanding lead in this case for Vega we have other tests that we want to run but we'll revisit that topic later we finally did get blender working after working directly with one of the cycles coders from the blender foundation to troubleshoot issues with system hands and blue screens finally using the build at 2.7 8-9 cd6 b03 we were able to get the render to complete on Vega and on the Titan XP which had been completing no problem already we've been in comms with them throughout to troubleshoot further but for now we've actually got finished charge to look at this shows that power thermals and time to complete the render all-in-one the Titan XP finished this render in fifteen point one minutes which is why its wine ends sooner and the Vega card finished the render in nineteen point eight minutes for a thirty-one percent longer duration this was with tile sizes at 256 by 256 though the latest OpenCL optimizations in blender ensure that the GPU should remain at one hundred percent load regardless of pile size in the future so we optimize for the GPU with 256 by 256 anyway thermally the Titan XP sits at 45 percent fan speed for a noise output of 43 DBA and they sustain the operating temperature of 73 Celsius Vegas it's at roughly 2000 rpm Auto fan speed for a noise output of roughly 47 DBA and steady state temperature of 80 to 81 Celsius power consumption on Vega averages roughly 390 watts total system power drought while a Titan XP average is 323 watts total system power draw both Nvidia and Andy maintain their clocks of 1848 and 1600 megahertz respectively throughout the entire test so there's no thermal throttling going on with either even with the auto fan we can't definitively state that this performance will apply to all blender renders given the differences in material types and types of workload that blender can produce but the cycles renderer with our benchmark performs this way your mileage may vary this gives a good early indicator of course there's still room for blender to update further getting to gaming and synthetics we first start with fire strikes to provide a baseline for performance with fire strife ultra we see Vega EFI plays at 4906 points when stock which ranks at about 2.4 percent behind the GTX 1080 founded Edition card with gtx 1070 founders edition cards at about 15.6 percent behind Vega FP aftermarket models will obviously score higher but we're focusing on reference for today compared to a 1080 TI Vega EFI is about 27% behind the 1080i reference card and about 30 1.7 percent behind the 1080i gaming ex-partner model just to illustrate the type of games you would see with the partners if you had them as for similarly priced Titan XP cards the Vega F II operates about 32% behind of the stock configuration or about 40% behind the overclocked configuration as shown in these chart the overclocking issues we faced on Vega ended up lowering performance overall not raising it we think this is primarily due to that HBM clock bug or whatever it may be that drops the reporting clock by 400 megahertz but we're not positive we reached out to AMD and haven't yet received an answer but hopefully should have one for you soon for the rest of the charts now that we've Illustrated that overclocking issue will leave OC and out fire strike extreme shows similar performance with Vega FPS sandwich between the 1070 EFI with a 1070 EFI about 13 percent behind Vega and a 1080 EFI cards with Vega EFI about 5.5 percent behind the 1080 F II that's 1080 EFI and Vega EFI where Vega has lost some ground at 1440p versus the previous test Vega EFI is about 29% behind the 1080 TI and about 34% behind the titan XP reference card for fire strike at 1080 P times by an FPS numbers that correspond to these check the article linked below for gaming let's start with AMD's most favorite benchmarking scenario then move on to other games with doom at 4k and ultra settings using async compute in Vulcan the Vega EFI card operates an average FPS of 60 4.2 which places it slower than a GTX 1080 Fe by about 5% slower than a 1080i F II card by 28% and slower than a Titan XP card by 32% due to the thermal limiter going to a hybrid mod the Titan XP performs much higher as we showed several months ago and using the 1080i gaming addict increases performance somewhat significantly over the TI fe just for reference the 1070 FP runs a 54.7 FBS average which is about 10 FPS lower than the Vega F II card moving to Ghost Recon wildlands at 4k with very high settings the Vega EFI card runs an average FPS of 37.7 1% lows of 34 and 0.1% levels of 33.7 fps in the very least and these frame times are reasonably consistent here so that's good unfortunately the card still runs about 12% slower than a 1080 F II at 43 FPS average and about 32% slower than the 1080 TI FPS 55 FPS and consistently times lows for reference with Titan XP card operates a 60 FPS average with a 1080 TI gaming X at 58 FPS average the 1070 operates at 35 average with lows around 30 and a gtx 980ti operates roughly the same 33.7 FPS average at 1440p with Ghost Recon the Vega F II card manages to maintain 62 FPS average which