The Review of NVIDIA GeForce RTX 4080 FTW3 and RTX 4090 FE

When it comes to choosing a high-end graphics card, one of the most significant factors to consider is the power target. The RTX 4080 FTW3 has a higher power target than its counterpart, the RTX 4090 FE, which means that for applications such as overclocking or native stock boosting, the FTW3 will give you a boost in the range of 15 to 30 megahertz higher depending on the measurement and application. While this number may not be significant enough to make a substantial difference in performance, it is still technically better than the FE.

One of the main reasons to choose the RTX 4080 FTW3 over the RTX 4090 FE is its superior noise normalized performance, particularly when it comes to power consumption. In some cases, this means that the FTW3 will be more efficient and less noisy than the FE. Additionally, the FTW3 is more likely to have an unlocked BIOS, which can make overclocking easier for enthusiasts.

However, the RTX 4080 FTW3 also has a smaller design compared to the RTX 4090 FE, which could be seen as a potential upside. Nevertheless, this limitation does not seem to be enough to justify the slightly higher price point of the FTW3 compared to the FE. The difference between the two cards is significant enough that it's hard to recommend the FTW3 over the FE unless you have specific needs or interests.

For professional users who require high-end graphics capabilities, such as those working with 3D art software, the RTX 4080 FTW3 may be a better option due to its increased memory capacity. However, this benefit comes with a catch: the FTW3's design does not support SLI bridging with another card of the same model, which means that users will need to purchase two cards if they want to use SLI. This limitation is due to the lack of standardized SLI bridge placement on both cards.

In contrast, the RTX 4090 FE supports SLI bridging, but only with cards of the same partner model. This means that users who want to take full advantage of SLI will need to purchase multiple cards from the same manufacturer. While this is not a deal-breaker for everyone, it can be frustrating for those who want to maximize their performance.

In terms of overclocking, the RTX 4080 FTW3 is technically better due to its power target and noise normalized performance. However, the FE's design limitations make it less desirable for enthusiasts who want to push the limits of their system. The fact that the FTW3's power consumption is more balanced between the PCIe slot and the cable also makes it a more appealing option.

For those who don't require the highest level of performance, the original RTX 4080 review provides a comprehensive analysis of its strengths and weaknesses. In general, the card is not worth the investment unless you have specific needs or interests that align with the FTW3's capabilities.

It's also worth noting that while both cards are high-end options, they may not be suitable for all users. The RTX 4080 FE has a slightly more powerful design and more features than the original RTX 4080 review suggested, making it a better option for those who want the absolute best performance without breaking the bank.

In conclusion, the RTX 4080 FTW3 is a high-end graphics card that offers superior noise normalized performance and power consumption compared to its counterpart, the RTX 4090 FE. However, its design limitations, such as the lack of SLI bridging with another card of the same model, may make it less desirable for some users. Ultimately, the choice between these two cards depends on your specific needs and interests.

"WEBVTTKind: captionsLanguage: enthe desk is an absolute mess because we've been using the card that we're reviewing today in an overclocking competition with jay and with paul from paul's hardware and you can check that content out separately but before we got into any of the overclocking and the mods that we're working on coming up like shunt resistor mods soldered onto the card we ran all the normal review numbers so that we could produce this look at the evga rtx 3090 fcw3 there is not technically an msrp that we've received yet and because this card is impossible to find anywhere on retail sites we're not 100 sure what it's supposed to cost but we have some ideas and we'll talk through today the value of the already strained 90 gpus when applied to a higher end partner model before that this video is brought to you by ek and the ekaio series we recently reviewed the ekaio drgb 360 as one of the top performers in our cpu cooler charts advantage from its fan performance and high quality pump internals that we found during our tear down if you're looking for a high performance rgb cpu cooler with longer steady state times check out the ekaio 360 and 240 at the links in the description below so as a reminder with partner models of any video card what you're really looking for is quality of life improvements and when we say that what we really mean is things like noise normalized thermals so who has the most efficient cooler when you set them to the same noise level not same type of noise necessarily but same noise volume who has the best thermal result who's got the highest sustained clocks as a result of the gpu boosting behavior where it's contingent upon thermals the the extent to which it boosts and then you're also looking at stuff like fan spin down some of the older reference designs didn't have that but partner models did now it's pretty much everywhere uh things like the thermistors that evga places on its cards so the fcw3 cards have icx thermistors those are ntc thermistors so it's like a smaller thermocouple basically it's placed in a few spots on the board and you can get a temperature measurement out of it these are