TomsHardware--GTX 1180 Expected in July

[xkm1948]

https://www.tomshardware.com/news/nvidia-turing-faq,37067.html

1080Ti still is some of the best choice until 1180Ti comes out.

 

[jboydgolfer]

My post "dissapeared"

Id guess either may or sept based on top tier gpu's past release dates.

 

[yotano211]

How is it possible for a July release when DDR6 memory wont start production until July.

 

[cucker tarlson]

I'm counting on 1170 matching 1080Ti, or at least within 5%. 1080Ti is a 250W card, overclocked it's hot and loud. 1170, if a 150-170W card, will run at max OC like a champ on air while staying quiet.

 

[xkm1948]

I'm counting on 1170 matching 1080Ti, or at least within 5%. 1080Ti is a 250W card, overclocked it's hot and loud. 1170, if a 150-170W card, will run at max OC like a champ on air while staying quiet.

Probably not gonna happen. RTG offers little competetion. I would be surprised if 1180 matches 1080Ti performance. 1170 would probably just be at 1080 level.

Only advantage of 1170 over 1080Ti would be hardware accelerated ray tracing tech IMO

 

[Vya Domus]

This will simply be a product stack shift like the second generation Kepler.

1180 ~ = 3584 shader GP102 equivalent
1170 ~ = 2560 shader GP104 equivalent
...

And so forth , everyone expecting more than that within a margin of error of lets say 10% is going to be thoroughly disappointed.

 

[Fluffmeister]

How is it possible for a July release when DDR6 memory wont start production until July.

Samsung announced they started mass production of GDDR6 back in January. Equally a month later SK Hynik confirmed impending availability.

 

[John Naylor]

Probably not gonna happen. RTG offers little competetion. I would be surprised if 1180 matches 1080Ti performance. 1170 would probably just be at 1080 level.

Only advantage of 1170 over 1080Ti would be hardware accelerated ray tracing tech IMO

1070 was faster than 980 Ti ... and 970 was about the same as the 780 Ti ... so not out of the realm of possibility

https://www.techpowerup.com/reviews/MSI/GTX_1070_Gaming_Z/26.html
https://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/27.html

 

[xkm1948]

1070 was faster than 980 Ti ... and 970 was about the same as the 780 Ti ... so not out of the realm of possibility

https://www.techpowerup.com/reviews/MSI/GTX_1070_Gaming_Z/26.html
https://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/27.html

Nvidia had some competetion back then. Now they have no real competetion at all. Vega was a major flop and Polaris is no longer competitive. It would really surprise me if Nvidia decide to make another big leap in terms of GPU performance.

 

[rtwjunkie]

I'm counting on 1170 matching 1080Ti, or at least within 5%. 1080Ti is a 250W card, overclocked it's hot and loud. 1170, if a 150-170W card, will run at max OC like a champ on air while staying quiet.

That’s not usually how Nvidia works. Normally one model down in the new cycle matches one model above from previous cycle. Only rarely does it deviate. So more than likely you will be ponying up for an 1180 to match or exceed 1080Ti.

 

[yotano211]

Samsung announced they started mass production of GDDR6 back in January. Equally a month later SK Hynik confirmed impending availability.

oh thanks, I thought I read somewhere that it started in June or July

 

[Caring1]

I'm not expecting any great leaps, just a smaller node process and newer tech.
I'm hoping power consumption drops.

 

[cucker tarlson]

That’s not usually how Nvidia works. Normally one model down in the new cycle matches one model above from previous cycle. Only rarely does it deviate. So more than likely you will be ponying up for an 1180 to match or exceed 1080Ti.

970 matched 780Ti, 1070 matched TX and outperformed 980Ti, so this is absolutely how it works. 980 beat 780Ti, 1080 beat 980Ti by +25-30% (talking all stock)
I get why you'd think that if that was a refresh, on same uarch, process and memory. But given it's new uarch,refined process and new memory nvidia would be stupid to stay at 1080Ti performance level. I don't think anyone at nvidia thinks 1080Ti perofrmance as the aim for 1180. Vega 7nm with 4 stacks of hbm2 and features that were supposedly there on vega is launching late this year. They'd be morons to hamper 1180. +15% over TXp is the bare minimum for me to even give a crap about 1180.

