Overclocking RTX 4070 Ti VRAM to Compensate 192-Bit Bus Width?
Vložit
- čas přidán 14. 06. 2024
- Support me on Patreon:
/ der8auer
---------------------------------------------------------
Music / Credits:
Outro:
Dylan Sitts feat. HDBeenDope - For The Record (Dylan Sitts Remix)
---------------------------------------------------------
Paid content in this video:
- /
Samples used in this video:
- /
---------------------------------------------------------
Timestamps:
0:00 Intro
2:28 3DMark Time Spy Extreem GT1
2:56 Cyberpunk 2077
4:24 Remnant: From the Ashes
5:10 PUBG
6:00 BF2042 Problems
6:36 VRAM Overclocking Benchmarks
7:26 Noise & Power Draw Gigabyte Eagle
8:44 Summary/Conclusion
9:36 Outro - Věda a technologie
GN: Der8auer did a thermal testing of AMD, we didn't have time to do that, go watch him.
D8: GN did a 4070ti testing, I didn't have time for it, go watch them.
You guys should do more cross-overs
Tech without the drama. Genuinely a rare find these days.
Keis
these days ? you only do CZcams now ?
I was wondering all the time about this topic comparing it to my 3070 TI with 256 bit bus! Thanks Roman!
Yes I was surprised they did this too.
Its significantly slower even with max overclock. Know this for a fact without even seeing the video, owned a 3070 Ti.
Is there any article explaining why the 40 series GPUs have narrower bus width? It was way back in GTX 10XX that the 1060 had 192 bit, and the X70 class cards have always had 256-bit width bus (except for the 970 fiasco).
@@namastyle they increased the on die cache (similar to what AMD did with the 6000 series). The issue that AMD ran into, and now nvidia is that no matter how much cache you have, lower bandwidth is still going to limit the amount of data that can be streamed to the card. If the 4080 had a 320bit bus, it’d be much closer to the 4090 than it is right now. I think they only did this because of how negligible a difference there was between a 3080, 3080ti, and 3090. They needed to make a larger gap between the cards to be able to launch skus to compete with AMD/Intel (maybe) that fill in the performance gaps that are between the 4070ti/80/90
IMO though, the 4070ti should’ve for sure been a 4070 with the bandwidth that it has. As I said though, there needed to be a big enough gap to justify the cards. The next round of DX12 UE5 and other games with crazy Raytracing and path tracing is going to force this card into the 1080p with DLSS territory very quickly
@@namastyle They did a similar thing that AMD did by lowering bus width and adding more cache (except NV added more L2$ whereas infinity cache is L3$) in order to increase VRAM capacity. Otherwise if AD104 had 256bit bus it would be limited to either 8GB or 16GB, and we all damn well know NV would chose the cheaper 8GB option. This way they can keep almost the same performance but increase VRAM size to 12GB
Similarly 5700XT was 256bit bus and 6700XT was 192bit (also why 5700XT was a better mining card than 6700XT)
It's a shame the prices are the way they are. Ada really is an impressive architecture. The performance, features, and efficiency are to be commended
Indeed but since people keep buying and Nvidia owning the vast majority of the pro and pc gamer market they have zero incentive to suddenly charge pre-2021 prices again.
Ada isnt really that great.
TSMC is.
@@flimermithrandir that's like saying Fermi 1.0 wasn't bad tsmc was
@@flimermithrandir 😂
@@virtualtools_3021 So would you say Samsungs Node was better than TSMCs Node?
I am curious. Tell me. Enlighten me.
4070 Ti, the GPU I want, from a manufacturer I loathe, at a price I refuse to pay.
Niave fingers crossed for 2024.
fa real tho
In 2024 you gonna pay 1200 dollars for a RTX 5050 Ti
@@SMGJohn
Most likely not, but $600 for 5050 sound plausible!
Just because 4050 will be the most popular middle range GPU at 2023 at price of $500!
😂
@@SMGJohn nVidia charging that and me paying that are very different things.
Maybe the 20 year low in GPU sales during the holiday season with brand new products on the shelf will cause a course correction. My wallet is firmly holstered, and my 3080 is still running great after 2 years.
Absolutely great video! very interesting performance from this card when setting the power target to 70%!
