Indeed NVidia decided that GDDR5X was the best option for a high performance card.
How much faster is the 1080 than any single GPU Fiji/HBM based card again, answer quite a lot.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Radeon RX 480 "Polaris" Launched at $199
- Thread starter Kaapstad
- Start date
Indeed NVidia decided that GDDR5X was the best option for a high performance card.
How much faster is the 1080 than any single GPU Fiji/HBM based card again, answer quite a lot.
Indeed NVidia decided that GDDR5X was the best option for a high performance card.
How much faster is the 1080 than any single GPU Fiji/HBM based card again, answer quite a lot.
Surely by this logic 256-bit GDDR5 is better than 384-bit GDDR5, because the 1070 is faster than the 980 Ti?
I don't get what you have against HBM memory, but clearly the answer is a combination of GDDR5X with a 256-bit interface is cheaper than HBM2 and/or the 1080 doesn't have the processing power to saturate more memory bandwidth than GDDR5X can supply.
Nvidia (and any profit making company) will not produce 'the best thing' they possibly can. They will make the cheapest thing they possibly can for a given goal, in order to make the highest profit margin.
If HBM2 would have given the 1080 0-3% more overall performance in games, but would have lowered their profit margin 10%, they would not do it.
And obviously there are further considerations like time-to-market vs your competition, and customer perception (like not using HBM1 since they could only get 4GB and customers may not buy it, or want to pay less).
This entire thread is mostly speculation
I can tell why AMD don't like this stuff going around - If it turns out to be true, everyone will be saying I told you so, and if it's untrue the release will flop out of overhyped expectations even though they still have a decent card...
Either way... One hundred pages back this thread turned into 95% rumours and NVIDIA hate/fanboys
HBM1
Unreliable
Expensive
Poor 1080p performance
Overrated
Poor overclocking
Limited capacity
Not even AMD or NVidia will touch it for their up coming cards !!!
HBM2 will be an improvement but it is not the answer to everything that some people seem to think.
This to be honest. Once cards get loads more powerful and we start going to silly resolutions like 8k then it will probably be needed. But this will probably be a few years from now. At the moment, the bandwidth that HBM can offer is simply not needed 99% of the time.
In fact 12Gbps GDDR5X chips will probably more than enough for this generation of cards, even on the big boys.
Wound that back 10yrs for you. Luddite haha.
Calling kaap a luddite.... This thread is hilarious.
Yeah kaap is well known for hanging onto tech for as long as possible.
Soldato
- Joined
- 17 Aug 2003
- Posts
- 20,160
- Location
- Woburn Sand Dunes
furys just dont overclock for toffee. how do we know that, and the 'poor 1080p performance' are down to the MBH and not the gpu itself?
Err? 1500Mhz means that is in line for a FuryX if not faster, but having double the VRAM.
If so, my brother going to have early Christmas this June, receiving a Nano.
Calling kaap a luddite.... This thread is hilarious.
I was being facetious unless you hadn't twigged. Kap is fabulous at flip flopping when it suits him.
It doesn't matter. HBM was first utilised on the Fury cards which were (to some people) a complete flop.
Therefore HBM = bad technology, causes cancer, etc. FACT.
And because HBM2 is just more of the same it will also suck. FACT.
HBM(x) is failure, GDDR5X is best memory ever, nVidia uses it.
I'll restate what i posted on HOCP here;
nVidia jumped early with a paper launch on a big die straight off a new process - a month later and still #### all stock anywhere. They will NOT be making great margins on these cards as clearly the yields are truly awful and you can't fit as many dies on a wafer at that size.
They did it to steal AMD's thunder like they did previously. This time AMD have done a really good job keeping the leaks to a minimum and it would appear that nVidia are about to lose a good deal of the market in that AMD have pretty much all the consoles that matter, are winning OEM contracts hand over fist and people are starting to code accordingly.
It's going to be an ugly 24 months for nVidia and even worse going forward if things keep going the way they are.
nVidia jumped early with a paper launch on a big die straight off a new process - a month later and still #### all stock anywhere. They will NOT be making great margins on these cards as clearly the yields are truly awful and you can't fit as many dies on a wafer at that size.
They did it to steal AMD's thunder like they did previously. This time AMD have done a really good job keeping the leaks to a minimum and it would appear that nVidia are about to lose a good deal of the market in that AMD have pretty much all the consoles that matter, are winning OEM contracts hand over fist and people are starting to code accordingly.
It's going to be an ugly 24 months for nVidia and even worse going forward if things keep going the way they are.
For the n-th to the n-th time, the poor 1080p performance when Fiji launched was nothing to do with HBM. And it appears to be far better with the latest drivers and in dx12. Yes capacity is not great but they needed to implement it on a part so they could create a manufacturing chain.
Save your fingers we've been here before with the old boy. Sometimes people with more money than technical understanding think they reserve the right to state their incorrect opinions, even when people have tried to correct them.
We both know fiji's issues, i predicted fiji's 3 faults before release and they all came true.
Best not feed him the attention and ignore it again.
TSMC 16nm yields are good including for Pascal, not amazing but definitely not truly awful - people perpetuating that the yields are poor have no idea what they are talking about - nVidia is also lesser priority on volume production to some bigger names and having to split its allocation over various different product areas which stretches what they can produce for consumer cards a bit.
This is not a paper launch - with a paper launch only really reviewers get cards and maybe 1-2 trickle through retail if you are lucky - there is supply of the 1000 series cards albeit not enough for the demand or massive quantities but most places are receiving 10+ a week (in some cases of each brand) but a lot of that is still going into fulfilling pre-orders - I'm sure someone from OcUK can give a better picture of what their supply levels are like.
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Even if Vega come along with HBM2 and smashed Nvidia to bits Kaap would still find fault with it somewhere, probably state the card would be even better with gddr5x etc
And then probably default back to Fiji to backup his argument
Everyone knows Fiji was terrible but I don't blame the memory for it, intact the memory is what helps it at higher res
And then probably default back to Fiji to backup his argument
Everyone knows Fiji was terrible but I don't blame the memory for it, intact the memory is what helps it at higher res