[TT] Pascal rumored to use GDDR5X..

[3DVagabond]

Maybe not issues with HBM but maybe issues with Nvidia getting HBM working. I am sure AMD had a lot of the delays for Fury because of getting HBM working correctly.

There was a report that the company that makes the TSV (Through silicon via) for Fiji only went into volume production about a month before Fury X released. That is what most likely has caused short supply.

 

[3DVagabond]

I don't understand your response to my post, care to elaborate?

You are saying that you don't mind old tech. I prefer state of the art tech rather than something that is old already at release.

 

[littleg]

Would a mid-range card be able to saturate a 448GB/s memory interface anyway?

 

[Pariah]

You are saying that you don't mind old tech. I prefer state of the art tech rather than something that is old already at release.

Why on earth does the age of the tech matter? I want whatever best fits my needs. I couldn't careless how old the tech is. I bought a plasma TV last year despite it being an EOL technology and older than LED TV's because it's simply a better TV. It's hard to comprehend how someone would base their purchasing decisions solely on how old the tech is. That's absurd.

 

[stuff_me_good]

Why on earth does the age of the tech matter? I want whatever best fits my needs. I couldn't careless how old the tech is. I bought a plasma TV last year despite it being an EOL technology and older than LED TV's because it's simply a better TV. It's hard to comprehend how someone would base their purchasing decisions solely on how old the tech is. That's absurd.

Like you said, you buy the best tech for you and you are right plasma is superior compared to LCD. Same goes for HBM, it's superior compared GDDR5X. If you even read all the benefits it has, no wonder enthusiasts wants it.

 

[Pariah]

Like you said, you buy the best tech for you and you are right plasma is superior compared to LCD. Same goes for HBM, it's superior compared GDDR5X. If you even read all the benefits it has, no wonder enthusiasts wants it.

The only thing I care about is performance. I don't care about size or power usage. My case can handle both fine, and I have no plans to replace it any time soon. HBM1 on AMD's cards has not demonstrated any measurable performance benefit in games vs GDDR5 in Nvidia's products. If Nvidia can produce a faster card with GDDR5x than AMD can with HBM2, I'm not going to buy the AMD card because it has newer memory tech in it. That would be idiotic.

 

[96Firebird]

You are saying that you don't mind old tech. I prefer state of the art tech rather than something that is old already at release.

If the old tech performs, why does it matter if it is old? Say they figured out how to increase bandwidth and decrease power consumption of DDR3, and called it Ultra-DDR3 or something. You wouldn't want it because it uses old tech?

I guess I just don't understand that mindset...

 

[stuff_me_good]

The only thing I care about is performance. I don't care about size or power usage. My case can handle both fine, and I have no plans to replace it any time soon. HBM1 on AMD's cards has not demonstrated any measurable performance benefit in games vs GDDR5 in Nvidia's products. If Nvidia can produce a faster card with GDDR5x than AMD can with HBM2, I'm not going to buy the AMD card because it has newer memory tech in it. That would be idiotic.

You care about performance and yet you don't care the faster tech? :whiste:

Let's assume like 15% faster but 2x the price, still idiotic to buy slower but cheaper product?

 

[Gikaseixas]

AMD could end up using GDDR5X too if HBM2 does not turn out as per plan. Think about that guys.

 

[Mondozei]

Rather than pushing the envelope regarding form factor, performance and efficiency, milking their loyal fanbase with yet another round of GDDR5

Nobody disputes that HBM is the future, but the implementation of HBM1 should temper anyone who believes we are looking at radical performance leaps in the early stages of the technology. We all heard the same propaganda pre-fury, how HBM(1) was going to change everything. It didn't.

Learn from that episode.

People defending the stagnation of tech with this move just proves that Nvidia is right with their thinking described above, or else we should start thinking that people posting here have underlying intentions regarding the commercial welfare of Nvidia (i.e we have disguised shareholders/shills).

You can do better than this, PPB. This kind of trolling should be below you.

 

[Goatsecks]

I don't see why this thread has had to become so partisan.

HBM has considerable performance advantages over DDR5 but is, currently, without its own problems, e.g., supply and cost. With this in mind it is not unreasonable for any company, Nvidia or AMD, to look to DDR5 for improvements if a reasonable supply of HBM is not on the horizon (this actually sounds pretty sensible, no?).

It would also be interesting to see an investigation in to memory bandwidth, what is currently required and what bottlenecks exist. Any one have any links?