 

[rtwjunkie]

970 matched 780Ti, 1070 matched TX and outperformed 980Ti, so this is absolutely how it works. 980 beat 780Ti, 1080 beat 980Ti by +25-30% (talking all stock)
I get why you'd think that if that was a refresh, on same uarch, process and memory. But given it's new uarch,refined process and new memory nvidia would be stupid to stay at 1080Ti performance level. I don't think anyone at nvidia thinks 1080Ti perofrmance as the aim for 1180. Vega 7nm with 4 stacks of hbm2 and features that were supposedly there on vega is launching late this year. They'd be morons to hamper 1180. +15% over TXp is the bare minimum for me to even give a crap about 1180.

No, it’s not how Nvidia works. Only one level up beats previous one level below. Never once did I mention the word refresh. You even started out wrong. I had two identical systems in the house with 780 on one and 970 in the other. The 970 by the smallest of margins managed to outperform the 780. So forget a 780Ti. 1070 vs 980Ti was an aberration so far. But let’s call the top TPU Nvidia people, regardless of what the points say: @Vayra86 and @Fluffmeister for a ruling, shall we?

 

[Vayra86]

Rulings? You give me too much credit lol.

But yes @rtwjunkie is very much correct. Maxwell > Pascal was the exception to the rule, or rather, an event you don't see on a yearly basis. Its also the reason Nvidia got away with the inflated MSRPs with Pascal; the performance jump was inflated too.

If you go back to Kepler this is what's what:

670 > 760 = not entirely equal (670 was faster in most situations; 760 is handicapped both in VRAM and shader count)
680 > 770 = very tiny jump forward due to faster memory but generally equal
770 > 960 = mostly equal (though the 770 was still slightly faster as it has more bandwidth, minimal gaps depending on game/engine, again the 960 was a bit of a dud, returning 'feature' of x60 really)
780 > 970 = mostly equal
780ti > 980 = mostly equal - Both 780 and 780ti versus both 970 and 980 are same-die cuts versus full enabled...
970 > 1060 = significant jump of 20-25% for the 1060 6GB
980 > 1070 = significant jump of 25-30% for the 1070

For a blanket statement, I usually stick to ~30% performance increase per gen per card tier. And we know there is generally a 30% performance gap between each product in the stack, so the logical conclusion is that yes, in a general sense, GPUs 'move a single tier' each gen.

GPU performance development is really quite boring. The pattern is clear; we can already roughly determine that the 1180 will match a 1080ti.

 

[trog100]

1070 was faster than 980 Ti ... and 970 was about the same as the 780 Ti ... so not out of the realm of possibility

https://www.techpowerup.com/reviews/MSI/GTX_1070_Gaming_Z/26.html
https://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/27.html

no it wasnt i have owned (and overclocked both) and the performance is very similar.. the 1070 uses a lot less power than the 980TI bit but they go about the same.. i swapped my pair of 980TI cards for a pair of 1070 cards purely because of the better power efficiency with less heat generation..

trog

 

[Tonim89]

Rulings? You give me too much credit lol.

But yes @rtwjunkie is very much correct. Maxwell > Pascal was the exception to the rule, or rather, an event you don't see on a yearly basis. Its also the reason Nvidia got away with the inflated MSRPs with Pascal; the performance jump was inflated too.

If you go back to Kepler this is what's what:

670 > 760 = not entirely equal (670 was faster in most situations; 760 is handicapped both in VRAM and shader count)
680 > 770 = very tiny jump forward due to faster memory but generally equal
770 > 960 = mostly equal (though the 770 was still slightly faster as it has more bandwidth, minimal gaps depending on game/engine, again the 960 was a bit of a dud, returning 'feature' of x60 really)
780 > 970 = mostly equal
780ti > 980 = mostly equal - Both 780 and 780ti versus both 970 and 980 are same-die cuts versus full enabled...
970 > 1060 = significant jump of 20-25% for the 1060 6GB
980 > 1070 = significant jump of 25-30% for the 1070

For a blanket statement, I usually stick to ~30% performance increase per gen per card tier. And we know there is generally a 30% performance gap between each product in the stack, so the logical conclusion is that yes, in a general sense, GPUs 'move a single tier' each gen.

GPU performance development is really quite boring. The pattern is clear; we can already roughly determine that the 1180 will match a 1080ti.

- As a 780 Ti owner, I can state that GTX 980 is superior than GXT 780 Ti, not by a huge amount, still by a relevant margin. Same goes for 780 vs 970.
- GTX 770 is a rebrand of GTX 680, with higher clocks, no real improvement. You can flash one's BIOS into another.

The rest is pretty much accurate.

About the new card, I expect the non-Ti to be superior than the actual 1080 Ti. Anything below that level is worthless the release.

