GTX 980Ti finally launched - MSRP $649 - Reviews

[Face2Face]

What I would like to know, is what would be faster if both were water-cooled and OC'd to the max without increasing voltage.

No voltage increase on water = Boring.

Honestly, there is way too much silicon lottery at play. You would need to take an average of multiple cards for this to be fair.

That will be down to the silicon lottery. BUT better VRM and components certainly can help...

Very true. I've put enough voltage through my 16 Phase Lighting that would make a reference card blow up and melt in a fire D:

 

[thehotsung8701A]

My question is why aren't you guys waiting to see what AMD Fury can bring to the table?

 

[moonbogg]

Nice upgrade. Should see at least 2x performance and maybe even closer to 3x? 3x VRAM too.

Yeah thanks man. The last time I spent over $600 on GPU's was with the 7800GTX's. Kept it between $3-$450 for a long time after that, but not this time though. I can ride these until big Pascal I'm pretty sure, but never any guarantees there, lol. I don't normally spend in this price bracket anymore but I had to grab those Titan X's since going on sale for nearly half the price. I had to bite.

 

[LTC8K6]

My question is why aren't you guys waiting to see what AMD Fury can bring to the table?

A pig in a poke is usually a bad deal.

Give me some solid leaks, and I won't buy a 980ti.

 

[exar333]

My question is why aren't you guys waiting to see what AMD Fury can bring to the table?

I certainly am.

For me, I want the best option ~$500-600 (yes, the 980Ti would roughly fit this budget...)

This includes the following:
-Base performance
-Custom PCB options
-Overclockability
-power consumption when overclocked (strictly to ensure I could add a 2nd card later, if I want, without swapping-in a new PSU)
-custom BIOS availability (NV only)

The last bullet is important because of NV's silly TDP cap nonsense.

Edit: Probably looking at 60-90 days before I make a decision...

 

[thehotsung8701A]

I certainly am.

For me, I want the best option ~$500-600 (yes, the 980Ti would roughly fit this budget...)

This includes the following:
-Base performance
-Custom PCB options
-Overclockability
-power consumption when overclocked (strictly to ensure I could add a 2nd card later, if I want, without swapping-in a new PSU)
-custom BIOS availability (NV only)

The last bullet is important because of NV's silly TDP cap nonsense.

I really hope the Fury is not 4GB of HBM. Anything less than 6GB is not acceptable considering console use 6GB of VRAM. Though if AMD messes this up, then they are going to lose even more share to Nvidia. If AMD user switch over to Nvidia, that not a good sign for everyone.

I really hope it 8GB of HBM with 4096 core and price point of around $700 to $800 then it would be a no-brainer and I don't need to upgrade indefinite.

Profanity is not allowed in VC&G
-Moderator Subyman

 

[B-Riz]

I'm confused as to why anyone would buy this in stock form right now. (unless SLI, but why not wait for CLC?)

This card has good (really good) OC potential, and the better performers will not be using the stock cooler.

 

[thehotsung8701A]

I'm confused as to why anyone would buy this in stock form right now. (unless SLI, but why not wait for CLC?)

This card has good (really good) OC potential, and the better performers will not be using the stock cooler.

What is consider stock and what isn't consider stock? As far as I know both AMD and Nvidia dont' actually make their own brand GPU. Also is the CLC which i'm assuming stand for custom the same price as stock or is it more expensive?

 

[2is]

I never said HBM's efficiency. I said Tonga's memory efficiency:
http://www.anandtech.com/show/8460/amd-radeon-r9-285-review/3

"AMD’s own metrics call for a 40% gain in memory bandwidth efficiency, and if that is the average case it would more than make up for the loss of memory bandwidth from working on a narrower memory bus. We’ll see how this plays out over our individual games over the coming pages, but it’s worth noting that even our most memory bandwidth-sensitive games hold up well compared to the R9 280, never losing anywhere near the amount of performance that such a memory bandwidth reduction would imply (if they lose performance at all)."

AMD (and nVidia for that matter) COULD improve memory efficiency but from what I understand, it's been cheaper for them to just slap on more vram rather than optimize for a smaller memory pool. IIRC, AMD stated that since HBM is more expensive, it actually makes financial sense (dedicating man hours) to optimize a smaller memory pool. I wouldn't discount 4GB of HBM so easily, especially on an unreleased card.

4GB is still 4GB, if something needs more it needs more. If the amount of vram you NEED exceeds the amount of vram available on the board your performance will tank, doesn't matter if it's HBM or not. Not sure why you think it's suddenly ok to run out of VRAM just because it's HBM.

If we adopt your philosophy and apply it to system ram, that's like saying 4GB of DDR4 should be just as good as 6GB of DDR 3 because it's DDR4. Sorry but no, that's not how it works. In both situations, if you require more than the amount of RAM available, performance will tank. I know I'm repeating myself, but it seems like you might not fully understand how ram works.

 

[B-Riz]

What is consider stock and what isn't consider stock? As far as I know both AMD and Nvidia dont' actually make their own brand GPU.