marks it about tied with a GTX 970 and average FPS though Vegas low-end performance is still remarkably consistent with a 980 TI about 8 percent slower than Vega F II 1080i cards are all in the 90s alongside the Titan XP making the Vega Fe card about 32% slower than the slowest 1080 TI our 1080p results will be in the article that is interest you let's move on to another benchmark that's historically been favorable to Andy and Sniper Elite 4 with DirectX 12 and async compute at high settings and 4k Vega Fe operates a fifty-three FPS average with the lows close by this marks the 1070 F e card 10.2 percent slower and marks Vega epi about 11.5 percent slower than the GTX 1080 F II card or Vega about thirty percent slower than the 1080 TI f e card moving to a Titan XP the difference now becomes a 39 percent deficits on Vega F II maybe then Ash's with DirectX 12 and high settings will show some difference with this game Vega Fe runs an average of its 69.50 s with lows at 40 4.6 and 40.7 zr1 % the GTX 1080 Fe card operates about 10 FPS faster at seventy eight point nine FBS and the 1070 Fe card operates about nine FPS lower than Vega Fe at 61 Titan XP is nearly 100 FPS average for its persistent difference of about a 30% deficit on Vega F II since the launch of this card it seems that some of the community has been a bit disillusioned by Vega FP a few things here first of all and the at least in their initial dem I was never said that they were competing with a 1080 TI at CES when the 1080 I did not yet exist the demo that they had set up was a GTX 1080 a Vega card with vsync on in some game demo so no you can't see the fps in those demos it is entirely possible that Vega FP was slower and probably was because vsync was on but the point was they're comparing it to a 1080 F II the community over the past couple weeks leading up to launch had amped itself at least in part to believing that Vega Fe would be competing with a Titan XP or a 1080 TI and gain reforms that's clearly not the case that doesn't mean it's a bad card what we can look forward to on the gaming side of things now that we've tested this and have a pretty educated understanding of performance of Vega hence buying it for the channel what we can learn is that we know the power consumption metrics we know some of the thermal behavior we know some of the clock behavior and in theory if Andy is on the same silicon which they should be maybe minus one HBM module or stack if they're on the same silicon with maybe a slight clock bump as has been rumored it is possible that our X Vega will be pretty competitive with a GTX ten eighty founders Edition at leaves because that's what we primarily tested if that's the case and if NV is able to make rx Vega in the 400 to 500 dollar range it would be pretty competitive so saying that there's no reason to panic it has been the wildermann and a whole lot of flame whores online about this card there's no reason for any of that on the gaming side our advice to you is not to speculate and wait because as it's clearly shown by this card there's a possibility it does okay in gaming when it's our X beta not when it's the frontier edition which costs thousand dollars much like this at twelve hundred dollars without the context of the 1080p I would it seemed quite a bit different than it does today as for performance the card the Vega EFI card is ahead and Cadia is ahead increo is ahead of medical and power software performance not in power consumption it's behind there as and consumes more power now with regard to gaming performance to give NVIDIA credit they do more or less crush Vega efi at the price point and performance if you look at something like a 1080i even a Titan XP that price goes out the window there but whether or not that's relevant is up to you entirely that Vega F II card is not something we would recommend for gaming just like the Titan XP is not something we would recommend for gaming wait for our X Vega or look at something else like a 1080 or 1080i we'll look at our x-ray guys soon enough but overall what we can look at here for our X Vega is the basics and then as far as performance and extrapolating performance my advice would be don't I would say we're not quite sure what this means for performance on RX Vega just yet there are still variables and keep your hopes realistic for RX Vega by which I mean don't hope for anything just wait and see we should have one to test if it's not sent to us we will buy one so we'll definitely have a review up and can followup with Vega at that time when it's appropriate to for gaming but for today we have a pretty good idea of thermal power and noise so that really lays out all stairs to save out there most are noise have some decent ideas from spec view perv and blender and then anything else will be done separately we have a hybrid model coming up shortly assuming we're able to mount it liquid cool it to this card so subscribe for that as always you can go to patreon.com/scishow and access to helps out directly with this type of testing thank you for watching subscribe for more I'll see you all next time you