not particularly actionable it's an extremely nice marketing point but there are a few places where they are useful so we find them useful for more extreme overclocking even for water cooled overclocking like putting the board under a water block which it's difficult to find for something like an fcw3 but they are coming out but the reason it's useful there is because you can just sort of mentally verify that you do have good contacts the memory to the mosfets things like that you don't have to really wonder about it and worry if you're burning something up so for extreme overclocking it's useful because when we're trying to figure out do we need to run a memory heater or is the memory running too hot without direct contact cooling we can look at those thermistor numbers and figure out okay well we're not really anywhere close to the 110 degree tj max so we're fine so that's kind of what they're useful for daily life using it out of the box no modifications not much overclocking to speak of the thermistors the the nine thermal sensors on the board are not really that useful but again we like them for other features outside of the stock usage so keep that in mind it might be useful to you it might be utterly useless but something kind of neat to look at and that's kind of about the end of it as for the rest of the card it looks like this will probably be in the range of 1800 we haven't gotten official confirmation if we get it before this video goes up but after it's edited we'll we'll put it right here on the screen but we're reviewing this under the assumption that it's about 1800 and i'll give you some alternative prices and where i think it stands towards the end of the video we don't expect many people will be buying 3090s but for those who are it'll be useful to look at how the partner models compare to the uh comparatively overbuilt versus previous generations 39d founders edition so this will give you a good look versus the fe card is the first 39d partner model reviewing don't know if we're going to review more but in the very least you get a look versus fe and where it lands so without further delay let's get into the thermal numbers acoustics and gaming performance frequency behavior overclocking behavior all that stuff and then if you want additional information on how our testing is conducted you can check our gpu testing methodology piece which we'll link in the description below let's get into it power testing's up first looking at averages and then we'll look at more depth in a moment the rtx 3090 fe ran 356 watts and a full load of furmark overclocking the fv got it to 408 watts making it one of the least efficient gains we've seen in a long time and contrary to nvidia's claims that you can clock this up somewhat efficiently the ftw3 starts higher at 384 to 391 watts and runs up to 464 watts with the fully unlocked shipping v bios we'd still like more power available for serious overclocking though the normal v bios is called lands 390 watts so there's basically zero change from the so-called ocv bios we'd also expect minimal changes in gaming benchmarks so this should theoretically mostly impact fan speeds based on what we're seeing here if you're curious a gaming workload reflects similar behaviors total war 3 kingdoms at 4k ultra lines up almost exactly the same way with everything in the range of 400 watts before any overclocking looking at the per rail power draw we observed the pcie cables at about 160 watts 135 watts and 115 watts the pcie slot draws a little bit over the official pci sig spec in this initial testing and hits 80 watts somewhat consistently it should be closer to 75 watts max this isn't much of a concern but if you were to run multiple of these anyway say three in a water-cooled config for multi-gpu processing you'll definitely want an auxiliary power into the motherboard to take some of the load off of the 24-pin cable and prevent any melting keep in mind that you can't sli three of them but you can still run three plus of them without a bridge for workstation applications where it might not really benefit from a bridge the power is just high on the slot overall and considering the 12 volt cables still have plenty of room if possible it makes sense to try and shift the load there we have a lot of thermals to cover but we'll start with the fan response to temperature under auto settings and as controlled purely by v-bios this is the out of the box noise test in this testing the rtx 3090 fe ramps slowly over a five minute period to about 37 dba and is more variable in its fan ramp over time the slower ramp period is the nature of having so much surface area and metal and it takes longer to reach steady state and soak as a result the ftw3 ramps faster kicking fans into gear sooner and also ramping faster the 3090 fcw3 levels out at about 39 dba at a 20 inch distance when fully auto that marks its auto fan speeds as a bit louder than the 3090 fe just within territory of being noticeable but that doesn't mean its cooler is any worse or any better we need to look at noise normalized thermals to determine the efficiency of each cooler before though we move into noise normalized thermal testing let's map the rpm to the noise that we just saw gpu temperature target on the fe card using a fire strike looping render landed at about 68 degrees celsius the evga 3090 ftw 3 hit 65 under automatic v bios configurations and note that again this is what the ambient controlled and monitored every second of the test at about 21 degrees celsius this is not a delta t measurement it's just the temperature the fan rpm for the cards is not directly comparable either because the fan count and the placement is different however the evga cards spun up to about 1900 rpm once at steady state