Gadgetory

this is something we haven't seen before this is Nvidia taking a relatively successful card the GT 1030 and sticking ddr4 on it so rather than gddr5 which basically every card barring HBM uses these days they've gone with system memory which is actually a tremendous downgrade in memory bandwidth the GT 1030 is named the GT 1030 and both of its current iterations the gddr5 version and the ddr4 version but it's got a few differences beyond just the memory - like the clock speed reduction this is something we've railed against with other cards in the past and the Istana and videos Don it they've both done this but this specific instance is particularly concerning because what you have is a device that is heavily dependent on memory bandwidth getting a reduction in memory bandwidth from 48 gigabytes per second to something like 16 gigabytes per second with some of the game performance results showing differences upwards of 2x so that's what we will be talking about today benchmarking the GT 1030 bridges the GT 10:30 with ddr4 it's unfortunate really because the GT 1030 had more success than it ever deserved because of the mining thing going on but despite having recommended it in the past we struggle to do so now because it's so easy to accidentally buy the wrong one and really get punished as a result of it so I'm not really sure what we can run for the ad for this particularly critical video but I'm sure we'll figure something out before that this video is brought to you by us and the gamers Nexus anti-static mod map the GN anti-static mana is a four foot by two foot surface two millimeters thick of high-quality industrial grade anti-static material and it includes a common ground point for Earth a grounding wrist strap and it has on its electrical wiring diagrams that may prove useful a GPU silhouette and grid for your teardown efforts and other useful items go to store like Aaron's exes dotnet to pick up a GN mod mat today the GD 10:30 with ddr4 is something we've railed against in the past and it's because it's one of the most egregious missteps and product marketing we've seen really honestly it deserves to be called something like GT 1020 or something like that because this inherently having ddr4 on a card it's not good for performance but it's something that we could forgive if it were very clearly marked as its own product category with any GT 1030 right now the only difference you legitimately see is if you look closely enough on new AG or Amazon or whatever you might see that it says d4 or something like that instead of g5 most consumers are gonna have no idea what that means it's just going to look like any other randomized product serial number basically so that's the problem we have with it the other problem we have with taking a product changing it and keeping the same name is that it makes us look bad in media because you're going to have consumers especially the low end who don't know what they're buying see performance numbers from us from others for a GT 1030 buy something and get extremely different performance results and that's not because of us it's because the product changed later and the name was kept the same so you can still buy both versions of the GT 10:30 you just have to be very careful about it graphics devices are actually bound by memory bandwidth so we go over some of the basics as to why the differences are so prevalent here before we get into the actual performance numbers so if you want to learn about why they're different typically speaking anything that is really dependent on memory operations going across the bus will be something like textures a large file that you're pulling from somewhere to later recall and apply to an object or a mesh in the game and these things take memory bandwidth when you're calculating memory bandwidth between these two devices it's roughly 48 wears just 16 gigabytes per second the lower beam ddr4 and that difference is pretty easy to to come up with so basically you take the memory bus width divide it by 8 to convert 2 bytes just for ease and then after that you multiply it by the memory clock you multiply it by 2 for DDR or double data rate and then you multiply by 2 for gddr5 if it's a thin and the results again 48 versus roughly 16 and the difference alone is massive but NVIDIA has also changed the core clock of the ddr4 version so the reference baseline clock for this card the d4 version is about a hundred megahertz or a little over below what the original GT 1030 gddr5 card was and that's also something that's worthy of note but not really a huge impact at a performance considering the rest as for what's trains memory bandwidth it's just about everything in computer graphics texture sampling is a big one or anything else that collects samples for processing anti-aliasing temporal effects that need to keep frames in memory scales and transforms of textures where data it needs to be retained in memory and so forth the differences should be exaggerated in games where shaders are lighter load as memory bandwidth is likely to be more utilized anything that relies on heavy memory occupancy will strain the GPUs memory bandwidth most of the time that be higher quality textures anti-aliasing effects and some fixed function processing the CPU and geometry pipelines become more bound by draw calls and geometric complexity whereas the GPU becomes more bound by large files getting calls for application to objects and meshes or files that might remain resident in memory for further transactions and things like transforms as always getting into the numbers testing methodology and the test bench can be found in