You are right there, any cuts to AMD's budget over the last 3-4 years will be felt in the next 2-3 years. We have already seen this with 200 series renamed to 300, lackluster 300, and a half-arsed Polaris lineup missing half the market segments and massively disappointing performance per watt. If their is an increase in GPU R&D budget (although i wonder if there is), it wont be know for another few generations.
-
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.
AMD VEGA confirmed for 2017 H1
- Thread starter Doom112
- Start date
- Status
You are right there, any cuts to AMD's budget over the last 3-4 years will be felt in the next 2-3 years. We have already seen this with 200 series renamed to 300, lackluster 300, and a half-arsed Polaris lineup missing half the market segments and massively disappointing performance per watt. If their is an increase in GPU R&D budget (although i wonder if there is), it wont be know for another few generations.
Caporegime
So basically your speculation trumps all - it's almost like you want them to spend less on GPU development for some reason. Only on ocuk forums do companies spend more on modifying existing uarchs then making new ones from scratch.
You might want to look at how often Intel has developed new uarchs from scratch as opposed to the iterative way of doing things. Intel has been conservative for a reason.
It's the same with lots of modern engineering - it's why we still use multistage rockets and not SSTO.
Also the fab market has changed as there has been major consolidations and Intel is hitting problems too. Intel's competitors are closer than they have been for years.
Ultimately amd has one window of opportunity next year TBH. It's a blessing in disguise that kl is 14nm and desktop 10nm chips wont be around for s while.
Plus regarding Vega you might want to go onto beyond3d and talk to some of the posters there - they appear to have more of a clue than what many are purporting to have here.
I am going by what is being discussed over there.
Last edited:
Caporegime
I think many here also seem to ignore part of the direct r and d spend has been covered by the console contacts too. Looking at how the PS4 PRO essentially had a down clocked rx480 in it with some additional features and it would not surprise me one bit the next GPU amd releases will be based on the Xbox Scorpio GPU.
This is why I think vega might have a gddr5 version unless the purported rx490 is using a bigger Polaris based chip. The Xbox uses GDDR5.
Then you get parts like Tonga which apparently had enhanced opencl performance which Apple got the best parts for and the Polaris 11 is more of the same with the die being made thinner and again Apple getting the best parts. It makes me wonder how much Apple was involved in that.
Last edited:
Where is this shortage for AMD? AMD have priority to HBM2 just because nVidia experienced a shortage doesn't mean AMD will. And nVidia still have released HBM2 cards. Why does this mean AMD will have to delay? As always you briefly read a article and post your own conclusions. So 1080Ti is releasing in January? Can you link me this info too? Or is that another conclusion pulled out of your.... Thin air?
Some (very) knowledgeable people there but if you are referring to what I think you are then keep in mind those same people literally days before Pascal launch were still claiming nVidia were going to be 6+ months later and struggling with yields and heat issues, etc. Just because they might be saying what you want to hear doesn't mean they are right
Although the half-missing segment does make one wonder, the part about disappointing performance per watt couldn't be further from what everyone else is seeing. If anything, AMD have hugely improved on their performance per watt ratios with Polaris, leaving negligible difference between each camp.
That in it self is a bit of an achievement given the high power requirements for the previously released GCN lineup and even going further back.
Caporegime
Some (very) knowledgeable people there but if you are referring to what I think you are then keep in mind those same people literally days before Pascal launch were still claiming nVidia were going to be 6+ months later and struggling with yields and heat issues, etc. Just because they might be saying what you want to hear doesn't mean they are right
They were talking about uarchs not yields and some of them seem to have been right on the money before(they obviously have worked in electronics engineering) - plus the last sentence is ironic since you obviously don't like what they are saying and want it to be as wrong as possible.Also as you know Polaris for coding has the same flags as Tonga and Fiji - people have found this out.
Vega apparently does not and I know many here are annoyed about it. So ergo it's a bigger change.
