1660Ti Reviews Thread

[LTC8K6]

Well, where's the 1680ti then?

1650
1650ti
1660
1660ti
1670
1670ti
1680
1680ti

NV has this on the 1660ti page:

16: The New
Supercharger

 

[Ranulf]

Yea, AMD falls even further behind in performance per watt (and pretty much every other metric as well). Bring on the obligatory "power consumption doesnt matter" from the AMD camp.

The only reason "power consumption doesn't matter" exists is because people use it to justify their argument any time that is the main metric their new beloved product wins out. I've seen it go both ways and its been annoying for the last 4+ years. Any one on this forum who buys an OC factory clocked video card or overclocks anything in their PC obviously doesn't care that much about perf per watt except as a way to justify a product based on the fact it will let you overlock it to eek out more perfomance or run your games well while staying quiet and cool.

If you care about perf/watt so much you'd never overlock your PC and you'd damn well have led light bulbs in every fixture in your house because you'll save far more in energy costs there than worrying about power draw from your PC components.

 

[NTMBK]

The only reason "power consumption doesn't matter" exists is because people use it to justify their argument any time that is the main metric their new beloved product wins out. I've seen it go both ways and its been annoying for the last 4+ years. Any one on this forum who buys an OC factory clocked video card or overclocks anything in their PC obviously doesn't care that much about perf per watt except as a way to justify a product based on the fact it will let you overlock it to eek out more perfomance or run your games well while staying quiet and cool.

If you care about perf/watt so much you'd never overlock your PC and you'd damn well have led light bulbs in every fixture in your house because you'll save far more in energy costs there than worrying about power draw from your PC components.

It's not necessarily about cost- for me it's mostly about noise. I'm not interested in having a wind tunnel under my desk! Hence the RX 480 is the highest TDP card I've ever owned.

 

[amenx]

The only reason "power consumption doesn't matter" exists is because people use it to justify their argument any time that is the main metric their new beloved product wins out. I've seen it go both ways and its been annoying for the last 4+ years. Any one on this forum who buys an OC factory clocked video card or overclocks anything in their PC obviously doesn't care that much about perf per watt except as a way to justify a product based on the fact it will let you overlock it to eek out more perfomance or run your games well while staying quiet and cool.

If you care about perf/watt so much you'd never overlock your PC and you'd damn well have led light bulbs in every fixture in your house because you'll save far more in energy costs there than worrying about power draw from your PC components.

Not really about energy costs, but HW owners care more about heat and noise. GPUs with higher power draw require beefier cooling solutions that become more difficult to keep quiet the faster the fans spin.

 

[VirtualLarry]

From an "AMD fan" (but not to the extreme of fanboy), I have to admit, this seems like a very compelling value proposition, and is probably going to be my next card. Well, as long as they are going to detect my 4K UHD 40" Avera TV properly, and not downgrade the colors, like my GTX 1050 3GB and GTX 1060 3GB cards do. That's one (very important!) reason that I've stuck with my RX 570 cards on these rigs for so long. (The other has been a lack of value propositions for new cards in the market. Well, until now.)

https://www.anandtech.com/show/13973/nvidia-gtx-1660-ti-review-feat-evga-xc-gaming/2

Reading that review / arch. overview of TU116, what really stands out to me, is that they added the ability for concurrent Int and FP within the SM, which, reminds me a lot of the GTX 460 launch, and how they "tweaked" Fermi, in that GPU, to allow for (some) super-scalar operations, much like TU116.

And, if it means anything, GTX 460 was a monster hit of a card. Hopefully, the GTX 1660 (ti) will be as well. Looking forward to trying one out, hopefully I can afford one in a couple of weeks.

 

[Guru]

At about 35% faster than a gtx1060 6gb/rx580 the 1660ti is a good upgrade for........
Gtx1060 3gb owners 45% faster.
Rx570 43% faster
R9 290 owners 45% faster
Gtx970 owners 42% faster
Gtx 780ti owners 45% faster
Gtx780 owners 65% faster
7970,7950,rx280,rx285 owners 85% faster
Gtx680,670,760,770 owners 95% faster.
Gtx960 4gb, gtx1050, 1050ti, gtx950 owners 100% to 120% faster.