the article links in the description below if you want the written format of it or more information on how we tested some information here we used an i7 7700 K which is our old standardized GPU bench be replaced eventually and we use that to eliminate the Seaview bottleneck as even a possibility we have some numbers with a G 4560 just to give you a reference point of where would perform with a more reasonably priced CPU and then we also have some apu numbers which of course use completely different hardware because it's a different platform so we've got all that in here some of these numbers are older like the APU numbers are from launched they haven't been retested performance might have improved a bit but the point isn't to compare these versus AP use so we didn't rerun it the point is compare them against each other and these are the two devices that were run primarily for this benchmarking today well start with rocket League for this one rocket league is a fairly playable game overall previously we found rocket League to operate between 50 and 60 FPS average on the and the r3 and our five AP use and we found the GT 1030 and G 45 60 to operate at around 62 FPS average removing the CPU bottleneck for safety we found the GT 1030 r5 GPU and 7700 K we're at 63 FPS average representing AG view bottleneck this is what the g5 one again 1% lows and 0.1% lows were in the 40s the ddr4 version of the GT 1030 stock ran nowhere close to that frame rate it slumped to 28 FPS average a 55 percent decline in performance from the gddr5 version of the GPU or if you do the math the other way it's about a 2x increase to move to the g5 version with an overclock offsetting the core and memory by 250 megahertz each we managed to push 235 FPS average but frame times suffered as a result the GT 1030 ddr4 card can't even achieve half of the performance of the original GT 1030 and they share a name and that's not right it can't even operate anywhere close to the level of an AR 320 200 G stock apu this thin is getting obliterated by its memory bandwidth limitations this next chart shows the frame time performance between the gddr5 and ddr4 cards 16 milliseconds would represent a 60 FPS throughput and we'd ideally see a flat line as that represent the smoothest frame delivery or in another ideal scenario the least amount of variance frames of frame as possible in frame intervals in this instance the ddr4 card struggles to maintain a constant throughput and chokes regularly on frame to frame intervals resulting in an overall stuttery unplayable mess when compared to its ddr5 alternative that costs about $10 more dota 2 is next at 1080p the R 3 and R 5 numbers from launch benchmarks put the APS in the 40s to 50s for frame rate with the GT 1030 and G 45 60 getting stuck around 63 FPS average unrestrained by the CPU we measured the GT 1030 gddr5 performance at 67 FPS average with lows at 60 and 43 the ddr4 version managed shockingly to carry almost 50% of the original performance our GT 1030 ddr4 ran an average frame rate of 32 compared again to 67 on the model that costs a couple dollars more and somehow carries the same name so far this is just embarrassing csgo is up next another valve title for this one the AMD APU is at long stuck around the 80s to 90s for average frame rate pre overclocked the GG 10:30 and G 45 60 did well operating at 120 FPS average and originally receiving our recommendation for an ultra budget csgo machine at the time unrestrained at the GD 10:30 with gddr5 managed 120 2.7 FPS average with a low is relatively disparate at 85 and 54 fps the GT 10:30 with ddr4 embarassingly managed 55 FPS average with a 0.1% lows at 25 FPS the translation of these low values is again a stutter II massive frame throughput will clarify with frame time plots momentarily 55 FPS versus 123 is a giant reduction in performance from the original card and overclocking the ddr4 card helps but not in a way that can begin to defray the degradation once overclocked swear at nearly 2000 megahertz core versus the 1800 megahertz core rough max of the GT 10:30 with gddr5 and yet performance is still 47% lower than the original version of the car here's that frame time plot although neither card is the pinnacle of fluidity the ddr4 variants can't keep its frames consistent enough to provide an enjoyable experience without graphics quality or resolution reductions moving on to ashes of the singularity no never mind this error popped up when we tried to benchmark Ash's dx12 with the built-in benchmark the card can't even officially run it because it's not officially supported as a result of its lack of gddr5 Ash's thinks that this card is from the G ddr4 era and just won't even let it run without further manual modifications to the benchmark so we'll move on Sniper Elite for at 1080p high with dx12 and async compute acts like a synthetic test here realistically it dropped to medium settings for playability but we can still use some stand-in synthetics and that's how we're treating this one for this one the GT 10:30 with ddr4 again operated out exactly half the speed of the gddr5 GT 1030 further shaming the poor cards existence we're treating doom the same way basically synthetically the reason for this is because we already have data racked up for other low-end devices with these settings though the settings are way to abuse it for the 10:30 we can easily compare it to other devices trouble with doom is that it's 200 FPS physics tap means that we have to impose higher settings to restrict high-end devices because they'll hit the cap otherwise and appear artificially worse anyway looking at it synthetically the GT 10:30 with