thermals become tricky when normalized for noise anyway we equalize the fan noise levels so the cards produce the same noise but they aren't producing the same work the power is not normalized as we saw previously and the extra 40 watts allowed on the evga 3090 fcw3 means that even a noise normalized test is not truly a test of purely the coolers the 3090 ftw3 therefore has a limitation but we can still evaluate it with another metric the rtx 3090 fe cooler maintained about a 10 degree advantage at 40 dba with a 57 degrees celsius baseline versus about a 67 degree result for the gpu only on the fcw3 from this line plot alone the fe looks better and we can see how you would think that however if we look at the watts per degree celsius dissipated the evga ftw3 runs 5.8 watts per degree celsius while the fe cooler runs 6.4 the evga cooler is technically coping better when proportionately aligned for the power load but the fe gpu cooler is undoubtedly extremely competitive here and running cooler when considering its lower power target this next plot is one of memory thermals as measured at 40 dba for starters we'll look at the ntc thermistors included on the evga icx board and then we'll look at manually placed thermocouples these ntc thermistors are placed adjacent to some memory modules and we'll show the image of that now and if you're curious to learn more about these you can learn about it in our original review of the icx cards back when they launched in the 10 series these three ntc thermistors plotted around the same temperature across the modules which is good there's even cooling across the memory so the load is evenly distributed the result was about 70 degrees celsius for all three measurements there was one erroneous spike at around the 510 second mark the spike was to 6543 degrees celsius and is obviously incorrect the precision px1 software bugged out somewhere in the process of reading the thermistors because to our knowledge our lab did not in fact exceed the temperature of the sun during testing thus far the memory temperatures are well within spec even accounting for the delta between the inside and outside of the memory casing tj maxx is 110 degrees so 70 to 80 is a good place to be to verify the accuracy of these sensors and the software other than that bug we attached our own thermocouples to a hotspot top memory module and a memory module near the slot which often runs warmer because of its presence near the motherboard for this chart we averaged the icx thermistors to clean up the charts the average is just below 70 degrees our bottom gddr6x module measured about 66 degrees celsius as tested on the back of the pcb where the flip chip package is closer to the thermocouple as opposed to through the casing the top module measured about 73 degrees celsius averaged they come out to 70. so it's nearly identical to the icx results this is great news for two reasons first the icx sensors appear accurate but there is no electrical reason they wouldn't be so this isn't really a surprise unless evga were doing something screwy in software thermistors aren't an evga invention they're extremely common it'd be hard to screw them up although it is possible secondly the more important one is that it means our own testing is accurate so even without the interior junction temperature being exposed to us via software we can still produce numbers that are reliable and test it a delta of maybe about five to ten degrees celsius could safely be added for junction internal temperature this chart is from the same data but for the power components this includes five power measurements from icx and one from our own vrm mosfet thermocouple we placed this on the exact center thermocouple and we determined which one would be hottest by taking some thermography we'll plot these line by line since it's kind of messy the power measurements heat up to low 60s at best with the bulk of them in the range of 65 to 70 degrees and the worst measurement at about 78 degrees celsius our own vrm mosfet measurement plots last and is a much more stable line thanks to the placement being less sensitive and it also plots right in the middle of the pack once again we're mostly internally excited that our own measurements are proving useful here it doesn't really say anything about the card in terms of its usefulness or its quality but we're just happy to see that the numbers we're producing are good all of these measurements are well within spec for the components on evga side and again as a reminder this is 40 dba normalized although auto was running pretty close to this mark anyway we next need to determine the frequency of the cards to evaluate out of the box performance plotted in a frozen frame render and 3d mark the 3090 fe card shown in blue averaged in the range of 1890 to 1905 megahertz the evga f230cv bios flattened out around 1935 megahertz so it's holding about 30 megahertz higher than the stock fe some of this comes from the boosted stock power consumption the normal v bios for whatever that means plots right between them although it makes this chart completely illegible a ride to the middle and is not significantly different from either we're not really sure that evga meant to do it this way it seems kind of like a screw-up in the v-bios programming but they're mostly the same as each other as we get into games just a reminder that we're keeping a truncated list of games for partner models that's because the performance doesn't change much card to card so if you want the full scope of the 3090 performance we'd recommend checking our initial 3090 review which also has 8k benchmarks for partner cards the focus is mostly on thermals it's also on noise and overclocking but we'll give some basics for games here anyway