We have 14nm Vega until the mid-2019 arrival of Navi on 7nm - How about a "Gigahertz Edition" redo?

[Pieter3dnow]

I would like a Vega Nano 56 and 64. Just undervolted and underclocked out of the box.

You prepared to pay for it .

 

[Shadowed]

You prepared to pay for it .

Not at all. But it would make Vega seem a lot more power efficient in benchmarks.

 

[ccityinstaller]

I would like a Vega Nano 56 and 64. Just undervolted and underclocked out of the box.

I can get you a VEGA 56 Nano right now...My first 56 will clock @ 1408Mhz/1100Mhz HBM @ .9V and draw less then 140W while doing so. Need more power you say? Well clock it up to 1575/1100Mhz @ 1.05V and you are good to go..The current crop of V56s can easily do it on 8/10 samples I would be willing to wager. Just got another 64 with EK block for $680 on eBay...Gonna give it some undervolt/overclock love and see what it will do before it starts slaving away in the mines.

 

[KazeoHin]

I can get you a VEGA 56 Nano right now...My first 56 will clock @ 1408Mhz/1100Mhz HBM @ .9V and draw less then 140W while doing so. Need more power you say? Well clock it up to 1575/1100Mhz @ 1.05V and you are good to go..The current crop of V56s can easily do it on 8/10 samples I would be willing to wager. Just got another 64 with EK block for $680 on eBay...Gonna give it some undervolt/overclock love and see what it will do before it starts slaving away in the mines.

Vega, Like Ryzen, hits a wall where you get diminishing returns for extreme voltage.

It runs great at lower voltages, though. The problem is that the R9 fury series did the same thing, and now they're pretty much garbage cards.

 

[Shadowed]

Problem is Vega shipped at such high voltages that it made it seem far less efficient than it can be. The reviews for Vega made it seem quite unappealing. A factory undervolted version would have made it seem efficient, just like Fury Nano did.

But it is too late for this imo.

 

[Maddness]

Problem is Vega shipped at such high voltages that it made it seem far less efficient than it can be. The reviews for Vega made it seem quite unappealing. A factory undervolted version would have made it seem efficient, just like Fury Nano did.

But it is too late for this imo.

The thing is, not all Vega cards will run stable with undervolting. There have been a couple of members here that have tried it and run into stability issues.

 

[Pieter3dnow]

Not at all. But it would make Vega seem a lot more power efficient in benchmarks.

For your entertainment

 

[SighTurtle]

I was listening to AMD's earnings call today, unless AMD is keeping it close to their chest, it sounds like they are content with their current 14nm Polaris and Vega products until next year. The only future GPU product mentioned was the 7nm machine learning oriented Navi chip currently being worked on with TSMC. GlobalFoundries, if I heard correctly, will be making 7nm Epyc/Ryzen products at the least. Probably too much to hope for consumer based 7nm Navi parts from TSMC.

Hopefully they can finally bring Polaris and Vega back to MSRP, as Nvidia has reached back to MSRP some places.

 

[Maddness]

Yes that is what i have heard to.

 

[NKD]

I was listening to AMD's earnings call today, unless AMD is keeping it close to their chest, it sounds like they are content with their current 14nm Polaris and Vega products until next year. The only future GPU product mentioned was the 7nm machine learning oriented Navi chip currently being worked on with TSMC. GlobalFoundries, if I heard correctly, will be making 7nm Epyc/Ryzen products at the least. Probably too much to hope for consumer based 7nm Navi parts from TSMC.

Hopefully they can finally bring Polaris and Vega back to MSRP, as Nvidia has reached back to MSRP some places.

I think AMD is going the same route as nvidia now. Release a bigger chip for non gaming related and then gaming related navi. If they are releasing navi for machine learning first it seems like they are going that route. I am just happy they are moving navi to 7nm on TSMC. That in itself is good news.

 

[Pieter3dnow]

I was listening to AMD's earnings call today, unless AMD is keeping it close to their chest, it sounds like they are content with their current 14nm Polaris and Vega products until next year. The only future GPU product mentioned was the 7nm machine learning oriented Navi chip currently being worked on with TSMC. GlobalFoundries, if I heard correctly, will be making 7nm Epyc/Ryzen products at the least. Probably too much to hope for consumer based 7nm Navi parts from TSMC.

Hopefully they can finally bring Polaris and Vega back to MSRP, as Nvidia has reached back to MSRP some places.