Good for Nvidia if they managed to retain great performance despite cheaping out on the memory bus and VRAM amount. Now it should reflect on the price the consumer is paying, right? Right...
lol
Lol
Cheaping out....lol. it's a 850$ 70 series you clown.....cheaping out....lol
Lol, I’d be shocked if these cards cost even 400$ to manufacture for AIBs. With the corroborated rumor going around that the FE 4080 costs Nvidia a little over 300$ to manufacture, it really seems like they are just laughing their way to the bank with their customer’s money.
@@barryhoggle2354 Read again. It literally means that they cheaped out on production costs and yet they're selling it like a high end product.
Thank you for the video!
Are you still planning on selling the power consumption indicating plugs for the 12+4 pin connectors?
Been strangled by bus limits before from nvidia, Not a problem since i got A770, not saying its perfect, it just works better, (black desert being a bandwidth heavy game, also 3440x1440)
I hope Intel Do his Job. They must bring These Monsters AMD/Nvidia down.
@@m.m.3753 trust me, Intel wants to charge Nvida money as well. If the could manage to catch up and compete you'd have a generation or two before price hikes at best.
15 years ago. I put together a top of the line PC, all the best parts available for around €2500, now to do the same it will cost about €9500-10.000. Just a 4090 today is the total cost of a whole HEDT 15 years ago. GPU pricing in particular is completely bonkers
You`re talking about a life time in pc performance. Completely incomparible.
@@brugj03you're drinking the Kool-Aid.
9:08 bruh what??
I would give you $499 - 599 for that, not $800 - 900.
Does nvidia have a gun pointed at you behind that camera or what? From $399 msrp (1070 Ti) to then $599 msrp (3070 Ti) now to $799.
I've seen so many charts over years of youtube, and then Jay sent me your way. Now I see charts from an actual engineer and they contain wattage usage. You're speaking my language: Efficiency. Thank you. Instantly subscribed.
How are you just seeing this guy? His company produces some of the best enthusiast thermal paste and other related items for CPUs.
@@christophermullins7163 Please see a therapist about your condescending attitude, and a doctor about your adenoids.
@@christophermullins7163 maybe he's just getting into the hobby. plus for a long time searching for der8auer only brought up the german channel so if you didn't know he also did english versions you'd likely skip over it if you didn't understand german.
@@sirmonkey1985 you're right on a lot of points. I'm fairly new to the enthusiastic side, 2019, and there are so many tech youtubers out there. Some people drone on and provide nothing new. Then you find people like Roman thar give me what I'm looking for.
@@MechTechRW If you want pure data, you should look at Gamers Nexus. Steve Burke over there isn't as entertaining to watch, but his data is the most reliable on CZcams.
One mistake: you called the card the "Eagle" model when it's actually the "Gaming" model, which is the next step up in the lineup. The Eagle hasn't been released yet and actually says "EAGLE" across the bottom left front of the cooler.
Noticed this as well. Good catch. The Eagle doesn't have LED fans either.
@@OCX600RR I figured it wouldn't have the LED fans but I couldn't find the product page for it when I posted that comment, which probably went live with the launch.
@@OCX600RR Also, Gaming OC and Aorus Master/Elite come with an additional 1 year warranty when registered within 30 days making it a 4 year warranty vs the 3 you get on every other card on the market (Eagle included). This is a big part of why I bought the Gigabyte Gaming OC 3080 12gb, that extra year warranty.
You are a king amongst men.
How I have missed this channel I don’t know.
🎉🎉🎉
It's good to see the efficiency of these cards and the lack of coil whine. Hopefully the prices will become more reasonable throughout the year
Intel raised their prices of CPUs due to inflation, TSMC/Nvidia/AMD will follow...
when the real lowend 4080 comes out the 4070ti's prices will probably drop, because there isn't a space for what this card was supposed to be considering that this was supposed to be the lowend 4080 we got
@@keitomineproductions4133 hate to break it for ya, but 4070 ti is what was supposed to be low end 4080. you're not getting any lower end 4080, but we may get 4080 ti.
@@veduci22 Um... Intel lowered their prices of their GPU's while massively improving the performance of their ARC cards.
Really appreciate you were showing the ability for saving energy by limiting the power target. Would you recommend just to set the power target to 70% or setting up custom curve for GPU voltage? I tried that with my 2080 Super, but it wasn't very stable in Microsoft Flight Simulator and Star Citizen. However I ordered exactly this card because its very compact and I don't need to switch my case with this "small" card. We will see how it performs in Star Citizen in WQHD. Thanks for your advice in advance.