Last edited:
I know that isn't what I'm saying. There are some on there who can talk very knowledgeably about the low level details of a GPU and even have a good idea as to the shape of future stuff in that respect but don't have a great track record when it comes to the actual timeline, overall configuration i.e. number of SPs and VRAM, etc. or details of the node the core is made on but for some reason people are very quick to forget that.
Maybe my post isn't as clear as I think but would help if you actually understood my post rather than just picking random words out of it an assuming what I'm saying.
Caporegime
Which is again you now backtracking - I was talking about uarch design not yields and final configs,and what you don't seem to get,is that there have been leaks of Vega information elsewhere in databases. The flags for Polaris are the same for Tonga and Fiji and Vega isn't.
Ergo,a much bigger uarch change - if you really don't like that then complain to all the databases.
I think but would help if you actually understood my post rather than just picking random words out of it an assuming what I'm saying,and making up things like this:
which basically means:
YOU don't like what they are saying and are hoping Vega is a fail. The fact that you seem worried might be a good sign Vega is not too bad!
Last edited:
Where have I ever said anything that would indicate I hope Vega is a fail? again you are just taking what you want to see from my posts (like someone else here used to do) and putting your own assumptions on my stance and you are just as wrong as they were.
I'm simply urging caution as people around here seem very quick to forget the past.
I'm simply urging caution as people around here seem very quick to forget the past.
Caporegime
Nope,since you are the one making up assumptions about people and when called out about it,you are the one now backtracking.
I was referring to evidence that Vega is considered different from Polaris and Polaris was considered the same software wise as Tonga and Fiji.
Even an utter technophobe would mean that there are more changes going from Polaris to Vega.
That is what lead to your basic flamebait comment. I never talked about performance,or made any claims it would be 100000x better.
I just said it would be a bigger change. YOU made the assumption I was saying it was 100000x better.
Yet,unlike many here I have not made a prediction it would destroy a GTX1080TI? All leaks indicate it is basically a modified Fiji with higher clockspeeds and either HBM2 or GDDR5 memory. Now you do the maths and see where that might be - it fits in with the XBox Scorpio TFLOPs too especially if the part is sub 1GHZ which fits in what we have seen for the optimal effiency curves for all AMD GPUs made so far.
Just like the leaks from China from the same places hinted Polaris 10 was 2304 shaders and within 5% of a R9 390X which was broadly true at 1080p. Remember,unlike some of you I never purported Polaris 10 to be a GTX980TI beater. The most I ever said is it might get close to a Fury if the top end estimates of the clockspeed came true,which was increasingly dubious anyway.
Last edited:
I'm not back tracking at all (again you are just seeing in my words what you want to see) - as I said they might have a lot of knowledge about the low level stuff but they've been (in some cases very) wrong before and demonstrably so about the actual configuration that is launched - I'm not saying they are wrong just that be careful believing they are right just because what they say is what you want to see and your responses are just reinforcing how much you need to be cautious of that:
(EDIT: For example go back and read what some of them said about stuff like primitive discard accelerator - some great insight into the technical side of it but the actual implementation in Polaris in reality was quite different to the optimistic claims they made).
Show me anywhere in this thread I've said anything like that?
(EDIT: For example go back and read what some of them said about stuff like primitive discard accelerator - some great insight into the technical side of it but the actual implementation in Polaris in reality was quite different to the optimistic claims they made).
Show me anywhere in this thread I've said anything like that?
Last edited:
Caporegime
S Just because they might be saying what you want to hear doesn't mean they are right
Yes,you are - I am talking about the changes from Polaris and Vega and you made that comment. That is what they are talking about - not core configs,or core uarchs and it is based on what leaks we have seen about the Vega.
Its a bloody joke - you are attempting to bury the fact that Polaris has the same software flags as Tonga and Fiji and Vega does not.
You basically answered what I said about uarch changes with that post. Now you are doing as much as you can once you have been called out about.
Now you are saying to be cautious of what?? That Vega has bigger changes to the core. Have I talked anywhere here about what they bring to the table in terms of performance?? Can you show me where I might have said what advantages Vega might bring over Polaris. Nope.
Only that Tonga/Fiji to Polaris is a smaller change than Polaris to Vega.