Maybe content is not the right expression they booted people from RTG there trying to get the ball rolling and they can't seem to produce a lower nm version of anything they already made (in other words waste of time and money). That is the same approach we seen from AMD when it concerned their CPU. Rather then banging your head against the wall on products you have trouble making and selling.

My only gripe with that attitude is what are they doing in the PR department because they hired a lot of people. These better earn their money when it comes to launching new products.

 

[Pieter3dnow]

I think AMD is going the same route as nvidia now. Release a bigger chip for non gaming related and then gaming related navi. If they are releasing navi for machine learning first it seems like they are going that route. I am just happy they are moving navi to 7nm on TSMC. That in itself is good news.

Here is the thing AMD can't be happy with Navi for various reasons (why else would you skip a consumer version). It is limited by design and by drivers?

 

[Jedibeeftrix]

when did we arrive at the point where we are skipping a consumer version of navi?
the bigger chip for non-gaming is understood to be vega 20, not navi 20.

 

[NKD]

Here is the thing AMD can't be happy with Navi for various reasons (why else would you skip a consumer version). It is limited by design and by drivers?

I don’t think anyone ever claimed they are skipping gaming version. That is just total speculation on your part.

 

[gigaxtreme1]

It's the profit margin in HPC/AI they target.

 

[Pieter3dnow]

when did we arrive at the point where we are skipping a consumer version of navi?
the bigger chip for non-gaming is understood to be vega 20, not navi 20.

This year no consumer version of Navi.

I don’t think anyone ever claimed they are skipping gaming version. That is just total speculation on your part.

Navi was not confirmed for consumers this year was it? It is simple if it is good there is no need for AMD to keep it for instinct this year.

It's the profit margin in HPC/AI they target.

Maybe less embarrassing then what happened with Vega only 1 day where people could buy it for a decent price.
You recall how they had a frantic Vega media campaign where all of the new features for the Vega card were mentioned and discussed and even to date some of them aren't even implemented (software).

 

[NKD]

This year no consumer version of Navi.

Navi was not confirmed for consumers this year was it? It is simple if it is good there is no need for AMD to keep it for instinct this year.

Maybe less embarrassing then what happened with Vega only 1 day where people could buy it for a decent price.
You recall how they had a frantic Vega media campaign where all of the new features for the Vega card were mentioned and discussed and even to date some of them aren't even implemented (software).

Navi is coming to different market segment first. At no point it is ever said that there will never be a consumer version of Navi. Nvidia does this all the time. They release to higher end market first and then to consumer. If AMD said we will never have consumer version of Navi then you have a point but that’s not the case. Nvidia does this every time, nothing different if AMD takes the same route.

 

[Jedibeeftrix]

Again, where is there any suggestion that the initial Navi offering will [not] be a consumer part?
Vega 20 is said to be a AI/ML part, with a focus on FP64, but that is Vega. Not Navi.

 

[NKD]

Again, where is there any suggestion that the initial Navi offering will [not] be a consumer part?
Vega 20 is said to be a AI/ML part, with a focus on FP64, but that is Vega. Not Navi.

I didn't hear the call. But from what I have been reading it was vega on 7nm. I am unsure where navi came from. Everyone reported it was vega. Not to say that Navi may not be sampling by the end of the year but no one reported navi specifically.

 

[Pieter3dnow]

The past RTG events where they would talk about Navi and saying it would be the first where Koduri would make the difference.
So when you have a winner first thing you do is limit it to the professional segment?
You could conclude the following: the part is to slow for the consumer market.
There some other things spring to mind but nothing positive about the situation regarding the consumer segment. Because if AMD has any positive news regarding RTG and consumer parts they would be screaming it from the rooftops.

 

[IdiotInCharge]

Because if AMD has any positive news regarding RTG and consumer parts they would be screaming it from the rooftops.

I'd forgive them for being quiet after Raja's Vega folly. Over-hyping is a bad thing and they probably want to make absolutely sure that they get the messaging on their next consumer release right.

 

[N4CR]

The past RTG events where they would talk about Navi and saying it would be the first where Koduri would make the difference.
So when you have a winner first thing you do is limit it to the professional segment?
You could conclude the following: the part is to slow for the consumer market.
There some other things spring to mind but nothing positive about the situation regarding the consumer segment. Because if AMD has any positive news regarding RTG and consumer parts they would be screaming it from the rooftops.

AMD doesn't do the Raja-Like claims any more. Lisa stopped that.
Navi AI machine learning market will be a more profitable pipe cleaner for 7nm, instead of doing this the FE, then small scale production gaming route.