for shadow of the tomb raider the evga rtx 3090 ftw3 plotted a 106 fps average for the ocv bios and didn't really change much for silent it ended up about 1.3 ahead of the stock 390 fe overclocking got us to 111 fps average and that's with a lighter oc we could hold 1200 megahertz memory and plus 80 megahertz core offset for both of those in competitive overclocking but it was tough to keep that stable in all games this config we're showing was stable though the uplift from the oc is 5.2 percent in average fps in red dead 2 with dx12 the evga 3090 ftw 3 plotted at 92.5 fps average for the ocv bios or 91.2 for the silent v bios or normal as they call it the results flank the rtx 3090 founders edition the sli results as a reminder don't improve here because the dx12 version of this game doesn't support sli vulcan does but that's in a previous video if you care about us a lot there's basically no difference here overall between the fcw3 and the fe and overclocking the 3090 ftw3 gets a 21 fps average which is an improvement that's noteworthy of about nine percent over the stock ftw3 and that's actually very good and we also found that changing the memory tune here benefited the card significantly so perhaps red dead 2 at least with the x12 is maybe a little bit more sensitive to the memory increases than to core increases although we haven't fully validated this that's kind of our takeaway from the initial data so far overall though it's worth overclocking these cards sometimes in single card configurations we'll skip 1440p because it was cpu rainbow six siege at 4k has the fcw3 about 4.4 percent ahead of the 390 fp card that's a bigger difference than we've seen elsewhere but the fps range is also wider and so differences become less significant overclocking got us to 218 fps average and that's again about 2.4 percent so depending on how you dice the numbers up uh it's either not too exciting or we'll move on in hitman 2 the rtx 390 fcw3 ran 114 fps average for the ocb bios a 4.3 gain over the stock fe card the normal v bios like in rainbow 6 is technically below the ocb bios score that's expected but it maintains a rank above the 30 90 fe single card sli doesn't do anything here and once again you'd have to look at our previous testing for more of that and as for the overclock 3090 ftw3 that one ran 116 fps average a gain of 2.2 percent over the stock ocv bios not particularly worth it in this one unfortunately so conclusions then well if it's fifteen hundred dollars and you're definitely buying a 30 90 anyway then it's worth it in a sense uh the fe card is actually called well enough that it's not like the previous reference series designs where we would say buy literally anything except for the reference design and a blower car like anything else would be an upgrade for the most part but in this case the fe card for the 3090 is actually comparatively very very good it's competitive now the evga card was pulling about 40 watts higher and because of its v bios increase that's generally regarded as a good thing if you're not really familiar with it because it gives you a higher power target so for things like overclocking or for just native stock boosting you're going to boost higher and it does it boosts in the range of maybe 15 to 30 megahertz higher depending on when you take the measurement and in what application the extent to which that is useful is obviously limited it's it's not like that number is going to get you a ton for stock versus stock performance it is technically better a lot of people buy for the best this isn't necessarily that but it's going to get you pretty close without getting into class of say uh kin pin card territory if we're also talking ebg or hall of fame territory from galax something like that so the ftw3 if it were price equivalent with the fe we would probably recommend over the fe for a few reasons but one of them is that it does have technically superior noise normalized performance particularly when you're looking at the power consumption so if you were to power normalize it even better but it's also more likely for the f23 you'll be able to at some point find it unlocked to be bios on a forum somewhere if you're into the overclocking scene if you're more of a blender 3d artist type of user game engine user it's technically better in some aspects it's just that the extent to which it's better feels a lot more limited than past generations the card's smaller than the fe so that's a potential upside but it's again it's it's just it's weird right now to stand here and say that an fe or a reference card is not utter garbage that has never really been the case at least not for a long time radon 7 was okay but uh this is much different so that's kind of where they stand relative to each other eighteen hundred dollars or whatever plus minus a hundred bucks doesn't really matter at this point at eighteen hundred dollars or seventeen or nineteen hundred dollars it is really hard to justify unless you're specifically interested in the overclocking stuff if you're a professional user you're probably interested in things like the memory capacity you might need to buy two of the same partner model if you can't get two fees and you want to run an sli bridge because keep in mind the fe bridge it's the the bridge placement the sli fingers are not standardized that was true last gen 2 and probably end generations before that but because they're not standardized don't expect to buy an fe card with an ftw3 because you're not going to be able to do it not in a case anyway on an open bench where you can freely position the cards on the x-axis sure but that is a limitation to consider you you'll have to buy two of some card if you want to use sli to make it easiest so for pure overclocking