Power limit % creates its own safer curve, suggest over Manual edits.
SaltySpaceKitten
you never need high frames rates, only lots of mem you need, keep the old RTX card please, it can run multiple 4k TV's etc, getting a cockpit
Don't forget the onboard L2 48MB Cache which makes up for the 192-Bit Bus Width
upgraded from GTX 780 to a 4070Ti, did not regret that purchase.
lol 😆👍
No shit Sherlock.
@@Boogie_the_cat thanks friend! what a kind and insightful comment!
Any idea when the wireview is going to be released? I've been looking to buy one for my system since you first showed it off.
Great review ! But isn't that the "Gaming OC" model from Gigabyte ? Eagle models looks and performs different, as I know.
Yea, I thought I remembered the Eagle version having 'Eagle' Imprinted on it.
I was coming to say the same. The box and the look of the card are consistent with the usual look of the Gaming OC models.
I have this card, bought it in march for little less than €1k because it was the best price to performance card still in my budget. Its basically 3090ti with half the power draw and heat. The heatsink on this thing is huge which means you can tune your fans to start spinning at slightly higher temp if you want totally quiet pc in anything but high intensity gaming. or if you want to run games at 250+ fps.
I needed a 4000 series card to do AV1 encoding in Resolve so I replaced my 3080 TI with a 4070 TI and it is noticeably faster, 3DMArk was 18400 before and now its 22400. Bottom line the memory bandwidth difference is not as limiting for this card as it looks on paper. When prices come down I do want a card with more Vram.
Hold onto that 4070Ti as long as you can, every generation now is going to have additional AI-related features, if you can hold until the 6000 series, you'll be set, but if you buy again now or possibly even in the 5000 series, it might be missing some significant AI-efficient power or architecture to improve ray tracing, scaling, or something we don't even know about yet. Regardless of future unknowns, probably not a good idea to spend a second time in the current-gen at the very least.
but v ram limitation
@@arunachalpradesh399 The cores are so much faster that the vram "limitation" is not a big factor
@@garypranzo9334 thats a temporary satisfaction, the recent trend of new games demanding more v ram will increase more in next year, and even low end 4060 ti 16gb will beat your card . secound problem is the price you paid doesnt deserve 12gb period. 4070 ti is 4070 in real , comparing to die core size
@@arunachalpradesh399 bought this card months ago and completed over 40 video projects for my clients while you are waiting for a 16gb 4060 to play games. The 4070 ti was best option for my work at that time. Will be good until 50 series comes out at least.
So the smaller chip is well balanced with the choked memory bandwidth which really just drives home how well designed the architecture is, but also how much nvidia is overcharging to maintain profit levels with volume of sales dropping.
Hey, brother. For the XTX reference problem, would you consider ripping apart a know good one? I would totally enjoy that. Unfortunately, if it was just due to a lack of fluid, I think it’s possible you may not notice a difference once you open it up.
Great video. One visual remark: Dunno if this is related to PWM or the cyan light by the CZcams plaque is simply faulty, but I can see it flashing and it is annoying.
In Australia the Retail price on the Gigabyte Eagle (As that is the card you tested) is AU$ 1479 for the non OC and AU$ 1599 for the OC version.
Love that gigabyte listened for all feedback considering vram temperatures and now they are one of the best
$1499 for a strix model 4070 ti here in canada.
@Der8auer When are the frames and removal tools going to be ready for the new Ryzen 7000 series sir? We need them!! ♡♡
Interesting that even the lower end models are still having these gigantic cooler systems
Its because of the marketing fiasco. When it was the "4080 (12gb)", you can't have it look drastically different from its supposed 4080 brother, so they're designed to have roughly the same size. During the announcement Nvidia themselves picture the (non-existent FE) card with a 2-slot design. However, AIBs already manufactured the unnecessarily big card for the original release schedule - they aren't going to throw them away.
@@souledgar ah yeah that makes sense, I wonder what the 4070 (non ti) version will look like
Funny how "lower end" still costs $800+. Nvidia is going to ensure the next generation of gamers stick to consoles, very sad.
thats the new normal.
These 200w cards ran just fine with 2.5 slot dual fan coolers though. gets a little hot but they all were.