 

[RejZoR]

The thing is, with no competition, NVIDIA becomes its own competitor in a way. If they launch a new product and it's not a significant upgrade to old generation, people won't buy it. There is a small group of people who buy new generations because they already skipped one or two generations, but it's not the same as if everyone jump on them.

 

[cucker tarlson]

No, it’s not how Nvidia works. Only one level up beats previous one level below. Never once did I mention the word refresh. You even started out wrong. I had two identical systems in the house with 780 on one and 970 in the other. The 970 by the smallest of margins managed to outperform the 780. So forget a 780Ti. 1070 vs 980Ti was an aberration so far. But let’s call the top TPU Nvidia people, regardless of what the points say: @Vayra86 and @Fluffmeister for a ruling, shall we?

No,I'm sure it's you who got all the facts confused. 970 is barely faster than 780 is your own observation, based on comparing two rigs,which is not a very accurate method. What matters here is 780Ti was never that much faster than 780 to begin with. 1080Ti is like ~25% faster than 1080 on average, and 780Ti was not 25% over 780, not even close. More like 15%. So if 970 manages to beat 780 enough for you to notice on two different rigs then yes, it means it's at 780Ti's toes.
This was also a totally different situation back then. Why would they settle at 25% faster than 1080 ? They'd be sore losers in that situation. +2 years since 1080, new architecture, new memory, new (refined) process. There's just no way 1180 is 1080Ti performance and 1170 is 1080. For Christ's sake they've even got a 1070Ti at 1080 performance, and that's the same line of cards. 1170 at 1080 perfomance would be a pascal refresh,and a pretty measly one at that considering what I just said about 1070Ti.

 

[rtwjunkie]

@cucker tarlson it pains me, because I am just going to have to block you. See, this topic is one you a) are completely blind on, and b) you continue to misconstrue my own words to suit your NVIDIA agenda on releases, which I might add, you are the lone proponent of.

I never said equals. I said one level up matches or exceeds one level down. While you are at it, go read some frickin reviews and find out it was the 980 that beat 780Ti l, not 970. I gave personal experience so you could see some real world comparison instead of faceless reviews.

As for 1070Ti beating 1080, if overclocked and 1080 is at stock. Dafuq kind of comparison is that? Lol.

 

[cucker tarlson]

I said: "970 matched 780Ti ", "980 beat 780Ti ", "they've even got a 1070Ti at 1080 performance ". You keep insisting I said 970 beat 780Ti and that I'm comparing 1070Ti OC vs 1080 stock.

It's stupid to argue about the performance of a card that we know nothing about anyway.

 

[Vayra86]

Between 970 and 980 is at best a 10% perf gap just like it is between the 780 and 780ti. The situations in which the 970/980 excels are those that heavily push VRAM both in capacity and bandwidth; but the 780 and 780ti do have a stronger core. In older games the Keplers win, in more recent ones, the Maxwells win. Similarly in games that use heavy tesselation the Maxwells go in the lead.

About 1070ti and 1080, in a few years time those two will be in a similar dituation with regards to VRAM /bandwidth where the 1080 will last just a tad longer and produces a minor 5-10% perf advantage.

Both are bad examples of the 'generational leap"' of 30% per card; for different reasons; in the case of the 780ti it was due to a vibrant competition on the high end and with the 980 it was the flagship/halo product effect that was supposed to sell it. In reality, most people jumped on either the 970 or the 780, not the overpriced alternatives.

But in a general sense the idea still stands; last gen's performance drops one tier in the stack.

 

[blued]

No, it’s not how Nvidia works. Only one level up beats previous one level below. Never once did I mention the word refresh. You even started out wrong. I had two identical systems in the house with 780 on one and 970 in the other. The 970 by the smallest of margins managed to outperform the 780. So forget a 780Ti. 1070 vs 980Ti was an aberration so far. But let’s call the top TPU Nvidia people, regardless of what the points say: @Vayra86 and @Fluffmeister for a ruling, shall we?

You seem completely out of the loop. Check 970 vs 780ti performance across all resolutions. The last 2 gens (Maxwell and Pascal) have been x70 = or beats Ti previous gen. Regardless of your personal experience with your 780, 970 I will take TPU and the other many sites as the more believable reference.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1070/24.html

 

[rtwjunkie]

You seem completely out of the loop.

You don’t know a thing about me. Let’s keep it that way.

 

[Divide Overflow]

Someone's still frequenting Tom's Hardware?