Anybody else unimpressed with new midrange Nvidia GPUs, and much higher MSRP?

[sam_816]

I'm sure a more practical and reasonable list of poll options could be chosen, but so far I have seen people either love the new cards or hate them. People who don't understand that they really are mid range dies seem to love the cards. Those who realize they are mid range correctly hate the cards with god-like authority.



You mean 290x OC. Yes, the 290x OC makes absolutely everything from the entire Kepler and Maxwell generations, from top to bottom, look completely ridiculous.

agree with u and sad part is that even some of the well-versed and up-to-date (with technological advancements) have blind love for either NV or AMD.
i watched NV's presentation (& even at AMD's E3 pres) & couldn't understand why ppl (including tech channel/ gaming youtubers/press) were cheering at vague statements and claims without any live demo/benchmarks etc??
ppl learned nothing from gtx970's 3.5+5gb or fury's "overclocker's dream" fiascos?
ppl are applauding NV's marketing strategies but who will eventually pay?

 

[maddie]

I agree I think there should be a poll. I'd make one but there isn't room for yet another pascal thread anywhere lol. Too cluttered. Please add a poll.

"Are you happy with Nvidia's newly announced 1000 series?"

"Yes. They are worthy successors"
"No. They are overpriced, mid range garbage and should cost $350 and $250"

Go make it, OP.

In all my nightmares, I never thought WW III would start this way.

 

[tential]

A lot of people seem to forget that the reference 980Ti is significantly underclocked and the majority of the cards sold are about 15%-20% faster. Going by how Kepler and Maxwell aged I wouldn't be surprised if 1080 only matches fury in the near future. 25% faster than Titan X really makes it imperceptibly faster than actual 980Ti cards. I'm really furious that I burned my 980Ti as it seems that 1080 won't be any faster than my card was. The only saving grace I can think of is overclocking headroom under water cooling with voltage tweaks. I really hope it reacts to Voltage change very well and that 2.5GHz+ is possible under water. But that pathetic 256bit bus may spoil even that. If it overclocks well under water it might make it quite good, I only hope so because thanks to planned obsolescence I actually can't play latest games on a card that cost 1000$ even though it is watercooled and overclocked as far as it can be. I really need a new card, maybe I'll screw NV just like it screwed me thanks to their planned obsolescence, even 980Ti looks pathetic in DX 12. The state of Multi-GPU gaming doesn't look good right now it's a shame because 3x390X would be sweet if they scaled properly in games.

I was joking about the 390x beating the 980ti but now it may come to fruition. Ugh I hate this card when will it stop performing so I can upgrade!

 

[BlitzWulf]

agree with u and sad part is that even some of the well-versed and up-to-date (with technological advancements) have blind love for either NV or AMD.
i watched NV's presentation (& even at AMD's E3 pres) & couldn't understand why ppl (including tech channel/ gaming youtubers/press) were cheering at vague statements and claims without any live demo/benchmarks etc??
ppl learned nothing from gtx970's 3.5+5gb or fury's "overclocker's dream" fiascos?
ppl are applauding NV's marketing strategies but who will eventually pay?

Agree 100% they even gave a round of applause to JHH for literally doing nothing more than showing 1080p60fps gameplay footage of three already released games and describing the hard work of those devs as if they were all Nvidia achievements! That peanut gallery that was in attendance was worse than the "live studio audiences" that T.V. shows used to be recorded in front of.

 

[3DVagabond]

I find it really hard to explain that to people. They seem to want the best for years to come, I wonder if AMD is only starting to catch up to Nvidia in games because they already stopped working on the drivers. That way it is easier for review sites to show the newer video card beating the older ones.

I've been thinking this myself. I'm wondering how much optimizing nVidia has been doing lately for Maxwell.

Look at 970/980 vs. 290/290X.

If when GM104 was released the performance stacked up like this against Hawaii cards how many would have sold? Especially when you consider what the pricing was. I think the performance tanking is already underway.

I don't know how much performance AMD is going to be able to get out of a 232²mm chip. Seems too small to me. We also don't know how much performance 14FF is going to offer, but TSMC 16FF sure seems to be great. Able to hit very high clocks. AMD could be launching a much smaller chip at much lower clocks.

