NVIDIA to Steal AMD's Ryzen Limelight on Feb 28

[Grings]

I bet its £799+ in the uk, same as maxwell titan cost
1080 was same price as 980ti, 1070 same price as 980, and 1060 same as the 970, makes sense
Plus partners are allowed to make custom pcb's so we will probably see some closer to 1000
(and despite having really capable pcb's, i bet they all are hardlocked to overclock about the same, just short of titan performance)

Never mind ryzen, its vega i want to see breaking monopolies

 

[kruk]

The ironic thing about the licenses right now is that x86 is pretty much dead so Intel is like living off AMD in a roundabout way.

Isn't AMD64 just an extension of x86 and cannot exist without the latter?

 

[eidairaman1]

Isn't AMD64 just an extension of x86 and cannot exist without the latter?

It can Exist on its own, X86 is just for legacy stuff now. Once everyone is pushed away from 32bit instructions and memory spaces (all devices that have a CPU in them) then it will just be called X64 at that point. X86 started 8086, aka 80286, 80386, 80486, 80586(pentium)

https://en.wikipedia.org/wiki/X86

 

[ADHDGAMING]

Ha, NVIDIA wishes! GTX 1080 Ti is too expensive for most gamers to care. Ryzen, on the other hand, is upsetting the almost 11 year status quo. GTX 1080 Ti will be a footnote while Ryzen gets mainstream press coverage.

This behavior coming out NVIDIA is expected: very unsportsmanlike.

Dont we mean Vega???

 

[mouacyk]

It will be ~30% faster than GTX 1080 and cost ~40% more, so a good deal for demaning gamers.

By pricing the TXP at $1200, NVidia just cleared a pricing spot for the Ti. They are regretting the killing they could've done with the 980 ti but didn't because they practically re-offered the TXM at $650. It's called price anchoring, and the TXP pricing gave it all away when it came out.

I'm not budging from my 980 TI, unless Vega has a SKU that can offer +50%.

 

[efikkan]

Do you really think nVidia will charge only 800$ for a card allegedly faster than $1200 Titan X Pascal with Vega missing? I will be surprised if the card is under $900.

GTX 1080 Ti will not be faster than Titan X (Pascal), it will use a lower binning but with slightly higher clocks, giving marginally lower performance than the Titan X.

You basically don't know nothing about Vega and yet you keep telling us how fast it is and how much power it consumes. Please stop doing this until it releases, thanks!

AMD has demonstrated the performance of Vega 10, slightly beating GTX 1080 in AMD favoring titles like Doom, so we actually know. This is the Vega chip which will be just as big as a GP102, while matching the GP104 in performance.

BTW; The most efficient Pascal is >80% more efficient than the most efficient Polaris. There is no way Vega will close the whole gap.

A significant upcoming product it is, but not compared to Ryzen... Not anywhere close. You're looking at the first mass market octacore vs. a cheaper variant of the existing Titan line.

One product is ~30% faster and 40% more expensive, and the other is ~50% cheaper and beats/matches the fastest cpu on the market.

I never said it was, and if you actually read my post you'll see I argued it would be unfortunate for Nvidia to release it at the same time.

 

[uuuaaaaaa]

AMD machine right now:

 

[FordGT90Concept]

Dont we mean Vega???

Nope, Vega is gonna be a while yet (1H 2017 last I heard). NVIDIA is trying to rain on Ryzen's parade.

 

[Hotobu]

Nope, Vega is gonna be a while yet (1H 2017 last I heard). NVIDIA is trying to rain on Ryzen's parade.

I don't see how. They don't compete in the same sector. I guess the argument can be made that if people go out and spend money on a processor motherboard combo then that leaves less wallet share for spending on a video card, but I would think very rare is the case that someone says, "Let me skip this top of the line GPU to go buy a new motherboard/processor." My guess is that the people who can afford to drop close to 1K on a videocard are most likely very rarely going to be in a situation where they are CPU bottlenecked enough to have to make that decision. If anything new systems should encourage the purchase of video cards.

I think it's moreso the case that they're piggybacking off of AMD and not trying to steal the limelight.