Reference specifically.

This card needs the proper treatment from EVGA, Asus, MSI, Gigabyte, etc.

 

[monkeydelmagico]

No voltage increase on water = Boring.

Honestly, there is way too much silicon lottery at play. You would need to take an average of multiple cards for this to be fair.



Very true. I've put enough voltage through my 16 Phase Lighting that would make a reference card blow up and melt in a fire D:

Hard cap at 275w for GM200 cards?

 

[exar333]

I'm confused as to why anyone would buy this in stock form right now. (unless SLI, but why not wait for CLC?)

This card has good (really good) OC potential, and the better performers will not be using the stock cooler.

If you need an upgrade, you need it. I grabbed a cheap 970 on launch because my 670 wasn't cutting it at all, especially with my new display at the time. I threw the dice and got a pretty solid OC'er, but didn't have time to wait for the Lightning/FTW versions.

This time around, I am investing more $$$ and I truly think these cards will be the best for the next 14 months or so until we see 14nm/16nm. Who knows what delays we could see there, so I want to make sure to get the best one I can. I can afford to wait it out a bit. Not playing TW3 until its a few months old and bugs have been worked through and drivers optimized, so that will help too.

 

[B-Riz]

4GB is still 4GB, if something needs more it needs more. If the amount of vram you NEED exceeds the amount of vram available on the board your performance will tank, doesn't matter if it's HBM or not. Not sure why you think it's suddenly ok to run out of VRAM just because it's HBM.

If we adopt your philosophy and apply it to system ram, that's like saying 4GB of DDR4 should be just as good as 6GB of DDR 3 because it's DDR4. Sorry but no, that's not how it works. In both situations, if you require more than the amount of RAM available, performance will tank. I know I'm repeating myself, but it seems like you might not fully understand how ram works.

More than likely, AMD has some compression tricks no one knows about.

So, this 4GB sux0rs drumbeat is old. (Also, system RAM <> video RAM, not a good comparo)

Go read up on compression (MKV, ZIP, etc) and get back to us.

Also, DX12 is a huge unknown that could make compatible cards much better vs DX11.

 

[iiiankiii]

4GB is still 4GB, if something needs more it needs more. If the amount of vram you NEED exceeds the amount of vram available on the board your performance will tank, doesn't matter if it's HBM or not. Not sure why you think it's suddenly ok to run out of VRAM just because it's HBM.

If we adopt your philosophy and apply it to system ram, that's like saying 4GB of DDR4 should be just as good as 6GB of DDR 3 because it's DDR4. Sorry but no, that's not how it works. In both situations, if you require more than the amount of RAM available, performance will tank.

You're right. If the game really needs that much VRAM buffer, it will need it. However, from what AMD said, VRAM buffer management hasn't been that efficient. They claimed that it was less expensive to increase the VRAM instead of improving VRAM management via driver. AMD claimed that the high bandwidth (hence HBM) would allow them to have less redundancy in memory buffer to achieve a greater bandwidth. With standard GDDR5, you need to populate as much memory buffer as possible to get the needed bandwidth.

"You're not limited in this world to any number of stacks, but from a capacity point of view, this generation-one HBM, each DRAM is a two-gigabit DRAM, so yeah, if you have four stacks you're limited to four gigabytes. You could build things with more stacks, you could build things with less stacks. Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better. So if you have four stacks you're limited to four [gigabytes], but we don't really view that as a performance limitation from an AMD perspective."

"If you actually look at frame buffers and how efficient they are and how efficient the drivers are at managing capacities across the resolutions, you'll find that there's a lot that can be done. We do not see 4GB as a limitation that would cause performance bottlenecks. We just need to do a better job managing the capacities. We were getting free capacity, because with [GDDR5] in order to get more bandwidth we needed to make the memory system wider, so the capacities were increasing. As engineers, we always focus on where the bottleneck is. If you're getting capacity, you don't put as much effort into better utilising that capacity. 4GB is more than sufficient. We've had to go do a little bit of investment in order to better utilise the frame buffer, but we're not really seeing a frame buffer capacity [problem]. You'll be blown away by how much [capacity] is wasted."

 

[boozzer]

A pig in a poke is usually a bad deal.

Give me some solid leaks, and I won't buy a 980ti.

just buy it :thumbsup: and leave the rest of us alone

 

[LTC8K6]

just buy it :thumbsup: and leave the rest of us alone

That's no fun.

I'd rather rant a while yet.

 

[thehotsung8701A]

That's no fun.

I'd rather rant a while yet.

Does AMD have anything close to GAMEWORK?

I hear that Hairwork is much better than TressFX, is that true?

 

[2is]

More than likely, AMD has some compression tricks no one knows about.

So, this 4GB sux0rs drumbeat is old. (Also, system RAM <> video RAM, not a good comparo)

Go read up on compression (MKV, ZIP, etc) and get back to us.

Also, DX12 is a huge unknown that could make compatible cards much better vs DX11.