it is worthwhile on that aspect we were it's a little upsetting how limiting the v bios is from a power perspective we'd like to see maybe 50 more watts at least out of it more load balancing between the pcie slot and the cables would be great too if possible but the definitely power limited shunt resistor mods can fix that it's not particularly fun to do this generation there's a lot more shunts on the cards that's an nvidia level thing not just evga but it applies to them too and of course other than all of this this is all under the assumption that you are buying a 30 90 because you know for your use case it is actually genuinely very useful they exist 3d artists again our own andrew coleman who works for gn does a lot of 3d art and for him he wants a 30-90 because he works with scenes that are very high in memory requirements on the video card and he doesn't care about the titan rtx features or the higher end driver features that were turned off this time so if you're like andrew and you need that kind of card then well here's your review but otherwise if you're just thinking 30 90 for gaming in general check our original 30 90 review and that'll give you a good idea of what we think about that generally not worth it but there are people who just want the best to use the common quote we see and it's not something we'd entirely fault but we would encourage at least researching a bit and seeing if maybe the best might be something cheaper and then putting the delta of money towards something else unless at all unless you're jeff bezos in which case well i don't know why you're watching our stuff but i'd appreciate a free prime membership thanks for watching you can go to store it on camerasex.net to help us out directly or you can go to patreon.com subscribe for more we'll see you all next time youthe desk is an absolute mess because we've been using the card that we're reviewing today in an overclocking competition with jay and with paul from paul's hardware and you can check that content out separately but before we got into any of the overclocking and the mods that we're working on coming up like shunt resistor mods soldered onto the card we ran all the normal review numbers so that we could produce this look at the evga rtx 3090 fcw3 there is not technically an msrp that we've received yet and because this card is impossible to find anywhere on retail sites we're not 100 sure what it's supposed to cost but we have some ideas and we'll talk through today the value of the already strained 90 gpus when applied to a higher end partner model before that this video is brought to you by ek and the ekaio series we recently reviewed the ekaio drgb 360 as one of the top performers in our cpu cooler charts advantage from its fan performance and high quality pump internals that we found during our tear down if you're looking for a high performance rgb cpu cooler with longer steady state times check out the ekaio 360 and 240 at the links in the description below so as a reminder with partner models of any video card what you're really looking for is quality of life improvements and when we say that what we really mean is things like noise normalized thermals so who has the most efficient cooler when you set them to the same noise level not same type of noise necessarily but same noise volume who has the best thermal result who's got the highest sustained clocks as a result of the gpu boosting behavior where it's contingent upon thermals the the extent to which it boosts and then you're also looking at stuff like fan spin down some of the older reference designs didn't have that but partner models did now it's pretty much everywhere uh things like the thermistors that evga places on its cards so the fcw3 cards have icx thermistors those are ntc thermistors so it's like a smaller thermocouple basically it's placed in a few spots on the board and you can get a temperature measurement out of it these are not particularly actionable it's an extremely nice marketing point but there are a few places where they are useful so we find them useful for more extreme overclocking even for water cooled overclocking like putting the board under a water block which it's difficult to find for something like an fcw3 but they are coming out but the reason it's useful there is because you can just sort of mentally verify that you do have good contacts the memory to the mosfets things like that you don't have to really wonder about it and worry if you're burning something up so for extreme overclocking it's useful because when we're trying to figure out do we need to run a memory heater or is the memory running too hot without direct contact cooling we can look at those thermistor numbers and figure out okay well we're not really anywhere close to the 110 degree tj max so we're fine so that's kind of what they're useful for daily life using it out of the box no modifications not much overclocking to speak of the thermistors the the nine thermal sensors on the board are not really that useful but again we like them for other features outside of the stock usage so keep that in mind it might be useful to you it might be utterly useless but something kind of neat to look at and that's kind of about the end of it as for the rest of the card it looks like this will probably be in the range of 1800 we haven't gotten official confirmation if we get it before this video goes up but after it's edited we'll we'll put it right here on the screen but we're reviewing this under the assumption that it's about 1800 and i'll give you some alternative prices and where i think it stands towards the end of the video we don't expect many people will be buying 3090s but for those who are it'll be useful to look at how the partner models compare