 

[jabbadap]

Nope, Vega is gonna be a while yet (1H 2017 last I heard). NVIDIA is trying to rain on Ryzen's parade.

Well you could say If Ryzen is great success, then Nvidia is trying to ride on it by offering top performant graphic cards.

 

[kruk]

GTX 1080 Ti will not be faster than Titan X (Pascal), it will use a lower binning but with slightly higher clocks, giving marginally lower performance than the Titan X.

Maybe not the reference versions, but last time I checked the partner 980 Ti's were faster than Titan X Maxwell and thats what most people will buy.

AMD has demonstrated the performance of Vega 10, slightly beating GTX 1080 in AMD favoring titles like Doom, so we actually know. This is the Vega chip which will be just as big as a GP102, while matching the GP104 in performance.

BTW; The most efficient Pascal is >80% more efficient than the most efficient Polaris. There is no way Vega will close the whole gap.

I'm not going through this again. You have a black box GPU running a single game and you are able to predict it's performance and power consumption? No, just no.

 

[medi01]

AMD machine right now:

Thanks, not sure what Raja has to do with Ryzen though. He better work harder on Vega, Polaris was meh.

 

[medi01]

BTW; The most efficient Pascal is >80% more efficient than the most efficient Polaris. There is no way Vega will close the whole gap.

Fury X, a 28nm product, consumes 70% more than 1070, that it occasionally beats.

Power consumption difference between 480 and 1060 is about 20%.

"most efficient" vs "most efficient" is a silly comparison, as one might even not be comparing cards from the same tier.

 

[Zeki]

that is right - a 1 year old architecture, 1000$ GPU (that is not even the top card by any means) will steal the show. I and many will buy that 1000$ card and pair with old intel i5-2500k, because there was little to none improvement in these last 6 "generations" - and none of us cares about CPU's and platform, but that is ok, cuz nvidia told us so.

Lol so true, I went from gtx980ti & i5-2500k to gtx980ti& i7-6700k to gain a few fps, well I hope I did it's hard to tell.

 

[efikkan]

Fury X, a 28nm product, consumes 70% more than 1070, that it occasionally beats.

Power consumption difference between 480 and 1060 is about 20%.

That's not even remotely close to the truth.

"most efficient" vs "most efficient" is a silly comparison, as one might even not be comparing cards from the same tier.

You are missing the whole point. The least efficient Pascal (GTX 1060) is still ~55% more efficient than the most efficient Polaris (RX 480). Vega 10 have to compete with gp104 which is 80% more efficient than Polaris, so that is the benchmark!

 

[medi01]

That's not even remotely close to the truth.

Besides being apples to oranges (perf/watt is not the same as absolute power consumption, good job trying to twist reality here) that's an ancient review, a sad example of why you shouldn't trust TPU much:

1) that has rather remarkable power consumption difference (compared to majority of sites) with 480 consuming even more than 1070
2) has messed up charts (arithmetic mean instead of geometric)
3) 480 gained a lot of perf after that

Check recent reviews of AIBs in power consumption in actual games with actual drivers. E.g. on computerbase.

The least efficient Pascal (GTX 1060) is still ~55% more efficient than the most efficient Polaris (RX 480). Vega 10

Bullshit, even if you'd omit vega, which you can't know shit for sure about, in that ridiculous sentence.

 

[bobalazs]

Or the other way around; Amd steals the spotlight away from *vidia.

 

[efikkan]

1) that has rather remarkable power consumption difference (compared to majority of sites) with 480 consuming even more than 1070
2) has messed up charts (arithmetic mean instead of geometric)
3) 480 gained a lot of perf after that

Check recent reviews of AIBs in power consumption in actual games with actual drivers. E.g. on computerbase.

That's completely untrue. AMD fanboys keep telling a fairytale that the first reviews and the first cards were bad. I've actually checked >80 reviews of RX 480, there is no significant variance between the initial reviews and the more recent, just the normal few percent variance as you should expect.

You claim RX 480 has gained a lot since that, well it has not. It has gained a few percent here in select games and there through driver tweaks, but so has the competition too. The performance gap between RX 480 and GTX 1060 remain unchanged.