That's 18 cards to upgrade from .

Where are you getting this absurd 35% increase? It's on pair with the GTX 1070, it beats it in some games(Vulkan titles, as Pascal was crap at Vulkan), but losses in others. Yeah it's essentially matching the 1080 in Wolfenstein 2 and beating the 1070ti in rainbow seige, but it also loses to the 1070 in GTA 5, Witcher 3, etc....

Again all of the benches show the 1070 about 25% faster than the 1060 6GB, since the 1660ti is essentially equal to the 1070 overall across most games, its 25% faster than the 1060 as well.

Also the 1060 3GB is not 10% slower than the 1060 6GB, its more like 5-6% slower on average, in some games the difference is as little as 2%. According to hardware unboxed who test usually 30+ games the different between the 3GB and 6GB 1060 is about 5%.

About being a good upgrade for a 1050/ti is also absurd. Why would anyone buy such a card just 2 years ago or so, probably at about $170, only to spend $280 1 or 2 years later? If they had the money back then or better sense, they would have went for the RX 470/570 or just gotten a GTX 1060 3 or 6GB.

No one with a 1050 will upgrade to a 1660ti, no one. Its out of their price range and the performance difference is big, but not that big for that segment of buyers. It would have to be SAME performance as the 1660ti is now, but at $180 or less. This market segment doesn't magically have $100 more to buy this and 75% to 90% performance improvement for 2 years is also not enough in this range.

 

[VirtualLarry]

No one with a 1050 will upgrade to a 1660ti, no one. Its out of their price range and the performance difference is big, but not that big for that segment of buyers. It would have to be SAME performance as the 1660ti is now, but at $180 or less. This market segment doesn't magically have $100 more to buy this and 75% to 90% performance improvement for 2 years is also not enough in this range.

I think that you might be surprised.

What is your opinion on my possible purchase, of 1-2 of these cards, to replace RX 570 4GB and 8GB cards? Foolishness, or money well spent?

Edit: And what's the over/under on two-fan models versus one fan? Do these cards really "need" two fans? (Like, my 7950 800Mhz core cards, from HIS, had a single side fan in the center, and that was hardly enough cooling, to run two of those cards in one rig, Folding or whatnot.) I made a big mistake back then, buying a single-fan card, I don't want to make that mistake again. Then again, my two GTX 1060 3GB cards, each have one fan, and they don't seem to have too much in the way of thermal problems.

Where's the cut-off point, for needing two fans versus one fan? Would that EVGA XC card with one fan be an acceptable purchase, at $279? Or should I pony up $309 for a dual-fan? Also, how is Gigabyte for card quality? I heard "some things" about them, not all positive.

As one of the cards lower down the RTX 20 and now GTX 16 series stack, the GTX 1660 Ti XC Black also lacks LEDs and zero-dB fan capability, where fans turn off completely at low idle temperatures.

Maybe I won't get the EVGA XC card, it's 2.5 slot width, and lack of 0db fan-stop modes, seem to indicate that that card might not be the wisest purchase. My RX 570 cards have the fan-stop feature, when they're below 55C.

 

[Guru]

I think that you might be surprised.

What is your opinion on my possible purchase, of 1-2 of these cards, to replace RX 570 4GB and 8GB cards? Foolishness, or money well spent?

Edit: And what's the over/under on two-fan models versus one fan? Do these cards really "need" two fans? (Like, my 7950 800Mhz core cards, from HIS, had a single side fan in the center, and that was hardly enough cooling, to run two of those cards in one rig, Folding or whatnot.) I made a big mistake back then, buying a single-fan card, I don't want to make that mistake again. Then again, my two GTX 1060 3GB cards, each have one fan, and they don't seem to have too much in the way of thermal problems.

Where's the cut-off point, for needing two fans versus one fan? Would that EVGA XC card with one fan be an acceptable purchase, at $279? Or should I pony up $309 for a dual-fan? Also, how is Gigabyte for card quality? I heard "some things" about them, not all positive.