 

[godihatework]

Couldn't agree more. :awe:

Nvidia have a lot to learn from AMD when it comes to mobile GPUs.

like how to not sell any?

 

[LTC8K6]

You care about performance and yet you don't care the faster tech? :whiste:

Let's assume like 15% faster but 2x the price, still idiotic to buy slower but cheaper product?

Has any reviewer isolated the HBM effect on Fury performance?

Is Fury faster because of HBM, or because it has a lot more Shaders/TMUs?

 

[Gikaseixas]

I don't see why this thread has had to become so partisan.

Nvidia fanboy - Nvidia using GDDR5X = AMD tech sucks, bigger, older, higher consumption doesn't matter

AMD fanboy - Nvidia using GDDR5X = GDDR sucks, bigger, older, higher consumption matters now

 

[littleg]

I don't see why this thread has had to become so partisan.

HBM has considerable performance advantages over DDR5 but is, currently, without its own problems, e.g., supply and cost. With this in mind it is not unreasonable for any company, Nvidia or AMD, to look to DDR5 for improvements if a reasonable supply of HBM is not on the horizon (this actually sounds pretty sensible, no?).

It would also be interesting to see an investigation in to memory bandwidth, what is currently required and what bottlenecks exist. Any one have any links?

I was looking at this article from techpowerup:

http://www.techpowerup.com/forums/threads/article-just-how-important-is-gpu-memory-bandwidth.209053/

The author litters the article with caveats about the reliability of the data though so i'm not sure how accurate the numbers really are. Seems quite difficult to get actual data about the bandwidth usage in a GPU.

 

[Erenhardt]

No one knows the cost of HBM integration vs GDDR5 yet everyone is shouting how expensive it is.
We know that on paper HBM is a lot faster than high-end 512bit GDDR5 config, all while using considerable amounts of power less.

To all the hypocrites: I guess you bashed every nvidia premium card as vigorously as you try to bash HBM.

No you didn't, you were rationalizing the premium cost with better performance or performance/watt of performance/smoothness or performance/dB etc...

I would really like to see the cost comparison between hbm and high speed GDDR5/GDDR5X

I was looking at this article from techpowerup:

http://www.techpowerup.com/forums/threads/article-just-how-important-is-gpu-memory-bandwidth.209053/

The author litters the article with caveats about the reliability of the data though so i'm not sure how accurate the numbers really are. Seems quite difficult to get actual data about the bandwidth usage in a GPU.

One thing is HBM performance, the other card using HBM.
We don't know how would fury perform, had it GDDR5 onboard, nor do we know how much more watts would it require.

 

[Techhog]

Nobody disputes that HBM is the future, but the implementation of HBM1 should temper anyone who believes we are looking at radical performance leaps in the early stages of the technology. We all heard the same propaganda pre-fury, how HBM(1) was going to change everything. It didn't.

Learn from that episode.

Yeah, let's just ignore the fact that HBM1 allowed AMD to release a card 25-40% faster than the 290X in the same power budget, and 15% faster than the 290X while using 50-100W less power, all while using essentially the same architecture, and somehow not be memory bottlenecked at with 4GB in some games where slower cards with GDDR5 are. Nope, it did nothing at all.

I wonder if you'd be saying the same if Maxwell had HBM1 and the 980 Ti were a 200W card.

 

[Techhog]

I don't see why this thread has had to become so partisan.

Welcome to *insert any tech forum here*!

 

[Mondozei]

Yeah, let's just ignore the fact that HBM1 allowed AMD to release a card 25-40% faster than the 290X in the same power budget, and 15% faster than the 290X while using 50-100W less power, all while using essentially the same architecture, and somehow not be memory bottlenecked at with 4GB in some games where slower cards with GDDR5 are. Nope, it did nothing at all.

I wonder if you'd be saying the same if Maxwell had HBM1 and the 980 Ti were a 200W card.

How much of that is due to new to the Fury arch and how much is due to HBM1? You're attributing all the gains to HBM1 and I wonder how you can source those claims. Let's see.

More worrying is the write-up implies that a GP104 may have 256-bit bus with 448GB/sec GDDR5X, while the flagship GP100 will be positioned as a professional Titan X/Tesla/Quadro product at first with HBM2 1TB/sec. That implies 3rd consecutive generation of milking the mid-range chip and making it a Marketing Flagship. D:

This is patently false and you know that. GM204 is seperate from GM200. Within each generation there are flagships. The GTX 980 Ti should be compared to the GTX 780 Ti(GM200 vs GK110) and the GTX 980 should be compared to the GTX 780(GM204 vs GK104).