Now, it is possible that the perf/GHz is much lower and/or that Pascal hasn't given much of a boost architecturally and Polaris has. AMD has claimed it's their biggest perf/W improvement ever. How well that'll translate in real terms I don't know. I suppose, for the sake of competition, we can hope that AMD has been really sandbagging it with the clock speeds and Polaris can be clocked up to gain raw performance. If we've seen Hawaii performance @ ~1GHz and it can clock appreciably higher then AMD could blow Pascal out of the water. We've seen nothing to suggest this though. It seems Polaris was designed to be purely Perf/W champion. That's not typically a chip that's designed for high clocks as higher frequencies and Perf/W are opposing metrics.

 

[Fallen Kell]

If when GM104 was released the performance stacked up like this against Hawaii cards how many would have sold? Especially when you consider what the pricing was. I think the performance tanking is already underway.

I don't know how much performance AMD is going to be able to get out of a 232²mm chip. Seems too small to me. We also don't know how much performance 14FF is going to offer, but TSMC 16FF sure seems to be great. Able to hit very high clocks. AMD could be launching a much smaller chip at much lower clocks.

Now, it is possible that the perf/GHz is much lower and/or that Pascal hasn't given much of a boost architecturally and Polaris has. AMD has claimed it's their biggest perf/W improvement ever. How well that'll translate in real terms I don't know. I suppose, for the sake of competition, we can hope that AMD has been really sandbagging it with the clock speeds and Polaris can be clocked up to gain raw performance. If we've seen Hawaii performance @ ~1GHz and it can clock appreciably higher then AMD could blow Pascal out of the water. We've seen nothing to suggest this though. It seems Polaris was designed to be purely Perf/W champion. That's not typically a chip that's designed for high clocks as higher frequencies and Perf/W are opposing metrics.

You have been missing the point entirely of Polaris and keep thinking it is going to be a performance part, it isn't. It is meant as a low cost, high yield, commodity part that will perform well as a low power, low heat, low noise solution so that the big system integrators like Dell and HP will decide to put it as the "upgraded graphics" solution on their $500-800 computer systems and the mobile version on their laptops. The high performance gaming card isn't coming out till next year.

AMD is focusing on the market segment where the big money is located. High performance gaming is a niche market with a very low percentage of the sales coming from the top end products. The real market is in the $500-800 computers which are the vast majority of the systems sold. It is a hard market to hit lately due to Intel's integrated graphics on CPUs, but it is still the vast majority of the market.

 

[Pneumothorax]

Not really, I might either hate the cards or just dislike them depending on the overclocking potential but nonetheless I might still buy one because just after getting a new monitor that is 3440x1440 I had a leak and my 980Ti burned and the original Titan is useless at that resolution right now so I really need something to tide me over until Vega is released. I probably won't buy GP100 because I don't want to find myself in the same situation as I did with the Titans and that 980Ti is just starting to. NV cards are only good for a year tops, then they drop precipitously in performance

Oh yeah, Nvidia has taken me to the cleaners for both the 680 (the first mid-range card to be priced so high) and the 780Ti (thanks to the planned obsolescence on Kepler). Sad to see it now being done so soon for Maxwell. Glad I got rid of 970 SLI when Newegg offered to take them back for store credit! Either Nvidia is doing it on purpose or they're designing their cards for a 1 year horizon. Either way, I'm happy how my AIO cooled 290x's are holding up and will wait for the big 16nm chips. Sadly, both AMD/Nvidia probably realize that 16nm is going to have to hold up for half a decade and have planned their roadmaps accordingly by milking us every year for it!

 

[Arachnotronic]

Oh yeah, Nvidia has taken me to the cleaners for both the 680 (the first mid-range card to be priced so high) and the 780Ti (thanks to the planned obsolescence on Kepler). Sad to see it now being done so soon for Maxwell. Glad I got rid of 970 SLI when Newegg offered to take them back for store credit! Either Nvidia is doing it on purpose or they're designing their cards for a 1 year horizon. Either way, I'm happy how my AIO cooled 290x's are holding up and will wait for the big 16nm chips. Sadly, both AMD/Nvidia probably realize that 16nm is going to have to hold up for half a decade and have planned their roadmaps accordingly by milking us every year for it!

Nah, expect to see 7nm TSMC GPUs in 2019/2020. I wouldn't be surprised if Volta were a 10nm product, tbh.

 

[tential]

@3DVagabond
How many times has Moonbogg said that the Maxwell Driver team was already fired?
You're going to see the effects soon!!!!

 

[provost]

And, I believed him... Lol since we certainly have a precedent with late releasing GK110s dropping support within a few months. If it worked before, why change it.... it's very profitable for NV... Lol...
I guess it's a personal choice whether one wants to "rent" performance or own it.

