Mistakes Were Made, RTX 4070 Ti Super Review Update

[Music] welcome back to Harbor unbox today I have a quick update video for you a quick re-review of the RTX 4070 TI super I guess you could say for those of you who missed it I noted in our TI super review that just a day before the review was set to go live Nvidia informed us that the MSI Ventus 3x model that we had been provided with for testing was underperforming by as much as 5% and this was a huge problem as we obviously wanted to evaluate the RTX 4070 super accurately but there simply wasn't enough time to fully retest it in fact testing aside these reviews take an enormous amount of time to create the week leading up to the 4070 TI super and 7600 XT reviews I was working 16 to 18 hour days pretty much 7 days a week but that's not a complaint on my behalf I absolutely love this the point rather being that 24 hours just isn't enough time to reproduce just days and days of testing but we're also not going to post a product review with an incorrect conclusion due to inaccurate testing so with the little time that I did have available I slotted in the Asus tough gaming version of the RTX 472 super into our test system quickly validated our 40 40p data and after a few hours it was pretty clear that the data that we had was within a few per of the tough gaming and therefore our conclusion wouldn't change and this really was a bit of a boring release from Nvidia now I promised you guys I would go back and update all of our data using the tough gaming model and feature that in our RTX 480 super review next week but since I now have all of the data I figured why wait we can just quickly go over it now also the slightly underperforming MSI Ventus 3x graphics card wasn't necessarily my fault uh but I did make a sloppy error that was my fault and I'd like to apologize for that and then correct it shortly after releasing that video which went out at 1:00 a.m. local time mind you I was made aware of a sloppy error in my specs table where I got multiple specifications wrong stuff like the RTX 480 core count and L2 cach capacity of the RTX 4072 super so pretty basic stuff there bit embarrassing anyway I pinned a comment on the video at around 2 a.m. explaining the error I also then trimmed that section of the video out using the YouTube trim function but unfortunately this does require an entire uh well requires the video to be reprocessed which generally does take several hours so that being the case I was completely exhausted and I went to bed at around 3:00 a.m. so upon waking up I quickly headed out to the office and found that YouTube had processed the video I then updated the pinned comment and then I had to change all the time stamps as well because they were now out because about 2 minutes of the video had been cut removing that specification section that I got wrong I also updated the specs table and the script for the written version over at text spot and in the pinned comment I linked people who were interested in looking at the correct specification to go over to the tech spot review so that was my screw up sorted and before we move on here is the correct specifications table the RTX 470 TI has 13% fewer cores than the 480 and 10% more than the 4070 TI it also packs the same 48 megab L2 cache as the 470 with the same core clock speed the key upgrade of course is the memory bus which is now 33% wider at 256 bit but with slightly slower 21 GBS per second memory the bandwidth is just 6% less than that of the RTX 480 so those are the correct specifications and again sorry for the mistake in the original video which has now been cut out okay the next thing to address Performance but rather than go over all of the games tested individually as there are a dozen games tested for rization performance at two resolutions and then 10 games for R tracing I'm just going to show you the average data and then we'll go over all of the individual game data for the upcoming RTX 480 super review next week at 1080p we see almost no change just a 1% uplift with the new data using the Asus tough gaming RTX 470 TI super so there's not much more to say here let's just move on to the 1440p data now 1440p we're seeing a 3% increase with the tough gaming over the Ventus 3x going from 120 FPS to 124 FPS this means whereas previously the 4072 I super was just 3% faster than the 4070 TI on average it's now 7% faster which I believe is in line with most testing so just a very minor Improvement overall here and this is where those gains come from interestingly we did see a rather large 9% uplift in Hogwarts Legacy with the tough gaming with fairly minor gains elsewhere 4% Resident Evil 4 and a plague Tower aquum with little to nothing everywhere else now at 4K we're again looking at a 3% performance increase on average when using the tough gaming model taking us from an average of 68 fps to 70 FPS still this meant where as previously we found the RTX 470 TI super to be 6% fast than the 4070 TI it's now 9% faster so about what the difference and specifications would suggest we'd see this time the biggest Improvement was seen in Dying Light 2 where the tough gaming was 7% faster than Watch Dogs Legion with a 6% boost 5% for a plague Toil requium and then just 4% in C Punk 2077 and everything else was at 3% or less then we have the rate tracing data and here we're seeing a 2% increase when calculating using the geoman making the 4070 TI super now 6% faster than the standard 470 TI then if we calculate the data for each title individually and then average those results we end up with a 3% uplift it's near enough to a rounding era and typically we're looking at a 4% increase though there were a few titles that saw no performance change at all as for how these changes affect our review conclusion well they don't which is why we went about publishing the data that we had as you'd expect a 3% increase in performance has improved the value of the 4072 I super by exactly wait for it 3% and believe it or not that doesn't change our thoughts about the product but this is just the MSRP data so let's see what this does in the retail Market here we see that instead of costing 133% more per frame than the 7900 XT the 40702 super really costs 9% more per frame which is certainly better but it doesn't change my thoughts about this product and the standard 4070 super still represents better value as I said in my original review so there you have it our updated data using the Asus tough gaming version of the RTX 472 super and of course that fixed specifications table again sorry for the mess up there I still feel that the 4072 super is a very underwhelming product and while it's nice and I'd suppose necessary at the $800 price point to go from 12 GB to 16 GB of vram after a year though you'd probably expect Nvidia to deliver a bit more than an extra 4 GB of vrm and a mere 6 to 9% performance uplift depending on the resolution used for testing as for the MSI Ventus 3x model or really all MSI RTX 47 TI super graphics cards you can now download a firmware update which takes around 20 seconds to execute and it will address the missing performance although neither MSI nor Nvidia have explained what the issue was I suspect it's power related and this is because the affected MSI models were all provided to the media well in advance so they very early uction models and it's possible there was a late change in specification it appears to me that the MSI models were shipped with a lower TDP which can result in a performance loss despite clock speeds appear Inc correct and this is a result of nvidia's advanced power gaing technology which can gate off some of the SMS to save power so while the GPU might be running the cores at 2.7 GHz not all cores are active due to a limited power budget and this explains why 1080P or some of the less demanding games we saw little to no difference in performance between the Juventus 3x and tough gaming while in the more demanding scenarios we consistently saw a performance improvement with the Asus model and this would also explain why I heard reports that some Ventus 3x models were 5% slower on average While others were 1 to 2% slower basically the reports were all over the place but this makes sense as silicon quality does come into play here higher quality silicon will perform better when power limited which is why power M for gpus and CPUs ensure that even the worst silicon can achieve the stock or Baseline performance whatever the case this was a massive headache but ultimately a bit of a nothing Burger as well at least in our case where we only saw a 2 to 3% Improvement overall if it was 5% or more than sure anyway I'd better get back to more testing as I said we have the 4080 super next week along with a few other new products so there's really no time to waste if you appreciated this update then consider becoming a harbor box member float plan patreon you can SL to either one of those things uh and it's a good way for us to stay in touch with you guys as well I was able to keep our members a bit more up to date uh I pinned all of these uh results in our Discord server a few days ago so people knew what was what and that we were a few per off there so anyway we get yeah access to our exclusive Discord server monthly live streams Q&A stuff and behind the scenes content so check that out if you're interested if it's perfectly fine and I would like to thank you for watching this video I'm your host Steve see you next time [Music] [Music]

Channel: Hardware Unboxed

Views: 162,307

Keywords: hardware unboxed

Id: 98ogL7ijrik

Length: 9min 50sec (590 seconds)

Published: Thu Jan 25 2024