Gtx 1080 benchmarked

From the source:
  • Every single instance of “Founder's Edition” can be replaced with the word “Reference,” using previous-gen nomenclature. There is not one difference in its market positioning. They are synonymous. NVidia has replaced its “Reference” name with “Founder's Edition.”
  • There are not two GTX 1080 models made by nVidia. Only the “Founder's Edition” exists; there is not a heaper card made by nVidia than the $700 Founder's Edition, which ships first.
  • Just to be clear: nVidia is making one official GTX 1080 and one official GTX 1070 model.
  • The “Founder's Edition” is not specially binned.
  • The “Founder's Edition” is not pre-overclocked.
  • The “Founder's Edition” uses the new industrial design and cooler from nVidia. Historically, this is what we would call the “reference cooler.” The cooler is more-or-less identical to the previous reference models. It's got vapor chamber cooling, a VRM blower fan, and a large alloy heatsink under the shroud. There is a backplate on the GTX 1080 Founder's Edition.
  • This card is not "limited edition," despite its name that would indicate as much, and will run production through the life of the GTX 1080 product line.
 
Last edited:
That's what I figured. It's just a money-grab for people that can't wait an extra month.
 
razor1 said:
NVIDIA GTX 1080 & GTX 1070 'Founders Edition' explained | VideoCardz.com

Founder's editions are kinda of a waste, its just the shroud that is changed.
Click to expand...
Like I mentioned it is just a standard reference design, which all the AIBs will sell.
How the heck did every site-publication miss the NVIDIA news brief I have linked a few times that explained that :)

However I think they made a mistake by saying it is only the shroud different between reference and custom AIB.
There is also the blower-vapor chamber cooling design, which may be better quality than in the past.
It could be AIBs will be restricted from doing that type of solution.

I assume quality is probably higher than in the past with respect to cooler-vapor chamber-maybe capacitors used as they intend to sell this reference FE product until 1080 EOL.
Components such as capacitors,etc probably do not change because the reference design is usually the board used by AIBs, but throwing it out there as a possibility.
Cheers
 
I just noted the advertised floating point throughput of the 1070 is low.

I assumed the cut gp104 die would have roughly 80% of the sps like 970:980 that would put it at 2048

Now with that SP count you need only 1586mhz to hit 6.5tflops

So either this is some marketing stupidity calculating with the base clock (which is lower than 1080) or it has fewer sps than I imagined, which would be...

Interesting, but not a particularly pleasant surprise

Either the yields are low and they're disabling many SMs, or they're taking advantage of the absence of competition and 'gating' the performance for the cut die ; they don't want to offer much more than stock Titan X at 379$

Of course it could be clocked low and consequently have a lot of OC headroom, but I don't see that happening given 1080

Thoughts, gentlemen?
 
razor1 said:
I'm thinking the SP counts are at 1920 for the 1070
Click to expand...

How do you think people will react if Pascal has less headroom than Maxwell? It ships at 1733 stock boost, 2000mhz is 15% - 2150mhz is 24%, and if the founders edition cards are high binned cards maybe that will be considered a very high oc

970 - 1178
980 - 1216
980Ti - 1076
TitanX - 1086

Those are stock boost clocks, it can clock up to ~80 mhz higher on it's own, only the Titan X struggles to hit 1500mhz because of reference board design >_>
 
founder's edition are just reference cards.

I think they will still have a similar overclockability to their maxwell counterparts, also nV's partners can make their own designs too.
 
razor1 said:
founder's edition are just reference cards.

I think they will still have a similar overclockability to their maxwell counterparts, also nV's partners can make their own designs too.
Click to expand...

Yeah but for once it was heavily suggested the reference board designed was changed to actually be competitive, and HH on G3D said he understood it to be a kind of high binned card, makes sense for WC, decent VRM performance on par with msi lightning etc etc

980, 970 not uncommon to hit 1550, some cards even go near 1600mhz - on air 31% fir 970, 28% for 980 both 1550

980Ti 1500 on air is not uncommon, on water you can fiddle with 1550+, 39%
 
Well I'm sure it will be all cleared up by the time reviews are out. Everyone is saying different things right now lol. But personally I don't think they are doing anything special with bins.
 
Ieldra said:
Yeah but for once it was heavily suggested the reference board designed was changed to actually be competitive, and HH on G3D said he understood it to be a kind of high binned card, makes sense for WC, decent VRM performance on par with msi lightning etc etc

980, 970 not uncommon to hit 1550, some cards even go near 1600mhz - on air 31% fir 970, 28% for 980 both 1550

980Ti 1500 on air is not uncommon, on water you can fiddle with 1550+, 39%
Click to expand...
Your percentages are way off as you seem to be making the mistake at looking at "advertised" boost clocks. For example a reference 980 ti already goes to 1200 not 1075 like its advertised.
 
Disappointing if this is going to be another 780ti->980ti level upgrade. Apparently the days are gone when a new generation means a 50% bump or more.
 
Xeth said:
Disappointing if this is going to be another 780ti->980ti level upgrade. Apparently the days are gone when a new generation means a 50% bump or more.
Click to expand...

780ti -> 980ti was over a 50% bump I thought? Same with 980 -> 1080... (As far as we can tell...)
 
misterbobby said:
Your percentages are way off as you seem to be making the mistake at looking at "advertised" boost clocks. For example a reference 980 ti already goes to 1200 not 1075 like its advertised.
Click to expand...