1660ti is about 35% faster than a RX 570. Again depending on what game you look at, it could be up to 45% faster on something like GTA 5 or only 25% faster on a game like Modern Warfare or Resident Evil 5.

What does a 35% average faster performance translate to in FPS in games? Well its a wide ranging number, but on average it would be something like 30fps difference. If the RX 570 is running Division 1 at 50fps, 1660ti is going to run it at around 80fps or so.

Is your gaming suffering with a RX 570 or can you still play games at high settings and 60fps? You don't need to max every single detail, even the RX 580 and 1060 6GB can't do it in newer games. The 1060 6GB used to be able to max every single game an run it at 60+fps, but a lot of the newer games it averages lower than 60fps.

Do I think the 1660ti is overall a good upgrade for the RX 570? Absolutely NOT! I think a RTX 2060 is a lot more worthwhile "upgrade", it offers about 60% faster performance over the 570, +/- 10% more/less performance depending on game and it only costs $70 more. So if you have $280 right now(most partner cards are around $300 actually), you'd be better off saving for a month or two to $360/70 and purchasing a RTX 2060. Its still a better value and it still comes with 1 free game, so that is essentially $60 dollar value you get back.

 

[sontin]

The 1660TI is more than 50% faster than a RX570: https://www.techpowerup.com/reviews/EVGA/GTX_1660_Ti_XC_Black/28.html

 

[NTMBK]

No one with a 1050 will upgrade to a 1660ti, no one. Its out of their price range and the performance difference is big, but not that big for that segment of buyers. It would have to be SAME performance as the 1660ti is now, but at $180 or less. This market segment doesn't magically have $100 more to buy this and 75% to 90% performance improvement for 2 years is also not enough in this range.

Not necessarily. If a teenager in 2016 was building their first gaming PC, they might have got a 1050 and a Skylake i3. Now two and a half years later they don't want to build a whole new system, but upgrading to a bigger GPU is an easy way to improve performance (especially if they just got a nice cheap 4K monitor). Plus the 1050 is still current, meaning you can resell it and put the money towards that bigger GPU.

 

[rainy]

What does a 35% average faster performance translate to in FPS in games? Well its a wide ranging number, but on average it would be something like 30fps difference. If the RX 570 is running Division 1 at 50fps, 1660ti is going to run it at around 80fps or so.

Sorry, in that case it would be 60 percent higher performance of 1660 Ti.
If you want 35 percent gap, then you have 50 vs 67.5.

 

[tajoh111]

Where are you getting this absurd 35% increase? It's on pair with the GTX 1070, it beats it in some games(Vulkan titles, as Pascal was crap at Vulkan), but losses in others. Yeah it's essentially matching the 1080 in Wolfenstein 2 and beating the 1070ti in rainbow seige, but it also loses to the 1070 in GTA 5, Witcher 3, etc....

Again all of the benches show the 1070 about 25% faster than the 1060 6GB, since the 1660ti is essentially equal to the 1070 overall across most games, its 25% faster than the 1060 as well.

Also the 1060 3GB is not 10% slower than the 1060 6GB, its more like 5-6% slower on average, in some games the difference is as little as 2%. According to hardware unboxed who test usually 30+ games the different between the 3GB and 6GB 1060 is about 5%.

About being a good upgrade for a 1050/ti is also absurd. Why would anyone buy such a card just 2 years ago or so, probably at about $170, only to spend $280 1 or 2 years later? If they had the money back then or better sense, they would have went for the RX 470/570 or just gotten a GTX 1060 3 or 6GB.

No one with a 1050 will upgrade to a 1660ti, no one. Its out of their price range and the performance difference is big, but not that big for that segment of buyers. It would have to be SAME performance as the 1660ti is now, but at $180 or less. This market segment doesn't magically have $100 more to buy this and 75% to 90% performance improvement for 2 years is also not enough in this range.

The GTX 1070 is closer to 35% faster than a GTX 1060 than 25%. 25% is really low and doesn't reflect the change in difference in specifications.

The GTX 1070 has 50 percent more shaders than GTX 1060 which easily translates into more than 25% more performance. How much more? somewhere in the 35-40% region overall.

https://www.computerbase.de/2019-02/geforce-gtx-1660-ti-test/3/#diagramm-performancerating-2560-1440

39% more performance.