So based on an unknown AMD compression that you know nothing about, up to and including it's existence, and unknown DX12 you have surmised that 4GB is adequate.

Talk about a logic fail. You'll have to do better than that to convince anyone who isn't trying to delude themselves.

I didn't say VRAM = System ram, I said performance will tank if you need more than you have of either one and bandwidth does not make up for capacity. If you think this is a false statement then you need to do some more research.

 

[thilanliyan]

4GB is still 4GB, if something needs more it needs more. If the amount of vram you NEED exceeds the amount of vram available on the board your performance will tank, doesn't matter if it's HBM or not. Not sure why you think it's suddenly ok to run out of VRAM just because it's HBM.

If we adopt your philosophy and apply it to system ram, that's like saying 4GB of DDR4 should be just as good as 6GB of DDR 3 because it's DDR4. Sorry but no, that's not how it works. In both situations, if you require more than the amount of RAM available, performance will tank. I know I'm repeating myself, but it seems like you might not fully understand how ram works.

See B-Riz and iiiankiii's posts. VRam is more than just the physical ram. You have the memory controller, drivers, etc. and how they all interact. If AMD is able to use the same tricks used in Tonga (did you read the AT article I posted?) for Fiji, why couldn't a smaller buffer be sufficient? In the AMD 285 2GB case, it is performing near a 280 3GB card, with much fewer resources.

As posted above from AMD,
"Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better."
That is what I'm getting at. To be clear, I'm not saying it WON'T be a problem, but I'm taking a wait and see approach. AMD wouldn't be stupid enough to make a high end card that tanks at ultra resolutions these days IMO.

 

[lavaheadache]

$650 to play what AAA game. There is nothing else coming out this year that would use that shiny 980 Ti. I already posted, no new Mass Effect, no new Elder Scrolls, no new Fallout (unless Bethesda just stealth drops it which I seriously doubt) Just Cause 3 is likely next year. High vs Ultra in Witcher III is next to irrelevant beyond screenshots, er - ? AAA releases are drying up - V and Witcher III have launched and those playing Day One like me are already finished or a third of the way through respectively.

And I doubt I'd consider AMD, it took them a year+ to release WHQL drivers, Nvidia has Day One drivers (and guides) for big releases. That alone is worth a premium.

And I'd pass on any reference card. You buy top shelf you want a top shelf cooler not some stock rubbish. Its like buying a K processor and keeping the stock cooler. And the Ti is gimped too, less cores and less textures units. Meh. Damn I'm getting old and grumpy.

If there aren't games to push you to buy the card, surely a new monitor would.... 34" ultra wides are nice... also a 4k would do the trick

 

[Enigmoid]

See B-Riz and iiiankiii's posts. VRam is more than just the physical ram. You have the memory controller, drivers, etc. and how they all interact. If AMD is able to use the same tricks used in Tonga (did you read the AT article I posted?) for Fiji, why couldn't a smaller buffer be sufficient? In the AMD 285 2GB case, it is performing near a 280 3GB card, with much fewer resources.

It does when it hits the vram wall. Look at shadow or mordor for example.

 

[2is]

See B-Riz and iiiankiii's posts. VRam is more than just the physical ram. You have the memory controller, drivers, etc. and how they all interact. If AMD is able to use the same tricks used in Tonga (did you read the AT article I posted?) for Fiji, why couldn't a smaller buffer be sufficient? In the AMD 285 2GB case, it is performing near a 280 3GB card, with much fewer resources.

As posted above from AMD,
"Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better."
That is what I'm getting at. To be clear, I'm not saying it WON'T be a problem, but I'm taking a wait and see approach. AMD wouldn't be stupid enough to make a high end card that tanks at ultra resolutions these days IMO.

I did, and responded. B-Riz post was completely ridiculous. "I don't know anything about AMD's compression or DX12 so that must make 4GB adequate" is essentially what he said. You referencing a post like that as if to say he made a good point makes this one equally ridiculous.

iiiankiii's post had more merit, but still doesn't address my concerns.

 

[96Firebird]

Are people trying to say that compression happens before the assets are allocated?

With the way color compression works on Tonga (delta color compression), you'd still need the whole asset in VRAM... The color compression only helps bandwidth, you still need the capacity.

 

[xthetenth]

AMD (and nVidia for that matter) COULD improve memory efficiency but from what I understand, it's been cheaper for them to just slap on more vram rather than optimize for a smaller memory pool. IIRC, AMD stated that since HBM is more expensive, it actually makes financial sense (dedicating man hours) to optimize a smaller memory pool. I wouldn't discount 4GB of HBM so easily, especially on an unreleased card.

It seems like for the most part the need for more bandwidth has driven the addition of enough memory to generally keep ahead of things so it's rarely been a focus. The current generation of compression seems to be about increasing effective bandwidth, rather than making more of a limited memory pool.

 

[biostud]

Inno3d is launching a GTX 980 Ti with Accelero hybrid cooling.