TPU: nVidia readies 1660TI

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
I can see a cut-down 1660 Ti going for $230, but surely a 3 GB GDDR5 card would be less than $229. That card would struggle even now in some games..

The rumors are for the 1660 (non Ti) to have 6GB of GDDR5, not 3GB.

I hope 3GB is a thing of the past except for they lowest end media cards.

If they want to cut it down more, they could go for 4GB or even 5GB has been done before IIRC.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Out of curiosity, if performance is close to GTX1070 why get the 6GB 1660ti at $279 over the 8GB GTX 1070 currently found at newegg at $299 ???

Well, since it isn't even officially announced yet, it is hard to say exactly where performance falls, and if there will still be GTX 1070 cards when the 1660Ti ships.
 

happy medium

Lifer
Jun 8, 2003
14,387
480
126
Well, since it isn't even officially announced yet, it is hard to say exactly where performance falls, and if there will still be GTX 1070 cards when the 1660Ti ships.

its easy math, just figure out how much faster a Touring core is than a Pascal core.

rtx 2060 1920 cores vs gtx 1660ti 1280 cores =80% or 20% slower.
They said the 1660ti will have higher clocks than the 2060 So I'll use a 2060 overclocked model.
It should be right around a gtx1070 on this chart 2060 100% vs 1660ti at 80%


a.png
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
Honestly I think the 1650 will be 1024 cores ,4 clusters cut, 4gb ,128 bit
,is that even possible ? :] That would be very close to rx 470/480 speeds.

The 1650 (or whatever is below the 1660) may end up being a different chip. There's a TU117 that was also in an earlier rumor/leak about the product stack.

TU116 isn't going to be a very large chips since it doesn't have the extra ray tracing hardware that the 2000 series chips have. Couple that with a mature process and I don't think they'd naturally have many chips where they'd need to cut 4 clusters to get a functional part.

An 8 cluster chip does seem like the next logical stepping point though. I've no idea what they'll call it, but it would be the ~1030 replacement as well.

The rumors are for the 1660 (non Ti) to have 6GB of GDDR5, not 3GB.

I'm just going off of the data from videocardz article (https://videocardz.com/newz/nvidia-gtx-turing-teased-by-leaked-photograph) that you posted. The 1660 is listed as having "6GB/3GB GDDR5".
 

linkgoron

Platinum Member
Mar 9, 2005
2,286
809
136
Because you pay for hardware that is not usable by 99% of current games.

If you're paying 80% 2060 for 80% performance 2060 without RTX, you're actually getting less features for your money. The die will probably be significantly smaller without RTX and only 1536 cores (assuming 1660ti is full TU116), so even the "more money because of die size" nonsense won't be relevant.
 

Furious_Styles

Senior member
Jan 17, 2019
492
228
116
IMO there is Zero chance of that happening. NVidia is all in on RT. Lower end cards are the exception because functional RT HW would just blow the transistor budget in that chips in that price range.
I don't think it's a zero chance, definitely a low chance though. I can still dream at least!
 

maddie

Diamond Member
Jul 18, 2010
4,722
4,624
136
its easy math, just figure out how much faster a Touring core is than a Pascal core.

rtx 2060 1920 cores vs gtx 1660ti 1280 cores =80% or 20% slower.
They said the 1660ti will have higher clocks than the 2060 So I'll use a 2060 overclocked model.
It should be right around a gtx1070 on this chart 2060 100% vs 1660ti at 80%


View attachment 2708
You wrote:
"rtx 2060 1920 cores vs gtx 1660ti 1280 cores =80% or 20% slower."

How does this calc work? 2/3 rds (66%) of the cores get 80% performance?
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
Because you pay for hardware that is not usable by 99% of current games.

This is odd, because I do recall our back and forth regarding DX12. You were upselling AMD's dominance as far back as Hawaii/Fiji, 2015. You kept citing DX12 performance for a handful of games.

Be consistent atleast. Either new hardware/features maters at launch or it doesn't.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
6,792
5,749
136
If that 19% increase over 1060 is true how can Nvidia get away with calling this a 60 Ti card? Really hoping it's BS and it's the $229 1660 that's only around 20% faster than the 1060.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
This is odd, because I do recall our back and forth regarding DX12. You were upselling AMD's dominance as far back as Hawaii/Fiji, 2015. You kept citing DX12 performance for a handful of games.

Be consistent atleast. Either new hardware/features maters at launch or it doesn't.

He is consistent - AMD good, NVidia/intel bad.