Bullshit, even if you'd omit vega, which you can't know shit for sure about, in that ridiculous sentence.

What? Try reading your own post before submitting.

Vega 10 has to compete with GP104, so the efficiency of GP104 is the benchmark. GTX 1080 is 80% more efficient than RX 480, so that's the gap AMD need to close. I also want to remind you that even though AMD promised ~2.5× performance/watt with Polaris, it was just a few percent more efficient than Fiji, and that's even with a node shrink! For comparison, the "boring" Pascal achieved a massive gain in efficiency vs. Maxwell. Try reading the numbers yourself.

 

[medi01]

I've actually checked >80 reviews of RX 480, there is no significant variance between the initial reviews and the more recen

Oh really:
http://www.hardwarecanucks.com/foru.../73945-gtx-1060-vs-rx-480-updated-review.html

It improved for two reasons:
1) Ref card was poor (throttling)
2) Driver bump

Oh, 290x is faster than 780Ti nowadays, isn't it? Miracle.

You claim RX 480 has gained a lot since that, well it has not.

Yeah, "has not". And for the power consumption, Anno game, total system power consumption (in watts):

From https://www.computerbase.de/2017-02...-ti-kalmx-test/2/#abschnitt_leistungsaufnahme

1080 is 80% more

Oh dear.
1) %-s don't work like you think
2) With your approach you have 2 different values from your own slides.
3) For real figures of real cards in real game see chart above.

 

[medi01]

Total system consumption in watts while running Anno
Come show me that giant power efficiency gap.

480 AIBs:

https://www.computerbase.de/2016-12...tt_so_viel_watt_kosten_die_hoeheren_taktraten

1060 AIBs:

https://www.computerbase.de/2016-10...tt_so_viel_watt_kosten_die_hoeheren_taktraten

 

[londiste]

Oh really:
http://www.hardwarecanucks.com/foru.../73945-gtx-1060-vs-rx-480-updated-review.html

It improved for two reasons:
1) Ref card was poor (throttling)
2) Driver bump

it improved primarily for one reason - different selection of benchmarked games.
in games benchmarked both in this review and previous ones, rx480 improved 2-3% averaged across games (which is pretty good result in itself).

regarding these power consumption graphs, anno 2205 is a pretty strange game to use for power consumption test. if nothing else, that game is pretty consistently favoring nvidia cards.
look at the previous page in both these reviews - 1060 gets 44 fps, rx480 gets 33 fps, that is a 33% difference.
in a test machine with an overclocked 6700k, 4 ram sticks and 3 drives (this setup should consume above 100w at load even without graphics card), it would seem that neither card is fully utilized by anno 2205.

 

[medi01]

it would seem that neither card is fully utilized by anno 2205.

Mind blown.

 

[medi01]

Note Vishera in this lits

 

[efikkan]

Oh dear.
1) %-s don't work like you think

Look at the link I provided:

The other one:

As everyone can see, GTX 1080 is ~80% more efficient, since it provides ~80% more performance per watt. This is the very definition of energy efficiency. You seriously can't disagree with basic math! I'm sorry, but this is completely ludicrous.

2) With your approach you have 2 different values from your own slides.

No, again see the link I provided and this one;
GTX 1080 182% performance per watt compared to RX 480, and 178% in the other. Different benchmarks showing exactly the same.
I have not manipulated any data, just compressed the charts to show the relevant models. Being dishonest does not help your case.

3) For real figures of real cards in real game see chart above.

So you are accusing Techpowerup and the sites behind 80+ other reviews which arrived with the same conlusion for fake reviews?
I think you should learn about statistical outliers is before you start cherry-picking.

 

[medi01]

As everyone can see, GTX 1080 is ~80%

You have been told, what was wrong with this benchmark.
You ignored it or didn't comprehend.
You were told, %-s don't work like you think.
You got to an idea that repeating the same statements somehow makes it better.

You were provided with benchmarks and links showing actual power consumption in actual game across AIB 480's and AIB 1060's being roughly in the same ballpark, which you again have ignored.

What are you expecting from me to tell you now?