to the uh comparatively overbuilt versus previous generations 39d founders edition so this will give you a good look versus the fe card is the first 39d partner model reviewing don't know if we're going to review more but in the very least you get a look versus fe and where it lands so without further delay let's get into the thermal numbers acoustics and gaming performance frequency behavior overclocking behavior all that stuff and then if you want additional information on how our testing is conducted you can check our gpu testing methodology piece which we'll link in the description below let's get into it power testing's up first looking at averages and then we'll look at more depth in a moment the rtx 3090 fe ran 356 watts and a full load of furmark overclocking the fv got it to 408 watts making it one of the least efficient gains we've seen in a long time and contrary to nvidia's claims that you can clock this up somewhat efficiently the ftw3 starts higher at 384 to 391 watts and runs up to 464 watts with the fully unlocked shipping v bios we'd still like more power available for serious overclocking though the normal v bios is called lands 390 watts so there's basically zero change from the so-called ocv bios we'd also expect minimal changes in gaming benchmarks so this should theoretically mostly impact fan speeds based on what we're seeing here if you're curious a gaming workload reflects similar behaviors total war 3 kingdoms at 4k ultra lines up almost exactly the same way with everything in the range of 400 watts before any overclocking looking at the per rail power draw we observed the pcie cables at about 160 watts 135 watts and 115 watts the pcie slot draws a little bit over the official pci sig spec in this initial testing and hits 80 watts somewhat consistently it should be closer to 75 watts max this isn't much of a concern but if you were to run multiple of these anyway say three in a water-cooled config for multi-gpu processing you'll definitely want an auxiliary power into the motherboard to take some of the load off of the 24-pin cable and prevent any melting keep in mind that you can't sli three of them but you can still run three plus of them without a bridge for workstation applications where it might not really benefit from a bridge the power is just high on the slot overall and considering the 12 volt cables still have plenty of room if possible it makes sense to try and shift the load there we have a lot of thermals to cover but we'll start with the fan response to temperature under auto settings and as controlled purely by v-bios this is the out of the box noise test in this testing the rtx 3090 fe ramps slowly over a five minute period to about 37 dba and is more variable in its fan ramp over time the slower ramp period is the nature of having so much surface area and metal and it takes longer to reach steady state and soak as a result the ftw3 ramps faster kicking fans into gear sooner and also ramping faster the 3090 fcw3 levels out at about 39 dba at a 20 inch distance when fully auto that marks its auto fan speeds as a bit louder than the 3090 fe just within territory of being noticeable but that doesn't mean its cooler is any worse or any better we need to look at noise normalized thermals to determine the efficiency of each cooler before though we move into noise normalized thermal testing let's map the rpm to the noise that we just saw gpu temperature target on the fe card using a fire strike looping render landed at about 68 degrees celsius the evga 3090 ftw 3 hit 65 under automatic v bios configurations and note that again this is what the ambient controlled and monitored every second of the test at about 21 degrees celsius this is not a delta t measurement it's just the temperature the fan rpm for the cards is not directly comparable either because the fan count and the placement is different however the evga cards spun up to about 1900 rpm once at steady state thermals become tricky when normalized for noise anyway we equalize the fan noise levels so the cards produce the same noise but they aren't producing the same work the power is not normalized as we saw previously and the extra 40 watts allowed on the evga 3090 fcw3 means that even a noise normalized test is not truly a test of purely the coolers the 3090 ftw3 therefore has a limitation but we can still evaluate it with another metric the rtx 3090 fe cooler maintained about a 10 degree advantage at 40 dba with a 57 degrees celsius baseline versus about a 67 degree result for the gpu only on the fcw3 from this line plot alone the fe looks better and we can see how you would think that however if we look at the watts per degree celsius dissipated the evga ftw3 runs 5.8 watts per degree celsius while the fe cooler runs 6.4 the evga cooler is technically coping better when proportionately aligned for the power load but the fe gpu cooler is undoubtedly extremely competitive here and running cooler when considering its lower power target this next plot is one of memory thermals as measured at 40 dba for starters we'll look at the ntc thermistors included on the evga icx board and then we'll look at manually placed thermocouples these ntc thermistors are placed adjacent to some memory modules and we'll show the image of that now and if you're curious to learn more about these you can learn about it in our original review of the icx cards back when they launched in the 10 series these three ntc thermistors plotted around the same temperature across the modules which is good there's even cooling across the memory so the load is evenly distributed the result was about 70 degrees celsius for all three measurements there was one erroneous spike at around the 510 second mark the spike was to 6543 degrees celsius and