Yet,I doubt if anybody said Volta was a bigger change than Pascal,nobody would question that,right?
Last edited:
Soldato
Surely this must be tiring for you two?
Anyway: Very much looking forward to Vega as that'll hopefully be the upgrade jump from my very power-hungry 390X
Caporegime
Common wisdom is Vega is a ground up new GPU, just like Zen is a ground up new CPU.
This is getting seriously frustrating - feels like we've been waiting for Vega and Zen for forever and we still have no idea how they are going to perform compared to the competition. Considering the GTX 1080 ti is launching soon (in January), now would be a good time for AMD to start showing some numbers. I really hope AMD won't make the same mistake they did with the Fury and set the launch price too high.
Caporegime
Its tiring since Rroff is just arguing about information that has been known for yonks. Polaris identifies the same as Tonga and Fiji.
Vega does not - so there are greater changes than with Polaris.
This is what forums like Beyond3D,etc found out and they are discussing what these changes might be.
So he answers it with posts like this:
S Just because they might be saying what you want to hear doesn't mean they are right
Basically I don't see him arguing so much in Nvidia threads that Volta probably is a bigger change than Pascal.
I have not even purported to try and put a prediction on what these changes might do,but Beyond3D is a good start we cannot simply look at Polaris and make the same performance estimations. AMD might focus those changes on improving performance or dropping power consumption.
He is talking about "caution" - I don't understand what he is going on about. Its not like I am saying Vega will be a GTX1080TI beater. We already have leaks of a "new" 4096 shader GPU being developed. So that either hints XBox Scorpio is probably that GPU in one form or another and so probably will one of the Vega chips. It might use GDDR5 or HBM2,and thats about it.
The XBox Scorpio uses GDDR5.
So if there is a prediction I have made it probably will be something more akin to a GTX1080 not a GTX1080TI for the smaller Vega chip,because I expect it will be a GDDR5/GDDR5X part with less bandwidth than a Fury X but with a higher core clockspeed and additional improvements,and IMHO I think any uarch improvements will be used to drop power and to pare down the chip as much as possible to make it cheaper to make so AMD can fight a price war with Nvidia better.
Nvidia with Maxwell did a very clever strategy - many lines were cheaper to make than their Kepler ancestors(GTX970 might have bucked the trend but I think Nvidia wanted to make sure they gained more people upgrading to a higher and more profitable price bracket).
AMD moved away from this and has had dies getting bigger and bigger.
I see them moving to smaller chips,where possible.
OFC,I might be off but that is at least what I see it probably being.
Last edited:
Caporegime
The 1080TI will be a $900 GPU, only about 1 in 5,000 give a #### about it.
Its not relevant.
Yes,you are - I am talking about the changes from Polaris and Vega and you made that comment. That is what they are talking about - not core configs,or core uarchs and it is based on what leaks we have seen about the Vega.
Its a bloody joke - you are attempting to bury the fact that Polaris has the same software flags as Tonga and Fiji and Vega does not.
You basically answered what I said about uarch changes with that post. Now you are doing as much as you can once you have been called out about.
Now you are saying to be cautious of what?? That Vega has bigger changes to the core. Have I talked anywhere here about what they bring to the table in terms of performance?? Can you show me where I might have said what advantages Vega might bring over Polaris. Nope.
Only that Tonga/Fiji to Polaris is a smaller change than Polaris to Vega.
Yet,I doubt if anybody said Volta was a bigger change than Pascal,nobody would question that,right?
Again you are inventing your own specifics into things I've posted that I don't even mention or allude to - seriously re-read your own posts and then try and find anything like for instance where I've made any mention of software flags or whatever directly or indirectly. My posts are both a lot more general and a lot less anti-AMD than you are perceiving them to be.
I've no interest in seeing Vega fail for instance as if nothing else nVidia need a kick up the rear to stop them trickle feeding consumers I'm merely being somewhat realistic.
EDIT: PS I've been very quiet in the Volta thread for a reason - it'll become more clear but remember what I said about people not understanding what Volta is all about :S
Last edited:
- Status