The performance compared to power draw is amazing on the 40-series.
Just a quick question.
You have grizzly minus pad extreme for sale. I can not find the specs for it on your website?
Minus pad 8 is, well, 8w/mk.
On your site it says the extreme version is 260% better than the minus 8.
Please correct my math, but would that not make the extreme version 20.8w/mk?
Just trying to get confirmation on what the w/mk rating is. Thank you.
Thanks for this!
So Nvidia used a 60 class die in the 4070ti and priced it like an 80 class card....
I have the PNY XLR8 4070 Ti and I love this card. It runs ALL my games, unlike my previous 2070.
Danke fur das video
any ML/tensor benchmarks yet? how does this compare to 3080/3090 in workloads?
I realize you work with hardware, but I was wondering if you had a recommendation for anyone to watch in regards for aftermarket tuning software on new AMD GPU. It seems the upper limit on the GPU is astronomical but memory clock takes a massive hit, and vice versa, due to power distribution through the AMD provided (drivers?bios?). I hypothesize that given appropriate cooling (no intended to be a joke... given the recent hotspot discussions, aftermarket cooling solutions seem absolutely necessary anyway), the appropriate software replacement these cards can be beastly... no warranty, of course, but beastly nonetheless.
i still have the rx 5700xt issues and it is worst now if my game crashes then i get blue screen and i gotta install windows from scratch so should i wait for 4060 ti ?
What's your cable for this card ? It looks like to be a 90° adapter but I didn't find one... 9:19
I remember cards used to have 512 bit bus, the last card I had with that configuration was the 560Ti, I wonder why they stopped using the high width bus... I wonder if it would be a benefit for any modern card.
You're misremembering. The 560 ti was definitely not 512 bit. I think the GTX 285 was the last 512 bit gaming card from Nvidia and that was 15 years ago. ATI/AMD did have the hd2900 XT, then the 290x/390x more recently but even those were a very long time ago.
Memory has gotten much denser and faster to the point where the die space cost isn't worth it. Memory interfaces don't scale in size with new process nodes like logic does so there is a strong pressure to reduce bus width over time.
Thanks for this
So it is sufficient eg in Msi Afterburner to lower PL bar to 70% for more or less the same performance as stock??
Are your power measurements for games the entire system power or just PCIe and 12-pin combined?
It’s just the GPU power draw
What is this 12pin 90 degree power adapter you are using?
It's available at MSRP in Poland, and there are no problems with buying them, some of them are still available. Bought one (Gigabyte Eagle OC) and I am really excited to see it in my pc.
so the bus is okay? i know id definitely want more vram for higher res/hz gaming in the future. but like you said, its really a 1440p card.
If I add the two memory chips missing do I get the 256bit bus and 16gb?
Thank you for avoiding the histrionics that have the other youtubers acting like jilted teenagers. Just the facts, ma'am. ;)
Last time i remenber 192 bit is a 60 class bandwidth
What card should I get for my 3440x1440 120hz monitor? I have a Ryzen 7700x
The main factor I use to evaluate a gpu is the Bus width. The 3060ti was way more similar to the 3070 than the 3060 because they both had the same bus whereas the 3060 has a reduced bus. Ignore the name and check bus width first then look at shader performance.
I have a 3060ti OC LHR and I'm very glad I went for the 8gb at 256 bit rather than the 12gb at 192.
the 3060ti and 3070 also use the same GA104 chip. Bus width isn't the only determining factor, keep in mind for example the 3080 12gb got an increase from 320-bit to 384-bit but they perform about the same. It just so happens that bigger and more powerful cores tend to be given more bandwidth to play with because they're more prepared to use it.
In Romania, prices for GPUs are always worse than other parts of the world. Can't find a 4070 Ti for less than 1000 EUR here.
Also, the cheapest 7900 XT is 1130 EUR here.