is obviously incorrect the precision px1 software bugged out somewhere in the process of reading the thermistors because to our knowledge our lab did not in fact exceed the temperature of the sun during testing thus far the memory temperatures are well within spec even accounting for the delta between the inside and outside of the memory casing tj maxx is 110 degrees so 70 to 80 is a good place to be to verify the accuracy of these sensors and the software other than that bug we attached our own thermocouples to a hotspot top memory module and a memory module near the slot which often runs warmer because of its presence near the motherboard for this chart we averaged the icx thermistors to clean up the charts the average is just below 70 degrees our bottom gddr6x module measured about 66 degrees celsius as tested on the back of the pcb where the flip chip package is closer to the thermocouple as opposed to through the casing the top module measured about 73 degrees celsius averaged they come out to 70. so it's nearly identical to the icx results this is great news for two reasons first the icx sensors appear accurate but there is no electrical reason they wouldn't be so this isn't really a surprise unless evga were doing something screwy in software thermistors aren't an evga invention they're extremely common it'd be hard to screw them up although it is possible secondly the more important one is that it means our own testing is accurate so even without the interior junction temperature being exposed to us via software we can still produce numbers that are reliable and test it a delta of maybe about five to ten degrees celsius could safely be added for junction internal temperature this chart is from the same data but for the power components this includes five power measurements from icx and one from our own vrm mosfet thermocouple we placed this on the exact center thermocouple and we determined which one would be hottest by taking some thermography we'll plot these line by line since it's kind of messy the power measurements heat up to low 60s at best with the bulk of them in the range of 65 to 70 degrees and the worst measurement at about 78 degrees celsius our own vrm mosfet measurement plots last and is a much more stable line thanks to the placement being less sensitive and it also plots right in the middle of the pack once again we're mostly internally excited that our own measurements are proving useful here it doesn't really say anything about the card in terms of its usefulness or its quality but we're just happy to see that the numbers we're producing are good all of these measurements are well within spec for the components on evga side and again as a reminder this is 40 dba normalized although auto was running pretty close to this mark anyway we next need to determine the frequency of the cards to evaluate out of the box performance plotted in a frozen frame render and 3d mark the 3090 fe card shown in blue averaged in the range of 1890 to 1905 megahertz the evga f230cv bios flattened out around 1935 megahertz so it's holding about 30 megahertz higher than the stock fe some of this comes from the boosted stock power consumption the normal v bios for whatever that means plots right between them although it makes this chart completely illegible a ride to the middle and is not significantly different from either we're not really sure that evga meant to do it this way it seems kind of like a screw-up in the v-bios programming but they're mostly the same as each other as we get into games just a reminder that we're keeping a truncated list of games for partner models that's because the performance doesn't change much card to card so if you want the full scope of the 3090 performance we'd recommend checking our initial 3090 review which also has 8k benchmarks for partner cards the focus is mostly on thermals it's also on noise and overclocking but we'll give some basics for games here anyway for shadow of the tomb raider the evga rtx 3090 ftw3 plotted a 106 fps average for the ocv bios and didn't really change much for silent it ended up about 1.3 ahead of the stock 390 fe overclocking got us to 111 fps average and that's with a lighter oc we could hold 1200 megahertz memory and plus 80 megahertz core offset for both of those in competitive overclocking but it was tough to keep that stable in all games this config we're showing was stable though the uplift from the oc is 5.2 percent in average fps in red dead 2 with dx12 the evga 3090 ftw 3 plotted at 92.5 fps average for the ocv bios or 91.2 for the silent v bios or normal as they call it the results flank the rtx 3090 founders edition the sli results as a reminder don't improve here because the dx12 version of this game doesn't support sli vulcan does but that's in a previous video if you care about us a lot there's basically no difference here overall between the fcw3 and the fe and overclocking the 3090 ftw3 gets a 21 fps average which is an improvement that's noteworthy of about nine percent over the stock ftw3 and that's actually very good and we also found that changing the memory tune here benefited the card significantly so perhaps red dead 2 at least with the x12 is maybe a little bit more sensitive to the memory increases than to core increases although we haven't fully validated this that's kind of our takeaway from the initial data so far overall though it's worth overclocking these cards sometimes in single card configurations we'll skip 1440p because it was cpu rainbow six siege at 4k has the fcw3 about 4.4 percent ahead of the 390 fp card that's a bigger difference than we've seen elsewhere but the fps range is also wider and so differences become less significant overclocking got us to 218 fps average and that's again about 