 

[xpea]

Latest rumor from eastern AIBs is that top-end custom GTX1080s will clock up to 2.4GHz :awe:

exit: not first on this one

 

[DDH]

Latest rumor from eastern AIBs is that top-end custom GTX1080s will clock up to 2.4GHz :awe:

exit: not for first on this one

Link please?

 

[R0H1T]

Nah, expect to see 7nm TSMC GPUs in 2019/2020. I wouldn't be surprised if Volta were a 10nm product, tbh.

That'd be unexpected, still inexplicably nice, but I doubt they'd push Volta to 2018. They also wouldn't want to cede HPC contracts to Intel, nor AMD for that matter, not anymore than they absolutely have to

Thinking about it some more, they could (paper) launch small Volta towards the end of 2017 along with big HPC dies in 2018, but still that ought to be 16nm.

 

[maddie]

Seen on HardOCP: http://www.hardocp.com/article/2016/05/09/nvidia_founders_edition_cards_yea_or_nay

"I have literally been told for years, that NVIDIA either breaks even or loses money on NVTTM/reference cards sold into the market."

Well look at that. So this explains the $100 premium.

Ripleys effect.

 

[pj-]

I hope nobody here who is perpetually furious at nvidia ever goes to an audiophile forum. Your head would explode. I may lose a couple hundred bucks a year by chasing the top spot in benchmarks, but at least there are measurable improvements doing so. Those people spend thousands on audio gear that makes almost to literally no difference on the sound.

If 1080's OC headroom is good, dx12 perf is good, and the VR claims materialize, is AMD going to have anything remotely comparable before vega? If I can afford it and I want the best experience, what reason is there to not stick with nvidia?

 

[Ajay]

Seen on HardOCP: http://www.hardocp.com/article/2016/05/09/nvidia_founders_edition_cards_yea_or_nay

"I have literally been told for years, that NVIDIA either breaks even or loses money on NVTTM/reference cards sold into the market."

Well look at that. So this explains the $100 premium.

Ripleys effect.

Good article, and that cooler must be pretty special because:

Looks like a good overclock and decent temps (for an exhaust only cooler).

 

[Stuka87]

Good article, and that cooler must be pretty special because:

Looks like a good overclock and decent temps (for an exhaust only cooler).

The real question will be how will cards sold to the public OC? I have no doubt the demo units given to reviewers are going to clock well. Should have a decent idea in 1-2 months when they start to get into the hands of more than just a few people.

 

[Arachnotronic]

Seen on HardOCP: http://www.hardocp.com/article/2016/05/09/nvidia_founders_edition_cards_yea_or_nay

"I have literally been told for years, that NVIDIA either breaks even or loses money on NVTTM/reference cards sold into the market."

Makes sense, esp. since NVIDIA seems to be all-too-eager to get the reference cards out of the market ASAP.

 

[tential]

I hope nobody here who is perpetually furious at nvidia ever goes to an audiophile forum. Your head would explode. I may lose a couple hundred bucks a year by chasing the top spot in benchmarks, but at least there are measurable improvements doing so. Those people spend thousands on audio gear that makes almost to literally no difference on the sound.

If 1080's OC headroom is good, dx12 perf is good, and the VR claims materialize, is AMD going to have anything remotely comparable before vega? If I can afford it and I want the best experience, what reason is there to not stick with nvidia?

I don't know what audiophile forums you hang around but with DIY, I always see people make upgrades that "I would do". But I'm an audiophile... so maybe I don't have a limit.

And you should pick the GPU you want....

 

[swilli89]

I hope nobody here who is perpetually furious at nvidia ever goes to an audiophile forum. Your head would explode. I may lose a couple hundred bucks a year by chasing the top spot in benchmarks, but at least there are measurable improvements doing so. Those people spend thousands on audio gear that makes almost to literally no difference on the sound.

If 1080's OC headroom is good, dx12 perf is good, and the VR claims materialize, is AMD going to have anything remotely comparable before vega? If I can afford it and I want the best experience, what reason is there to not stick with nvidia?

Unless some miracle happened with Polaris and it could swap blows with 1080, you definitely stick with nVidia at the absolute high end if you are after a single card.

However with more seamless multi-gpu promised in dx12, its natural ability at VR (split screens) and potential top end perf and perf/$ dual Polaris 10 may be the best high end solution at $>600.