g5 manages 16 FPS average which is 20% faster than the GT 10:30 with ddr4 the gap closes a bit with an overclock and comparatively this is miles behind the RX 560 and gtx 1050 which operate at 65 point seven and 44 FPS respectively it's probably time we start analyzing the clock differences here and most of our previous benchmarks he saw the ddr4 version was overclocked and the g5 version was not well it's because we didn't have to overclock the g5 version on the ddr4 card overclocked with an offset of at least 250 megahertz on memory and on the core the card was never even close to catching up to its gddr5 origin this one the first card even with that overclock so that tells us this is severely crippled by memory bandwidth and using fire strike we can start to show why and how it's crippled and in what ways it can attempt to catch up here's an overtime chart for fire strike we'll add more lines as it goes the first two lines are the gddr5 and ddr4 GT 1030 s without modifications you'll notice that they're fairly close overall indicating that fire strike would be a good clock for clock tests without needing manual modification our differences never exceed 30 megahertz which is reasonable it's obvious that a 30 megahertz difference will never amount to a 50% change in performance so we can live with this we did another test pass with the EVGA GTX Core throttled with a 90 percent power limit so we can add that line next this one regularly dips down below the ddr4 card score clock but as you'll see in a moment it still outperforms the score significantly finally this next line shows the overclock to GT 1030 ddr4 is core clock which was offset by 275 megahertz for this test it's peaking at 2,000 mega Hertz which is an impressive gain of nearly 300 megahertz in some instances even still performance is behind the gddr5 card and the core can't make up for the choke created by memory bandwidth particularly on texture filtering let's look at the fire strike scores associated with these numbers the GT 1032 gigabyte ddr4 card is an embarassment and operates with a graphics score of 1922 and combined score of 537 the GT 10:30 with gddr5 also stock managed a 37 93 graphics score which is a 97% performance improvement nearly 2x we know that fire strike is intensive on GPU memory and this fully solidifies the ddr4 card's weakness had there been any doubt just to be sure we reduced the GT 1030 gddr5 cards power allowance to 90% resulting in a 36 87 graphics score that's still nearly 2x higher than the ddr4 card overclock can be ddr4 card by two seventy-five megahertz offset results in a 25% increase over the stock ddr4 card that's not enough to clawback the tremendous loss versus the gddr5 device just in case you were wondering here's a fire strike chart with some other devices on it and we'll highlight all the 1030 devices for clarity this is tested on the same bench and provides an easy comparison to other devices so to be clear here this thing the ddr4 card don't buy this it's garbage it's like a $10 price difference the GT 1030 as noted earlier had way more success than it typically would deserve but we were ok with recommending the gddr5 version of the GTS on 30 and actually we would almost never recommend a card of that low of a price class because it just makes so much more sense to step up to something like a GTX 1050 or maybe an rx 550 or something like that but with the mining market the way it was and GPUs and all of that it made sense in some instances to recommend a GT 1030 so we gave it the benefit of the doubt it was actually a decent card for the price for something like a g40 560 not something we like to recommend but given the circumstances a few months ago something we had the recommend for the price class now however with a ddr4 version out there to confuse buyers especially buyers in the budget market who don't probably don't even watch us so they're never gonna see this message they're not going to know the difference between the two cards because a lot of the time the difference is in the name to the extent of something like GT 1030 whatever OSI Edition lp4 g5 or LP 4d for those things look like just just a serial number basically so there's no reasonable way you can expect the average consumer to differentiate between them because it has the same freaking name otherwise and it performs way worse so do not buy the ddr4 version of the GT 1030 and at this point almost just just to send a message I tape I don't buy the gddr5 version either because it's just kind of messed up to have the same name for two extremely different products imagine what would happen if say Intel released an i7 8700 K special edition that's actually it has two cores and is hyper threaded and that's it or if AMD released an r7 1700 V - that's an r3 product it's it's an extreme that it's at that big of a difference because the memory bandwidth dictates everything for the performance of these low-end devices especially so we take issue with it and would encourage you not to buy this garbage product with ddr4 memory but otherwise if you're curious about the performance you've got it now and hopefully on video considers renaming it something like GT 1020 or literally anything else that's not already used so that's it as always if you want to support us so that we can continue buying things that vendors definitely won't send us you can get a store documents access net and pick up one of our mod mats the next round will be in shortly or go to patreon.com/scishow and access for a smaller contribution and to join our discord server subscribe for more as always thank you for watching I'll see you all next time