** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[Stanners]

I wonder if this is what Nvidia are going to use on the **80 and below cards while saving HBM for the Ti and Titan cards. Or it could all be rubish

 

[nashathedog]

There's nothing surprising there really, It's a smart move, Gddr5 is still okay and like I mentioned a while ago I can't see either brands cards using all hbm2. A mix of hbm1 and hbm2 cards makes sense and if they also have another range of gddr5 with better bandwidth why not.

 

[Marine-RX179]

8 Pack blew the low ROPs count theory out of the water when he benched a Fiji card with the memory running over 1000mhz and getting good gains from doing it. This highlights the fact that HBM1 running @500mhz is way too slow.

Ah right. Haven't really been following his testing. It's good to know.

 

[humbug]

8 Pack blew the low ROPs count theory out of the water when he benched a Fiji card with the memory running over 1000mhz and getting good gains from doing it. This highlights the fact that HBM1 running @500mhz is way too slow.

What did he get from a 100% memory overclock?

If its more than 40% i could believe its the memory bottlenecking the card but i doubt that.

 

[P.B]

more importantly what did he do to get 100% over clock ?

 

[Marine-RX179]

more importantly what did he do to get 100% over clock ?

Something that none of us can?

The title of No.1 Overclocker is not just for show

 

[P.B]

lol very true must have put it under ln2

 

[Marine-RX179]

lol very true must have put it under ln2

Yea...LN2 is great for quick bench run, but they are not actually consumer level to be use for everyday gaming...and also I'm not sure how many people will be comfortable gaming near LN2

 

[Boomstick777]

Can def see Nvidia doing this, maybe X80, X80 Ti and Titan Pascal will be all be based on HBM 2.0. While everything below will use GDDR5X.

 

[Jeps]

So GP100 cards will have HBM2 and GP104 cards will have GDDR5x. Makes a lot of sense really.

 

[CAT-THE-FIFTH]

It does make me wonder which rumours of HBM2 allocation were true!

 

[Marine-RX179]

Can def see Nvidia doing this, maybe X80, X80 Ti and Titan Pascal will be all be based on HBM 2.0. While everything below will use GDDR5X.

Yea skinny GPU=GDDR5X, and full fat GPU=HBM2 looks likely.

 

[Freddie1980]

8 Pack blew the low ROPs count theory out of the water when he benched a Fiji card with the memory running over 1000mhz and getting good gains from doing it. This highlights the fact that HBM1 running @500mhz is way too slow.

Was the Fiji card giving FPS performance in excess of the HBM card? From what I understand clockrate on video memory can't be looked at in isolation as it needs to be measured against the bus which gives you the overall memory bandwidth. Sort of like PSU's were you determine it's wattage by multiplying the volts by the amps.

 

[bru]

Surely the chips would need to have different memory controllers built in, so they would have been designed like this from the start and yet this is the first rumour we hear about it. I calling billyhooks on the whole thing. HBM across the board from both camps for next gen I reckon.


Disclaimer: I reserve the right to be completely wrong about this..

 

[Gregster]

Makes sense for the mid/low cards to come with GDDR5. Why add all that expense onto those range of cards?

 

[varkanoid]

Maybe the decision has come about due to availability. If GDDR5X is more plentiful than HBM2 then Nvidia will take the decision to go with whatever will allow them to maximise sales.

Better improvement with HBM2 yet low availability or a smaller improvement with GDDR5x but high availability. Nvidia always want to shift high volume so if GDDR5x allows this then they will go for it. Still will be an improvement over GDDR5.

Dont blame them. AMD should think the same then they may see sales improve in 2016. Know thy enemy and all that.

 

[queamin]

Surely the chips would need to have different memory controllers built in, so they would have been designed like this from the start and yet this is the first rumour we hear about it. I calling billyhooks on the whole thing. HBM across the board from both camps for next gen I reckon.


Disclaimer: I reserve the right to be completely wrong about this..

I agree with this and the design would have been done ages ago, they cannot just change the design at this late stage as it would set them back months.

 

[Boomstick777]

Makes sense for the mid/low cards to come with GDDR5. Why add all that expense onto those range of cards?

Maybe the decision has come about due to availability. If GDDR5X is more plentiful than HBM2 then Nvidia will take the decision to go with whatever will allow them to maximise sales.

Better improvement with HBM2 yet low availability or a smaller improvement with GDDR5x but high availability. Nvidia always want to shift high volume so if GDDR5x allows this then they will go for it. Still will be an improvement over GDDR5.

Dont blame them. AMD should think the same then they may see sales improve in 2016. Know thy enemy and all that.

Yeah if the availability and expense of adding HBM 2.0 to non top end cards is going to slow things down, then GDRR5X would absolutely makes sense, and leave HBM 2.0 for the higher end cards. That would mean more availability across the whole range until HBM matures and becomes readily available and less expensive. Either way, not really bothered if Nvidia or AMD do something like this. Anything to speed up progress and as long as solid availability of the higher end stuff it's all good.

 

[Insanties_birth]

Yeah if the availability and expense of adding HBM 2.0 to non top end cards is going to slow things down, then GDRR5X would absolutely makes sense, and leave HBM 2.0 for the higher end cards. That would mean more availability across the whole range until HBM matures and becomes readily available and less expensive. Either way, not really bothered if Nvidia or AMD do something like this. Anything to speed up progress and as long as solid availability of the higher end stuff it's all good.

Can't wait for the high end to have low availability which leads to scorching levels of price gouging and more experiments on new pricing levels. Just hope HBM isn't the ideal excuse some companies would like to put prices ever higher.

We can always wait and see but I just hope they iron out the kinks and that availability isn't quite the same issue it's been with the fury cards. Don't fancy paying an extra 100 or so just for the sake of it being hard to keep in stock and retailers taking advantage.

 

[D1craig]

Do any of you remember before the gtx 900s where released? I'm sure they where suppose to be on a die shrink but pretty close to their release they decided against it.

This all sounds familiar.