GP104 will be compared to GM204 and it will be priced accordingly. There is no "3rd consecutive generation" argument here, because in order to do that, you must mix the x04s with the x10s and that is FUD.

It's not idiocy because you're not an idiot. You're spreading FUD when you make those "3rd consecutive generation" claims. You can only get there by mixing x04 with x10 and that's FUD.

 

[sontin]

Yeah, let's just ignore the fact that HBM1 allowed AMD to release a card 25-40% faster than the 290X in the same power budget, and 15% faster than the 290X while using 50-100W less power, all while using essentially the same architecture, and somehow not be memory bottlenecked at with 4GB in some games where slower cards with GDDR5 are. Nope, it did nothing at all.

I wonder if you'd be saying the same if Maxwell had HBM1 and the 980 Ti were a 200W card.

nVidia did better with GDDR5.
Yes, HBM allowed AMD to build a card which is faster and uses less power but they needed HBM for this.

nVidia has changed the architecture from Kepler and has released a whole family of chips from low-end to high-end.

nVidia is faster with GDDR, uses less power and has more memory.

 

[Dribble]

http://www.tweaktown.com/news/47976/nvidia-rumored-use-gddr5x-next-gen-pascal-based-gpus/index.html

Take it with salt.

Other rumour sites use the much less exciting headline:
NVIDIA Pascal GPUs Rumored To Feature GDDR5X Memory – Focused on Entry-Level Options, HBM2 For Performance and High-End Cards
http://wccftech.com/nvidia-pascal-g...ns-hbm2-mainstream-performance-highend-cards/

So the top end get HBM 2, the lower rungs gets GDDR5X. This is better for the mid range cards that would never have got HBM anyway as GDDR5X is better then GDDR5. I suppose the only question is what do the GTX 970/980 class cards get - HBM or GDDR5X?

 

[gamervivek]

I don't see why this thread has had to become so partisan.

HBM has considerable performance advantages over DDR5 but is, currently, without its own problems, e.g., supply and cost. With this in mind it is not unreasonable for any company, Nvidia or AMD, to look to DDR5 for improvements if a reasonable supply of HBM is not on the horizon (this actually sounds pretty sensible, no?).

It would also be interesting to see an investigation in to memory bandwidth, what is currently required and what bottlenecks exist. Any one have any links?

Maxwell most probably is considering the alacrity with which nvidia jumped on HBM and the problems securing the supply.

AT's 980 review had separate improvements by core and memory, but at 1440p.

http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/22

Some of the 980Ti overclocked reviews topping out at 20% improvement despite close to a 30% improvement in core clock at 4k. Or eurogamer flat out stating that an overclocked 980 didn't scale that well at 4k.

http://www.eurogamer.net/articles/digitalfoundry-2015-amd-radeon-r9-fury-review

But then there are reviews going against the above conclusions as well. :|

 

[tential]

You'd prefer? What difference does it truly make if it puts framerates on your puter? You don't truly care about the memory technology being used, because that would make no sense. What would make sense is for a gamer like to you prefer high framerates. And if those framerates are delivered via HBM or GDDR5X it should make no difference to you.

I find a lot of people who have complaints against Nvidia and Intel resort to complaining about Die Size, Memory Tech used, etc. when they can't complain about the actual performance being offered by the company.

 

[littleg]

One thing is HBM performance, the other card using HBM.
We don't know how would fury perform, had it GDDR5 onboard, nor do we know how much more watts would it require.

True but we're talking about low / mid range cards. They generally wouldn't have the processing power to require such massive throughput from the memory.

At the high end the cards can process the data much faster and so they need the extra bandwidth to keep the processing units supplied with work, for the low and mid range cards that's not the case.

 

[Redentor]

Other rumour sites use the much less exciting headline:
NVIDIA Pascal GPUs Rumored To Feature GDDR5X Memory – Focused on Entry-Level Options, HBM2 For Performance and High-End Cards
http://wccftech.com/nvidia-pascal-g...ns-hbm2-mainstream-performance-highend-cards/

So the top end get HBM 2, the lower rungs gets GDDR5X. This is better for the mid range cards that would never have got HBM anyway as GDDR5X is better then GDDR5. I suppose the only question is what do the GTX 970/980 class cards get - HBM or GDDR5X?

So, it's confirmed.

http://semiaccurate.com/forums/showthread.php?t=8757