Techpowerup also shows a 38% difference in performance

https://www.techpowerup.com/reviews/MSI/GeForce_GTX_1660_Ti_Ventus_XS/28.html

https://www.anandtech.com/show/13973/nvidia-gtx-1660-ti-review-feat-evga-xc-gaming/16

Anandtech is showing a 37% increase in performance when the GTX 1660 ti is compared to the GTX 1060 6gb.

35% is a conservative number and 25% is just outright false.
As the techpowerup reviews show, the GTX 1660 ti is already 59 percent faster than a RX 570. You don't need an RTX 2060 for 60% better performance vs an RX 570.

One more correction is the different in performance between a GTX 1050 ti and GTX 1660 ti is bigger than 70-90%, the techpowerup shows this difference is 127-134% overall or 2.27x the performance. Compared to the GTX 1050 this difference rises to 197% which is triple the performance.

While the GTX 1660 ti is significantly more expensive than a gtx 1050 ti, the performance difference is even larger which makes the GTX 1660 ti a better buy than the GTX 1050 ti.

That's not to say a RX 570 is not a good buy. It's a great buy but the GTX 1660 ti is in a way different tier of performance.

 

[tviceman]

It doesn't have the same die size as the chip it replaces (GP106) though.

It's about 35% faster than the 1060, but also 42% bigger.

The closest die size wise is the GP104 at 314 (10% bigger than TU116), but that die is about 20% faster (as seen with the 1080).

In other words Turing is an ok improvement with regards to efficiency (roughly 33% better than GP106 and 15% better than GP104), but with regards to perf/mm2 it is at best a wash, and at worst a regression (unless of course the TU116 as seen in 1660 Ti is not fully enabled).

It's on the same process and has a 33% efficiency improvement. That's nearly as good as Maxwell's jump. Compare this with AMD's Vega 7 vs. Vega 60, and Nvidia actually made a better improvement in perf/w than AMD. Maxwell had a sizeable jump in die size for it's efficiency improvements on the same node, too.

Turing sans RTX is a great improvement over Pascal no matter how you try to spin it. Compared again to Vega VII's jump over Vega 64, which had an entire node increase to use for efficiency improvements, TU116 it actually did a better job at improving efficiency over GP106, which was already an extremely efficient design.

 

[tviceman]

Where are you getting this absurd 35% increase? It's on pair with the GTX 1070, it beats it in some games(Vulkan titles, as Pascal was crap at Vulkan), but losses in others. Yeah it's essentially matching the 1080 in Wolfenstein 2 and beating the 1070ti in rainbow seige, but it also loses to the 1070 in GTA 5, Witcher 3, etc....

https://www.techpowerup.com/reviews/MSI/GeForce_GTX_1660_Ti_Ventus_XS/28.html - 40% faster
https://www.anandtech.com/show/13973/nvidia-gtx-1660-ti-review-feat-evga-xc-gaming/16 - 36-37% faster
https://www.techspot.com/review/1797-nvidia-geforce-gtx-1060-ti/ - 38-40% faster

 

[StinkyPinky]

You wouldn't want to do that.

Why not? It's 40% faster, has more vram, and is a newer architecture so will likely be supported longer.

 

[happy medium]

Where are you getting this absurd 35% increase? It's on pair with the GTX 1070, it beats it in some games(Vulkan titles, as Pascal was crap at Vulkan), but losses in others. Yeah it's essentially matching the 1080 in Wolfenstein 2 and beating the 1070ti in rainbow seige, but it also loses to the 1070 in GTA 5, Witcher 3, etc....

Again all of the benches show the 1070 about 25% faster than the 1060 6GB, since the 1660ti is essentially equal to the 1070 overall across most games, its 25% faster than the 1060 as well.

Also the 1060 3GB is not 10% slower than the 1060 6GB, its more like 5-6% slower on average, in some games the difference is as little as 2%. According to hardware unboxed who test usually 30+ games the different between the 3GB and 6GB 1060 is about 5%.