Time will tell. I'm very curious to see the 1080's OC headroom, I'd love to be stunned with something like ~2.2 ghz overclocks on air with reasonable power draw. Whats got me really confused is trying to reconcile the fact that NV was able to coax 2.1 ghz out of a medium size chip on TSMCs 16nm while there were rumors of an AMD engineering sample being tested at 800mhz and nothing has been spotted elsewhere to counter that. Is there really a huge advantage to TSMCs process? Or is AMD going to release an uber clocked chip of their own? Possibly a chip that scales very well down to laptop speeds yet one that can clock really high on desktop? AMD chips at high frequencies have always been dangerous..

I know TSCMs chips made for apple clocked great and that was almost a year ago, but I have also heard great things about GF/Samsungs 14nm on its own mobile chips so something isn't adding up.

 

[iiiankiii]

Good article, and that cooler must be pretty special because:

Looks like a good overclock and decent temps (for an exhaust only cooler).

The demo ran with vsync on. Very likely temp never shot up due to low GPU load. Plus, the GPU has a tdp of 180w. I bet the cooler is the same as previous gen but with a different cover.

 

[pj-]

Unless some miracle happened with Polaris and it could swap blows with 1080, you definitely stick with nVidia at the absolute high end if you are after a single card.

However with more seamless multi-gpu promised in dx12, its natural ability at VR (split screens) and potential top end perf and perf/$ dual Polaris 10 may be the best high end solution at $>600.

Time will tell. I'm very curious to see the 1080's OC headroom, I'd love to be stunned with something like ~2.2 ghz overclocks on air with reasonable power draw. Whats got me really confused is trying to reconcile the fact that NV was able to coax 2.1 ghz out of a medium size chip on TSMCs 16nm while there were rumors of an AMD engineering sample being tested at 800mhz and nothing has been spotted elsewhere to counter that. Is there really a huge advantage to TSMCs process? Or is AMD going to release an uber clocked chip of their own? Possibly a chip that scales very well down to laptop speeds yet one that can clock really high on desktop? AMD chips at high frequencies have always been dangerous..

I know TSCMs chips made for apple clocked great and that was almost a year ago, but I have also heard great things about GF/Samsungs 14nm on its own mobile chips so something isn't adding up.

Multi GPU has some challenges with VR. Since there's only one video cable on current headsets, the second GPU has to copy its output to the primary gpu, which apparently adds a non-trivial amount of latency.

Aside from that, multi GPU VR is a niche within a niche so I am not counting on good support in the short term.

 

[Azix]

Makes sense, esp. since NVIDIA seems to be all-too-eager to get the reference cards out of the market ASAP.

The idea that they lose money on reference cards makes no sense. They set the MSRP for one thing and to suggest that the AIB partners with their fancy coolers and PCBs can sell for less and make a profit as well as give nvidia a profit on the sales is hard to believe. Those who do not use a custom PCB use the reference PCB, so somehow they are able to do this, make a profit and give nvidia a cut for less?

please

 

[beginner99]

Time will tell. I'm very curious to see the 1080's OC headroom, I'd love to be stunned with something like ~2.2 ghz overclocks on air with reasonable power draw. Whats got me really confused is trying to reconcile the fact that NV was able to coax 2.1 ghz out of a medium size chip on TSMCs 16nm while there were rumors of an AMD engineering sample being tested at 800mhz and nothing has been spotted elsewhere to counter that. Is there really a huge advantage to TSMCs process? Or is AMD going to release an uber clocked chip of their own? Possibly a chip that scales very well down to laptop speeds yet one that can clock really high on desktop? AMD chips at high frequencies have always been dangerous..

Can't compare different uArch with clockspeed. just like you can't do that with CPU eg Athlon vs Pentium 4 or SandyBridge vs Bulldozer.

GCN clocks lower already than maxwell in current gen on same 28 nm process. It's not made for high clocks it seems. As you can see with Fury nano or underclocked hawaii cards, lower clocks give a lot better performance/watt. 290/390 are basically factory OCed cards. Undervolt and reduce clock to 900 mhz and you almost half power usage.

So the tested sample might be an ultra small form factor Polaris 10 like the fury nano. AMD said there will be more products from a single die/chip design this round. I can see high performance P10 with GDDR5x, higher clocked and much worse performance/watt, a Fury Nano like P10 card, and same with cut-down dies. We will probably see a P10 Nano card with half the power use of the top P10 card and 80%-90% of the performance of it. Something that fits well with poor power supplies. They can sell it to OEMs or to consumers wanting to upgrade OEM PCs.

 

[airfathaaaaa]

Latest rumor from eastern AIBs is that top-end custom GTX1080s will clock up to 2.4GHz :awe:

exit: not first on this one

sure of micron gives them the absolute best modules
which by definition on will buy only them
so good luck finding the right model