2.4 percent so depending on how you dice the numbers up uh it's either not too exciting or we'll move on in hitman 2 the rtx 390 fcw3 ran 114 fps average for the ocb bios a 4.3 gain over the stock fe card the normal v bios like in rainbow 6 is technically below the ocb bios score that's expected but it maintains a rank above the 30 90 fe single card sli doesn't do anything here and once again you'd have to look at our previous testing for more of that and as for the overclock 3090 ftw3 that one ran 116 fps average a gain of 2.2 percent over the stock ocv bios not particularly worth it in this one unfortunately so conclusions then well if it's fifteen hundred dollars and you're definitely buying a 30 90 anyway then it's worth it in a sense uh the fe card is actually called well enough that it's not like the previous reference series designs where we would say buy literally anything except for the reference design and a blower car like anything else would be an upgrade for the most part but in this case the fe card for the 3090 is actually comparatively very very good it's competitive now the evga card was pulling about 40 watts higher and because of its v bios increase that's generally regarded as a good thing if you're not really familiar with it because it gives you a higher power target so for things like overclocking or for just native stock boosting you're going to boost higher and it does it boosts in the range of maybe 15 to 30 megahertz higher depending on when you take the measurement and in what application the extent to which that is useful is obviously limited it's it's not like that number is going to get you a ton for stock versus stock performance it is technically better a lot of people buy for the best this isn't necessarily that but it's going to get you pretty close without getting into class of say uh kin pin card territory if we're also talking ebg or hall of fame territory from galax something like that so the ftw3 if it were price equivalent with the fe we would probably recommend over the fe for a few reasons but one of them is that it does have technically superior noise normalized performance particularly when you're looking at the power consumption so if you were to power normalize it even better but it's also more likely for the f23 you'll be able to at some point find it unlocked to be bios on a forum somewhere if you're into the overclocking scene if you're more of a blender 3d artist type of user game engine user it's technically better in some aspects it's just that the extent to which it's better feels a lot more limited than past generations the card's smaller than the fe so that's a potential upside but it's again it's it's just it's weird right now to stand here and say that an fe or a reference card is not utter garbage that has never really been the case at least not for a long time radon 7 was okay but uh this is much different so that's kind of where they stand relative to each other eighteen hundred dollars or whatever plus minus a hundred bucks doesn't really matter at this point at eighteen hundred dollars or seventeen or nineteen hundred dollars it is really hard to justify unless you're specifically interested in the overclocking stuff if you're a professional user you're probably interested in things like the memory capacity you might need to buy two of the same partner model if you can't get two fees and you want to run an sli bridge because keep in mind the fe bridge it's the the bridge placement the sli fingers are not standardized that was true last gen 2 and probably end generations before that but because they're not standardized don't expect to buy an fe card with an ftw3 because you're not going to be able to do it not in a case anyway on an open bench where you can freely position the cards on the x-axis sure but that is a limitation to consider you you'll have to buy two of some card if you want to use sli to make it easiest so for pure overclocking it is worthwhile on that aspect we were it's a little upsetting how limiting the v bios is from a power perspective we'd like to see maybe 50 more watts at least out of it more load balancing between the pcie slot and the cables would be great too if possible but the definitely power limited shunt resistor mods can fix that it's not particularly fun to do this generation there's a lot more shunts on the cards that's an nvidia level thing not just evga but it applies to them too and of course other than all of this this is all under the assumption that you are buying a 30 90 because you know for your use case it is actually genuinely very useful they exist 3d artists again our own andrew coleman who works for gn does a lot of 3d art and for him he wants a 30-90 because he works with scenes that are very high in memory requirements on the video card and he doesn't care about the titan rtx features or the higher end driver features that were turned off this time so if you're like andrew and you need that kind of card then well here's your review but otherwise if you're just thinking 30 90 for gaming in general check our original 30 90 review and that'll give you a good idea of what we think about that generally not worth it but there are people who just want the best to use the common quote we see and it's not something we'd entirely fault but we would encourage at least researching a bit and seeing if maybe the best might be something cheaper and then putting the delta of money towards something else unless at all unless you're jeff bezos in which case well i don't know why you're watching our stuff but i'd appreciate a free prime membership thanks for watching you can go to store it on camerasex.net to help us out directly or you can go to patreon.com subscribe for more we'll see you all next time you\n"