About being a good upgrade for a 1050/ti is also absurd. Why would anyone buy such a card just 2 years ago or so, probably at about $170, only to spend $280 1 or 2 years later? If they had the money back then or better sense, they would have went for the RX 470/570 or just gotten a GTX 1060 3 or 6GB.

No one with a 1050 will upgrade to a 1660ti, no one. Its out of their price range and the performance difference is big, but not that big for that segment of buyers. It would have to be SAME performance as the 1660ti is now, but at $180 or less. This market segment doesn't magically have $100 more to buy this and 75% to 90% performance improvement for 2 years is also not enough in this range.

It looks like half the community answered you ,so I'll just agree with them.
Thanks guys.

 

[coercitiv]

Turing sans RTX is a great improvement over Pascal no matter how you try to spin it. Compared again to Vega VII's jump over Vega 64, which had an entire node increase to use for efficiency improvements, TU116 it actually did a better job at improving efficiency over GP106, which was already an extremely efficient design.

@antihelten wasn't spinning anything - in terms of die size and transistor count TU116 is equivalent to GP104. If that's all there is to TU116 then perf/mm2 and power efficiency aren't actually higher than Pascal, but I personally don't believe that to be true as I'm reasonably convinced there's a 256bit product to be had from TU116 (or at least a stronger 192bit one).

Even with the additional resources per core and increased caches, dropping from a max of 2560 units and a 256bit interface to somewhere close to 1536 units and 192 bit interface doesn't really add up considering similar die area, transistor count and the slightly denser 12nm process. Something close to 2048 shading units and a 256 bit interface for an uncut TU116 seems to fit much better with what we know so far.

I wonder how many of the 1660Ti boards will have 2 empty VRAM slots from the start.

 

[amenx]

Well, as long as they are going to detect my 4K UHD 40" Avera TV properly, and not downgrade the colors, like my GTX 1050 3GB and GTX 1060 3GB cards do. That's one (very important!) reason that I've stuck with my RX 570 cards on these rigs for so long. (The other has been a lack of value propositions for new cards in the market. Well, until now.)

With 4k (TVs only, not monitors), Nvidia cards default to RGB limited. Easily solvable by going to the NV CP resolution tab and selecting Nvidia colors/RGB Full.

 

[Arkaign]

Any news on 1660 non Ti?

 

[jpiniero]

Any news on 1660 non Ti?

Videocardz says Early March.

 

[Guru]

https://www.techpowerup.com/reviews/MSI/GeForce_GTX_1660_Ti_Ventus_XS/28.html - 40% faster
https://www.anandtech.com/show/13973/nvidia-gtx-1660-ti-review-feat-evga-xc-gaming/16 - 36-37% faster
https://www.techspot.com/review/1797-nvidia-geforce-gtx-1060-ti/ - 38-40% faster

I'm sorry to break it to you, but that is utter garbage!

1. Your numbers are way off, you didn't even copy-paste it right, techpowerup shows the 1660ti 29% faster over the GTX 1060 6GB.

2. Techpowerup tested the 1060 6GB with a completely different setup, they used 6700k, different mobo, different memory and used the FE edition of the 1060 6GB. There is no comparison between the two tests.

3. Same thing with techspot, used different setups when testing the two cards, so the direct comparisons are not comparable.

Do they rebench all these cards when making new reviews? I don't think so, benching even 1 card at such short time periods is challenging enough, benching 10 cards is impossible.

 

[tajoh111]

I'm sorry to break it to you, but that is utter garbage!

1. Your numbers are way off, you didn't even copy-paste it right, techpowerup shows the 1660ti 29% faster over the GTX 1060 6GB.

2. Techpowerup tested the 1060 6GB with a completely different setup, they used 6700k, different mobo, different memory and used the FE edition of the 1060 6GB. There is no comparison between the two tests.

3. Same thing with techspot, used different setups when testing the two cards, so the direct comparisons are not comparable.

Do they rebench all these cards when making new reviews? I don't think so, benching even 1 card at such short time periods is challenging enough, benching 10 cards is impossible.

Your doing the math wrong.

You cannot simply subtract the percentage away from the normalized 100% performance, you have to divide 100 by the percentage of the GTX 1060 represents in that graph.