3070ti with 256 bit GDDR6X is basically the older brother of the 16GB 4080
The 12GB 3060 is the older brother of the 4070TI, the main difference being the switch from 12GB GDDR6 to G6X, and the 100w OC, this card should cost around $260 for Nvidia to produce a founder's edition, and should retail for around $429 and 479 for partner cards
Both cards have a 250-300mm (60 class) GPU die
Both cards have a 192 bit (60 class) memory bus
When compared to the 3060TI,
3060TI: 392mm die, and 256 bit memory bus
4070TI:294mm die, and 192 bit memory bus
The 4070TI costs about the same to manufacture as the 3060TI which is why i suggest a $429 price point
"12GB 3060 is the older brother of the 4070TI" - also i have the 12gb 3060, so you're saying the 4070ti isn't worth the upgrade? can you explain? thanks
That's complete rubbish the 12gb 3060 is alot less powerful than a 3060ti let alone a 4070ti. I think your getting your info wrong
@@YeshuBenPandera He’s only talking about the Silicon and memory controller specs. Obviously this is way better than a 3060 with over double the performance. It is however, over double the price. So…. 🤷🏻♂️
@@HiTechReviewsUKI Probably the most informed non-insider.
Let me start this out by saying Performance does not set the classification of the card, the hardware does
The 3080 did not get re-named the 3060 after the 40 series launched, performance has, and never should be what sets the name and price of a card.
The class of card is set by the size of the die, the width of the memory bus, and the cost to manufacture, and this card costs about $260 to manufacture which makes it cheaper than a 3060TI to manufacture
the 3060TI has a 392mm die and a 256 bit memory bus
The 4060, i mean the 4070TI is 294mm and a 192 bit memory bus, this is basically the same as the 3060 non TI having a 276mm die and the same 192 bit bus.
The main reason the 4070TI should cost $449 instead of $349? The 100w OC and switch from 12GB GDDR6 to 12GB of GDDR6X
Again performance compared to previos cards has nothing to do with the class of card or price because
THATS WHAT NEW GENERATIONS OF CARDS DO
Give you more performance, at the same size and price
The problem?
They're trying to sell a 60 class card, for 80TI class price.
If this, lets face it, stupid argument were actually used, why isnt the 4070TI priced at $200,000? Its about that much faster than the GTX 8800 Ultra
@@YeshuBenPandera Sorry, i tried to make a short quip instead of making a long detailed comment .Basically this is a 4060 nonTI because the 4070TI has the same die size and memory bus as the 3060 non TI, Yes it is much faster, but thats what new generations of cards normally do(they just dont normally triple in price).
The 4070TI has a ~5% larger die than the 3060 and the same memory bus
When compared to the 3060TI, the 3060TI has a 30% larger die, and a 30% larger memory bus than the 4070TI.
Again, yes the 4070TI is faster, as it should be even if it were called a 4060, the problem here is that Nvidia is charging $799-$1049 for a card that should cost $429, because this card only costs them $230-$270 to manufacture slightly cheaper than the 3060TI to manufacture because of the higher power usage, more expensive node, and the switch from GDDR6 to GDDR6X though a reduction in the number of memory chips
If you're wondering why this is so much faster than the 3060TI while being a the same class of card as a 3060 nonTI, well this is because Nvidia jumped from effectively a 12nm node(samsung 8nm), down to 5nm(TSMC 4nm), they basically skipped 2 generations in performance.
If they had continued like normal going down to TSMC 7nm for the 40 series, this is the performance i would have expected from an RTX 5060, but if they had done that, they could have only chaged $400 for this card, Instead, they're trying to trick consumers to beleive this is a 70TI because of the larger than normal performance jump.
Yeah, the pricing, again, will be absolutely insane here in the Baltics....I could not find any 4070Ti GPUs on sale here, but 4080 is being sold at 1500-1600Eur. One reason to hate living here - everything related to PC parts is sold at usually about double MSRP, even years down the line. -_-
not only in Baltics, prices +- the same in the EU, and you could buy what you want in GE and then send it to your country via mail forwarding or event directly
I have had that same BF crash since BF3. Literally ever BF game I get that same crash.
The 3080 was $100 less MSRP with a 320 Bit bus lol.
These prices are bad enough to top the 20 series IMO. At least the 20 series had the excuse of being the first gen of RT. It's a shame because the performance jump really is solid.
Surely that Gigglebyte Gaming OC card does not need a cooler that long...
And surely the RTX4080 doesn’t need Quad Slot coolers with vapor chambers and 350mm lengths. These AIB cards are ridiculous. They basically looked at the MSRP, then looked at the specs, and thought to themselves, “holy shit, we can put like 200$ BOM cost into the cooler alone and still make money!”
i cant imagine the stock to be low when they have had a "release sufficient" ammount since it was the 12 gb 4080 and its been 5(?) months since then and i assume they didnt stop making cards til they figured out what to do. so if stock is low im not sure what the heck they are doing :p
My local did a sale of them recently (when I bought mine) and they sold out in that weekend
I wonder why they haven't released an ITX AIB if the PCB is that small and the GPU is also power efficient
Margins.
