Graphics Supply Constraints Solved – R9 Fury Series To See Wide Availability in Q4

[Kaapstad]

17 billion transistors on the new cards if the rumours are true.

That is for Big Pascal which I don't think we will see for a long time.

Small Pascal will be about 9 billion transistors which won't make it much faster than the GM200 cards. For anyone with a GM200 card not an upgrade worth waiting for.

Worse still Pascal could end up a bit like intel's latest few CPUs where ever decreasing size has concentrated the heat output in a smaller area making cooling and overclocking more difficult.

 

[LoadsaMoney]

Wonder if we'll also start to get the Fury non X below the £320 price of the 390X as well now, seen as its got 4GB less Vram.

Although dunno what they can do tbh, as the 390s have got 8GB as well, and they ain't much slower than the Xs, with a bit clock, they'll match em.

 

[r7slayer]

All you are crazy if you think any big pascal cards will be out soon. Will be midrange cards at most. a lot will buy this because its new tech and will appear amazing. Look at the 970 and 980 cards which came out they sold like crazy. Then the titan x and 980Ti which are also selling well. It will be low end and mid range first to get those sales in then about a year later which will be about 6 months after these mid range cards, we will see the bid daddy pascal come out so about 12 months maybe slightly longer from now. Don't forget that titan x and 980Ti will be currently much cheaper to produce than the bigger cards which are supposed to have HBM on them.

They would rather keep selling cheaper to produce top end cards then quickly release cards which cost more to produce and replace top end.

Yall mad if you think big end cards are comming soon.

 

[Gregster]

Well I bet I am closer to 5 months than Humbug is to 12 months

 

[tommybhoy]

I'd answer that but feel it would receive spin as an argument.

 

[Orangey]

The number of factors against a Q1 launch is huge.

So funny to see ignorant fanboys who know nothing about semiconductor production tell us otherwise.

 

[humbug]

The number of factors against a Q1 launch is huge.

So funny to see ignorant fanboys who know nothing about semiconductor production tell us otherwise.

Nooooo.... run it past FireStrik a few times and it'll be fine..... like lol.

 

[r7slayer]

I saw this before

Now that the rumor mill believes the GP100 GPU has been taped out it will likely soon be in the hands of NVIDIA’s engineering team for some last minute tweaks if needed. Once they ensure everything it okay they will give TSMC the green light to start mass producing wafers full of Pascal GPUs!
Read more at http://www.legitreviews.com/nvidia-pascal-gpu-with-hbm2-taped-out-gp100_165483#xXTYDM6ZSX7ckSV7.99

Not the usual site i get my news from but still.
As you can see a tape out doesn't mean ready to ship products rather more like a game in its beta stage. It's still in development but more a working prototype which is ready or next to ready to be mass produced and put into manufacture for products which are ready to ship. This isn't going to happen early 2016.

Also i'm sure i remember reading about HBM 2 being ready some time mid 2016 or rather Q2? That combined with time taken for pascal's production i just think early 2016 is too optimistic. Also AMD have priority to HBM from SK Hynix which is why nvidia was trying to source alternative sources of HBM such as that from Samsung. Is Samsung also ready to produce HBM 2.0?

 

[FoxEye]

The number of factors against a Q1 launch is huge.

So funny to see ignorant fanboys who know nothing about semiconductor production tell us otherwise.

Pot, kettle.

Q1 isn't as likely as Q2, but nobody here can say for a fact it's impossible.

And humbug's prediction of October 2016 for any Pascal cards is just laughable.

 

[r7slayer]

Pot, kettle.

Q1 isn't as likely as Q2, but nobody here can say for a fact it's impossible.

And humbug's prediction of October 2016 for any Pascal cards is just laughable.

True, there most likley would be your mid range cards first which from what i can tell will not be using HBM 2.0 so these cards don't have to wait on this. And most likley is what nVidia will push out first but time will tell only nVidia knows what they are doing lol.

 

[humbug]

Pot, kettle.

Q1 isn't as likely as Q2, but nobody here can say for a fact it's impossible.

And humbug's prediction of October 2016 for any Pascal cards is just laughable.

Thats Q4, i said Q3 thats July<, get your facts right.

 

[D.P.]

Except that's not how it works, is it? With everything engineered or built these days there, anything new is pretty accurately simulated before it even gets as far as being prototyped. In that talk JSH gave the other day they talked about the fact that one of the first things they did as a company was to buy out another startup that was simulating chips.

And since electronics either work or don't work (pretty binary outcomes, unlike testing an aeroplane's aerodynamics, for example), the idea that you might need to test for a year is still nonsense.

If the chips aren't getting so hot that they degrade, why would you want to test for longer than a couple weeks as DP said? It either works or it doesn't. They aren't using new materials with unknown properties are they.

Torture testing for a year -- lol, you made that up.

Exactly, certain posters are deluded as usual.

If a company y has to test a chip for a year before going Tinton production they will be bakrupt in no time while other companies get chips out in 5 months.

 

[D.P.]

lol....... you guys ^^^^^

PS: Maxwell is an evolution on the 28nm Kepler, 5 Months to tapeout an existing and already tested manufacturing process. 5 Months to tapout a new frock.

Pascal is much more similar to.maxwell than maxwell is to Kepler. Maxwell is an architecture that nvidia plan to use for years to come, pascal is the evoluriom, maxwell is the brand new architecture.

 

[Gerard]

Pascal is much more similar to.maxwell than maxwell is to Kepler. Maxwell is an architecture that nvidia plan to use for years to come, pascal is the evoluriom, maxwell is the brand new architecture.

Inventing words are ya?

 

[Stanners]

The number of factors against a Q1 launch is huge.

So funny to see ignorant fanboys who know nothing about semiconductor production tell us otherwise.

Who do you work for in the semiconductor business ?. Or is it a division of armchair expert?

 

[mmj_uk]

Who do you work for in the semiconductor business ?. Or is it a division of armchair expert?

Wikipedia.

 

[CAT-THE-FIFTH]

Even with the original large Kepler it was out nearly six months for commercial customers before the Titan was released.

Nvidia's high commercial cards are still based on a slightly tweaked Kepler GK110,the GK210, making the underlying tech around three years old.

Intel is releasing 14NM Xeon Phi using HMC soon:

http://www.itworld.com/article/2977101/hardware/article.html

So,unless its a 16NM pipecleaner part first with GDDR5X,if Nvidia is releasing a large Maxwell GPU with HBM2,who do people think they will target first??

A lower volume,higher profit margin market where Intel is going out full steam at,or a lower profit margin market with larger volumes where they are currently fine??

 

[Cambofrog]

I don't think NV will mess about getting Pascal out. After asyncgate, I don't think NV would like AMD cards to be perceived as having better support for DX12.
I'm not expecting NV to rush Pascal out, but I don't think they will be taking their time either

 

[D.P.]

Inventing words are ya?

Sorry, I was posting from my phone.

 

[mururoa]

Stocks improve but have the pump issues finally been resolved?

The pump issue plagued a very limited number of early production units and all problems were completely ironed out since early August.

Right, AMDMatt?