Ieldra said:
How do you think people will react if Pascal has less headroom than Maxwell? It ships at 1733 stock boost, 2000mhz is 15% - 2150mhz is 24%, and if the founders edition cards are high binned cards maybe that will be considered a very high oc

970 - 1178
980 - 1216
980Ti - 1076
TitanX - 1086

Those are stock boost clocks, it can clock up to ~80 mhz higher on it's own, only the Titan X struggles to hit 1500mhz because of reference board design >_>
Click to expand...
 
razor1 said:
depends on the overclock?
Click to expand...
If you go to the "Fury Series" category, it says average FPS 48 on the right.
The GTX 1080 category (one result obviously) says average FPS 50.

idk. Looks kinda shit to me but it's hard to interpret.
 
next week I'm sure we will be seeing many of these types of benches lol
 
I don't see Pascal as being so different from Maxwell that Nvidia scrapped the entire way that their basic drivers interface with their cards. Should be Plug n' Play.
 
cageymaru said:
I don't see Pascal as being so different from Maxwell that Nvidia scrapped the entire way that their basic drivers interface with their cards. Should be Plug n' Play.
Click to expand...

This is why Drivers still exist for DX12. The API allows access to more low-level features than before, but it is still DX code running on an API. The drivers still do all the communication between the API and the hardware. The drivers interpret just what the GPU needs to do to perform the action the API asks for.
 
I dug through the Ashes results a bit more and it turns out most of them are mGPU results (doh).
The 1080 being benched there seems to be about 10-15% faster than the other "Fury Series" results. It would probably be a bit higher with a larger sample size, assuming Pelly isn't pushing the card's OC at all (assuming the Fury cards are OC'd).

Here are all the results from the same user.
It seems to be 47~50fps for the GTX 1080 and 41~42 for the Fury (X). Around 15% faster.

 
TaintedSquirrel said:
I dug through the Ashes results a bit more and it turns out most of them are mGPU results (doh).
The 1080 being benched there seems to be about 10-15% faster than the other "Fury Series" results. It would probably be a bit higher with a larger sample size, assuming Pelly isn't pushing the card's OC at all (assuming the Fury cards are OC'd).

Here are all the results from the same user.
It seems to be 47~50fps for the 1080 and 41~42 for the Fury (X). Around 15% faster.

Click to expand...

Furys can OC!?
 
tybert7 said:
Same result, amd cards hold up better over time. Different motivation, makes nvidia look less ruthless and amd less noble, so it makes sense nvidias defenders would lean on that explanation. Either way, the result is the same though. AMD cards hold up better over time, and I fully expect maxwell cards to drop off after pascal unless they too are almost the same architecturally.
Click to expand...

can you say the same about older HD6000 versus GTX500 series? =) don't try to justify things, once they EOL a product they will not be "improving it via drivers" anymore, they only have been improving it's because the architecture has remain unchanged since HD7000.. with the r9 290/390 and R9 290/390X the was even more interesting as those cards already had matured architecture and drivers so they don't have other way to improve that just increase the clocks for both mem and core.. again if you test those cards with their relative competition of GTX700 series in the games of those years you will find that they keep performing the same..
 
My R9 290 gained a ton of performance over the years. Viva la Win 10 drivers.
 
Ieldra said:
I just noted the advertised floating point throughput of the 1070 is low.

I assumed the cut gp104 die would have roughly 80% of the sps like 970:980 that would put it at 2048

Now with that SP count you need only 1586mhz to hit 6.5tflops

So either this is some marketing stupidity calculating with the base clock (which is lower than 1080) or it has fewer sps than I imagined, which would be...

Interesting, but not a particularly pleasant surprise

Either the yields are low and they're disabling many SMs, or they're taking advantage of the absence of competition and 'gating' the performance for the cut die ; they don't want to offer much more than stock Titan X at 379$

Of course it could be clocked low and consequently have a lot of OC headroom, but I don't see that happening given 1080

Thoughts, gentlemen?
Click to expand...

Physical defects are not the only yield issue for manufacturing. For instance with a given power target more dies would be able to hit a lower clockspeed versus a higher one irrespective defects.

As a side this is why chatter about costs simply based upon die size, which seems to be commonly brought up nowadays, is flawed. A larger die could be more cost effective then a smaller one and have higher yields dependent on a number of other factors (and this is just the chip itself, not counting other costs related to making an entire graphics card).

1920 (30 SM) cores at ~96% the clock speed or 2048 (32 SM) at ~90% (GTX 670 to 680 ratio) the clock speed would match the given numbers. Although I wonder if they could disable 1 SM (or half a TPC), this at 970 to 980 clock speed ratio would also given the TFLOPS result.

Something to consider as well, well subject to change with further information, lower clock speeds would also given more differential in certain areas where hardware resources are the same (eg. ROPs).
 
TaintedSquirrel said:
Ashes of Singularity benchmark.

Ashes of the Singularity

Seems fake? I see 980 Ti's at 10k score.
Click to expand...
Considering the date yeah :)

We will probably see some leaked benches soonish with the cards distributed to reviewers/analysts, but I would heavily discount any benchmark "leaks-posts" dated before next week, only a very rare few will be true.
That said, DX12 games may need some patches to work optimally with Pascal (considering coding path done for Maxwell-Kepler) *shrug*.
Cheers
 
TaintedSquirrel said:
I dug through the Ashes results a bit more and it turns out most of them are mGPU results (doh).
The 1080 being benched there seems to be about 10-15% faster than the other "Fury Series" results. It would probably be a bit higher with a larger sample size, assuming Pelly isn't pushing the card's OC at all (assuming the Fury cards are OC'd).

Here are all the results from the same user.
It seems to be 47~50fps for the GTX 1080 and 41~42 for the Fury (X). Around 15% faster.
Click to expand...

I still think these are fake, reason I gave earlier.
Cheers
 
Just for the record, I ran new ashes of the singularity benchmarks at 1490/8000 recently to test new drivers with new windows install

Screenshot (3).png
 
You must log in or register to reply here.
Back
Top