E.g 100/71 = 1.408 or 40.8% for example.

That is if all performance was normalized against the GTX 1060 where the GTX 1060 was 100%, the GTX 1660 ti would be 40.8% faster.

This is doubly confirmed by the performance per dollar graphs. Note the GTX 1060 and GTX 1660 ti have near equal performance per dollar graphs with the GTX 1060 price at 200 and the GTX 1660 ti price at 280 dollars.

How much more expensive is the GTX 1660 ti over the GTX 1060 6gb? 40% more expensive. I.e $200 * 1.40 = 280 dollars. According to the performance per dollar graphs, the GTX 1660 ti has slightly better performance per dollar vs the GTX 1060? Why is this the case? in percentage terms, your getting 40.8% percent more performance for 40% more money hence the double confirmation.

Your simply wrong. Accept it.

It is not just techpowerup that is showing this percentage in improvement.

It's anandtech, computerbase.de, techspot.

Anandtech and computerbase definitely use the newest drivers and considering the performance increase is the same range, that is 36-40%, it provides strong statistical evidence in the accuracy of those other tests.

Your not going to get this 25% increase in performance unless your willing to limit game selection to 2 or 3 games for a review and select the worst titles out of the bunch. And this would be statistically inaccurate because error increases with smaller sample size and your intentionally try to sample bad performance which means data collection is biased and does not reflect reality.

Do not try to say your right any further. You will simply make a bigger fool of yourself logically while making yourself look incompetent when it comes to basic math and statistics.

 

[tviceman]

I'm sorry to break it to you, but that is utter garbage!

1. Your numbers are way off, you didn't even copy-paste it right, techpowerup shows the 1660ti 29% faster over the GTX 1060 6GB.

2. Techpowerup tested the 1060 6GB with a completely different setup, they used 6700k, different mobo, different memory and used the FE edition of the 1060 6GB. There is no comparison between the two tests.

3. Same thing with techspot, used different setups when testing the two cards, so the direct comparisons are not comparable.

Do they rebench all these cards when making new reviews? I don't think so, benching even 1 card at such short time periods is challenging enough, benching 10 cards is impossible.

Basic Math (100 - 71)/71 = .408 = 40.8%. The GTX 1660 TI is 40% FASTER than the GXT 1060.

But let us go a step further and do quick research on just how well the 1060 performs today vs. the review numbers you are refuting. Here is a GTX 1060 6gb GDDR5X review from December 17th 2018. https://www.techpowerup.com/reviews/KFA2/GTX_1060_6_GB_GDDR5X/31.html
Compared to the 2080 at 1400p in that review, the 2080 is 2.27x faster than the GTX 1060. On December 17th, approximately 2 months ago.

If we take that performance of the GTX 1060 and plop it into the GTX 1660 review, the GTX 1060 would wind up as 73% on that chart.So here is the math of the most recent GTX 1060 review (from 2 months ago) vs. the GTX 1660.

GTX 2080 score from GTX 1660 review at 1440p: 166% faster than a GTX 1660
GTX 2080 is 2.27x faster than GTX 1060.
166% / 2.27 = 73%

Sooo, with an updated score of a more recent and capable GTX 1060, the GTX 1660 is still (100-73)/73 = 37% faster.

 

[SMOGZINN]

We can argue all day about the price/perf ratio and if this is technically a good card at 1080p resolution, but the end result is that most of us are more than a bit disappointed. What we really wanted was a card that could do 4k at reasonable rates at a reasonable price and we did not get anything resembling that. I've been playing at 1080p for 3 years now, why do I need a new card that does it just a little bit better? Like the rest of the Turing this just is not an upgrade card, this only really makes sense if you are building or buying a new system, and even then it probably makes more sense to buy a Pascal card in your price range.

 

[Mopetar]

We can argue all day about the price/perf ratio and if this is technically a good card at 1080p resolution, but the end result is that most of us are more than a bit disappointed. What we really wanted was a card that could do 4k at reasonable rates at a reasonable price and we did not get anything resembling that.

Are you talking about Turing as a whole? I don't think anyone was reasonably expecting the 1660 Ti to be a capable 4K card.