Does using DLSS compensate for the lower bandwith? I would assume so, since the actual render resolution is lower, but maybe the chip still needs to communicate more with the memory when upscaling, I don't know.
It does, both reducing memory transfer speed requirements as well as memory usage which might come in handy in the future. I have anecdotal evidence of this while running Portal RTX on a 3070, and the performance with DLSS is about 3x that without DLSS, which points at a major memory issue, in this case not enough VRAM.
@@chuniac7513 Interesting, that's good to hear.
at intro i was thinking there is only cooler ;p Only first fan is over PCB other 2 fans are covering only radiator fins ? This card need such massive cooler ?
From all the benchmarks I have seen is the vram bus even making a difference?
This is the first 4070 Ti video that popped up on my screen. This should be interesting.
Looking at a few first-day reviews, I think that 4070ti might be a not-bad GPU for 1440p, especially for smaller cases. But the main concern is the price for sure.
Exactly, you don't pay 1000€ for a 1440p GPU. That's just ridiculous
@@Gollum1905 paid 909 euros for a 4070 ti eagle. Prices will dip for this card.
@@tensaims what kind of weird flex is this XD you can buy 4070 TI below 900€ which is still way too overpriced for that level of performance. If the RX 7900 XT and 4070 TI go below 750€ I'd still consider buying the 7900 XT. At that price range you're getting into 4k territory and the 4070 is demonstrably worse in 1440p and above. So not sure what your use case is, but even 850€ is a scam for a 4070 TI.
@@Gollum1905 flex? You named a price to illustrate your point and I am telling you there are lower prices for which you can get said card. Calm down nerd.
@@Gollum1905 I bought a 4080 to game at 1440p, why? because it will last me many more years at 1440p and I won't have to upgrade in a year. Saves money in the long run.
... This is WHY the more-unusual & RARER - AMD RADEON Pro WX 8200 … keeps getting BETTER, it does NOT have this issue… LOVE IT!
Man remember when *70 class was 350-400 USD?
You reviewed the Gigabyte Gaming OC 4070ti, not the Eagle as noted during the video and in the charts.
Is it possible to try a 4080 modded BIOS on it somehow. :D
Lol you are already making double content, you really dont need to apologize!
For sure could be a good water-cooling card for how small the PCB is, good for small factor cases looks like
I would not waste a single penny watercooling this thing. For that cost I'd rather have a superior air-cooled GPU. The fact that this thing is so close to a 3080 Ti, yet I can find 3080 Ti on the market that still have two+ years of warranty for $100-200 less, is pretty damn sad
Isn't he talking about the small PCB? Like if you take the cooler off and put a water block on it, you could put it inside a small case, whereas a 4080 or 4090 might not fit, so this would be maximum GPU power for that particular case.
not for the price, if you could buy just the pcb and save money maybe, but nvidia would never do something pro consumer like that
@@kasimirdenhertog3516 I just meant for future reference if the cost comes down but I mean hey you want a powerful card that PCB is the smallest I have seen from Nvidia for the past 2gens.
@Menti Capti OP was talking about using the card in a water cooled SFF build where heat dissipation (and total power use) and physical component size become a major factor. You have a close minded view about what water cooling is used for. Some people just like it for the aesthetics, others like it for the silence, some people do it for the size reduction etc.
Also if you think all of that stuff only costs "a few hundred dollars" I would avoid it, because cheaping out on water cooling parts is a great way to spring a leak and kill your system. Quality parts will be more than a few hundred for a full kit. Just the block alone would be $200+ for a full cover.
Bought one recently at a good price. Cheapest PNY, reference model no rgb bollocks. Still 3 slot and 3 fans cooler tho. That paired with 13600KF -0.09 offset, TUF Z790 D4, 48GB DDR4 3200 and Win 10 on Lexar NM800pro
Power limit only 100%, so not much oc potential there, BUUUUT... it can be quite economic.