I get the point you are making though...lots of hypocrites in this forum.
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
If that 19% increase over 1060 is true how can Nvidia get away with calling this a 60 Ti card? Really hoping it's BS and it's the $229 1660 that's only around 20% faster than the 1060.

Like you say - very unlikely. It would make no sense and leave them with a huge performance gap to the 2060.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
6,715
7,004
136
The naming certainly leaves something to be desired but again these cards are very interesting from an academic standpoint to see what the Turing arch sans RT cores is capable of doing (is it even the Turing arch at that point?)

I suspect these will be very potent little cards, only question that's left is if the price will be right (I say it too much maybe, but GTX 980TI performance for $300 2.5 years after I got a 980Ti for ~$300 is again not very inspiring).

$300 price point should really be somewhere between the 1080/Ti in terms of raw performance if we where on a standard cadence...
 

mohit9206

Golden Member
Jul 2, 2013
1,381
511
136
The naming certainly leaves something to be desired but again these cards are very interesting from an academic standpoint to see what the Turing arch sans RT cores is capable of doing (is it even the Turing arch at that point?)

I suspect these will be very potent little cards, only question that's left is if the price will be right (I say it too much maybe, but GTX 980TI performance for $300 2.5 years after I got a 980Ti for ~$300 is again not very inspiring).

$300 price point should really be somewhere between the 1080/Ti in terms of raw performance if we where on a standard cadence...
You got 980Ti for 300 used but most bought it at full price of 700 dollars so it is not an apt comparison.
 
  • Like
Reactions: happy medium

AtenRa

Lifer
Feb 2, 2009
14,000
3,357
136
This is odd, because I do recall our back and forth regarding DX12. You were upselling AMD's dominance as far back as Hawaii/Fiji, 2015. You kept citing DX12 performance for a handful of games.

Be consistent atleast. Either new hardware/features maters at launch or it doesn't.

In the past we had new features added for free (no added price per segment). RT cores came with double the price over previous segments (example : GTX 1080Ti vs RTX 2080Ti). So essentially we are paying extra for hardware that is not usable by 99% of games.

Hawaii R9 290X launched at $549 with superior Mantle and later DX-12/Vulkan performance against the GTX780 that was selling for $649. There was no added price tax for features like Asynchronous Compute , Mantle and later DX-12/Vulkan performance.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
6,715
7,004
136
You got 980Ti for 300 used but most bought it at full price of 700 dollars so it is not an apt comparison.

-That's fair, I got the card in the used card sellathon following the launch of the 1080. However the performance was there for the price, although it is not an apples to apples comparison. Those that bought the card for $700 paid a premium to enjoy the card a year and a half longer than I did or simply bought at a very inopportune time (immediately preceeding the launch of the 1080).
 

happy medium

Lifer
Jun 8, 2003
14,387
480
126
Hawaii R9 290X launched at $549 with superior Mantle and later DX-12/Vulkan performance against the GTX780 that was selling for $649. There was no added price tax for features like Asynchronous Compute , Mantle and later DX-12/Vulkan performance.
The 290x should have been priced at $349 , it was hot ,power hungry and a midrange card. The gtx980 used half the power was much faster for the same $550. For $150 more you could have waited for the much much faster 980ti.
See how this works? There is always something better cheaper on the horizon.
Mabe we should stay on topic and not talk about 5 year old cards.
 
  • Like
Reactions: tviceman

mohit9206

Golden Member
Jul 2, 2013
1,381
511
136
The 290x should have been priced at $349 , it was hot ,power hungry and a midrange card. The gtx980 used half the power was much faster for the same $550. For $150 more you could have waited for the much much faster 980ti.
See how this works? There is always something better cheaper on the horizon.

Not always. 2080 and 2080Ti proved this theory wrong. Better but not cheaper.
Even better is debatable in case of 2080.
 

AtenRa

Lifer
Feb 2, 2009
14,000
3,357
136
The 290x should have been priced at $349 , it was hot ,power hungry and a midrange card. The gtx980 used half the power was much faster for the same $550.

At the time of launch the R9 290X was competing with the GTX 780 and not the GTX 980.

R9 290X launched October 2013

GTX 980 launched September 2014
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Do we even know what Navi is supposed to be? Is it supposed to be the successor of RX570/580 or Vega 56/64?

570/80 replacements. Basically next gen 7nm console development lifted over to the PC market.

As such reasonably likely to be sensibly competitive in the low /mid range market. Hopefully they can get the power draw lower than polaris.