On 100% pwr limit in CP2077 it needs between 240-290 Watts(285W msx pwr according to specs) in 3440x1440 Ultra+frame gen+RTultra(no PT) and keeps 60+ frames, with 100+ FPS in some areas/situations.
On 80% limit it needs 220-230W tops, without any performance loss. Saw less fps when below 80% pwr limit. CP is very needy tho.
Most other games, when capped to 60FPS without DLSS consume around 140 Watts(God of War, Horizon, Witcher 3, etc). Which is 60W less than my previous RTX3060 while at stable 60 frames(3060 struggled a lot at that resolution). Even less power with DLSS, which at this resolution is hardly noticable.
Overall it is a really good card. Could be a bit cheaper tho, but I guess those days are far gone now😒
The reason it's not a problem having a 192 bits bus is because of the large amount of l2 cache of 48mb compared to 4mb of the 3070 TI
Der8auer make the LED of power consumption available for every one
Could you please add a 70% PT for the 3090 Ti as well, it doesn't feel fair to compare one undervolted card to a stock card
Not undervolted. Just power limited. If you want to know what a 3090ti will do at 70% PT just look at the 3080ti. It’s the same GPU with a 350W power budget.
"Hey, PUBG might be bandwidth-limited" - proceeds to show memory-OC results for everything but PUBG.
Why?
Been amazed with mine 4070Ti eagle, mostly because I didn't need to upgrade the 650W PSU Nor get a bigger case it was just a swap out the 1080Ti to the 4070Ti.
I play 1440p/165Hz
Oh that makes sense... dont need as big as a PSU as you would with a 3080 ti even.
Do u oc it? What oc if u do?
Those rgb lights around the fans would definitely throw me into a seizure.
8:15 aren't lower clock speeds only because the temperature is higher?
Yep. Silent mode seems to drop the fans by ~3-400 RPM, which means the card is running hotter.
I hope have plans to overclock the Strix 4070ti, using the ElmorLabs EVC2SE. That card has the voltage regulator chip.
"Sitting on my test bench *Right Meow*"
if it had 16gb it would be good.
Availability is insanely good here in norway. But the pricing is tragic, starts from $1100. All the cards are in stock.
good video
Umm isn't that the Gigabyte Gaming OC? The Gigabyte Eagle is smaller, and the box is different.
Can someone tell me why we are only seeing benchmarks for 4K performance while most of the gamers use either 1080p or 1440p?
It would make more sense to do efficiency in watts per frame per second (ie, W/FPS), rather than FPS/W.
1/fps/W = W/fps
annoying but if you want to see those numbers you can do that
I would love to see someone take a 3080 cooler and mount it on this card. This card doesnt need to be this big !
The bus and vram are definitely going to be limitations at 4k, especially for future titles. This card is $200-$300 overpriced.
That's not the Eagle, that's the Gaming (OC)
This is crazy when I can find 6800xt almost half the price
Battlefield 2042 froze my computer the other day and I was using 3080 ti, and I do not know what DICE did.
Hey... I was really hoping you would do a 7900xtx review..
At CES 2023 they said is more faster than 3090Ti, that is very confusing!
192-bit memory bus width? Then it's an RTX 4060 Super then.
4060 will have 128bit or 64bit LOL
@@jaggsta
Exactly! 4060 most likely have 128bit and 4050 that will be the new middle range king will have 64 bit memory bus ant $500 price point!
Cheers your the only review I've seen with 3080 ti results.
you are increasing clock by ~15% but you need 33% longer bus to be on par.
Bought the card day 1 and had no problem at all with BF2042. Actually I don't need a 4080 or 4090 because this card is really good in every game so far!
Playing with ultra settings and right now 128players maps with DLSS set to "Quality" and get 140-150fps. If I sett it to "ultra performance i get 180-200fps so I have nothing bad to say about this card
Yeah @ 1080p basically u will need nothing 😂😂😂....Still living in the stone age.
This is 4K Era...
@@abeerfaisal1986 4k is for losers 1080p is good enough
@spirit7792
1080p is for Beggars 😄😄😄😄😄
Seems a little misleading to compare current gen cards at 70% power target to last gen at 100% power target, would've been nice to see how they compare on an even playing field
The plots have both. They show the card at the 70% PT and default settings.
@@acidnynex Only for the 40 series, that's my point, there's should be 70% plots for both gens for a fair comparison.
This looks like the GAMING OC model and not the Eagle ??