 

[Maddness]

AMD doesn't do the Raja-Like claims any more. Lisa stopped that.
Navi AI machine learning market will be a more profitable pipe cleaner for 7nm, instead of doing this the FE, then small scale production gaming route.

Thank god for that. There was nothing worse than hyping up a product before launch and then it turns out to be not up to it.

 

[Pieter3dnow]

Thank god for that. There was nothing worse than hyping up a product before launch and then it turns out to be not up to it.

The hype is not the problem it was just not very proportional to the product being released.

AMD doesn't do the Raja-Like claims any more. Lisa stopped that.
Navi AI machine learning market will be a more profitable pipe cleaner for 7nm, instead of doing this the FE, then small scale production gaming route.

AMD doesn't do high end videocards either but there is something they should do that is be a little bit more honest to their goals for consumers.
If you want to call Navi a pipe cleaner then I'll take that maybe it will be just that.

 

[N4CR]

AMD doesn't do high end videocards either but there is something they should do that is be a little bit more honest to their goals for consumers.
If you want to call Navi a pipe cleaner then I'll take that maybe it will be just that.

I'm meaning more the initial product run pipe cleaner as they target high margin sectors. Then alter or strip it down for dGPU users.
AMD skirts high end or is high end in at least one title. But yes they need a consistent, competitive halo effort. We may not see that till MCM.
Keep in mind Vega supposedly already has 500gb/sec IF interconnects which are enough for testing..

 

[TheBlueChanell]

WCCFTech has some prelim benches that's showing 7nm Vega being up to 70% faster clock-for-clock than V64. Take that with a huge grain of salt since it's both an engi sample and WCCFTech. Would be spectacular if true but I'd really like to know what was done to gain said performance if true. It does align with the rumors that Lisa Su put together a skunkworks project for RTG to immediately and quickly give their hardware a shot in the arm to increase clocks, power efficiency / ppw AND that we could see the first fruits of that by the end of the year.

Regardless, 7nm is going to be exciting for everything. nVidia/AMD Intel/AMD etc. Hardware market was getting so boring, shits really gotten exciting starting with the launch of Ryzen last year. I feel like ever since the 20nm fiasco @ TSMC the whole market was on a boring downward spiral up until early 2017.

 

[TheBlueChanell]

The hype is not the problem it was just not very proportional to the product being released.

AMD doesn't do high end videocards either but there is something they should do that is be a little bit more honest to their goals for consumers.
If you want to call Navi a pipe cleaner then I'll take that maybe it will be just that.

That's not really a true statement. Vega 64/56 are high-end cards, they soundly compete with the 1070/Ti -> 1080. It's just the 1080 Ti outclasses everything. It was late but they most certainly compete in the high-end space and will continue to do so for the forseable future, albeit hopefully much more aggressively. The only time they really haven't was the lul between Fury X -> Vega which wasn't the initial plan either. A number of factors resulted in said lul/gap and caused Vega to be very late.

I don't think they intentionally misled either, as a response to your first statement, they just under delivered and a lot of that was Raja's overly-enthusiastic presentations. Look at their most recent CPU roll-outs as an example of the opposite.

Navi is not a pipe-cleaner either, not too sure why the person you replied to said/thinks that, they have specifically stated that the 7nm Vega Instinct is gonna be the pipe cleaner of their GPU roadmap. I do think Navi will be something special, it's going to further expand upon the IF-minded design of Vega and I believe we will see something akin to R700's design philosophy w/o all of the baggage and bullshit that has plagued both camps mGPU setups since their inception. DX12/Vulkan mGPU lends itself to my line of thinking here as well.

 

[Maddness]

WCCFTech has some prelim benches that's showing 7nm Vega being up to 70% faster clock-for-clock than V64. Take that with a huge grain of salt since it's both an engi sample and WCCFTech. Would be spectacular if true but I'd really like to know what was done to gain said performance if true. It does align with the rumors that Lisa Su put together a skunkworks project for RTG to immediately and quickly give their hardware a shot in the arm to increase clocks, power efficiency / ppw AND that we could see the first fruits of that by the end of the year.

Regardless, 7nm is going to be exciting for everything. nVidia/AMD Intel/AMD etc. Hardware market was getting so boring, shits really gotten exciting starting with the launch of Ryzen last year. I feel like ever since the 20nm fiasco @ TSMC the whole market was on a boring downward spiral up until early 2017.

I'd say it's more than likely the clocks have been misread. I can't see the 7nm Vega being 40% faster than the 1080Ti. Don't get me wrong, I for one would absolutely love that to happen. But I can't believe that is even possible with Vega, even with a large die shrink. IMO if it matches the 1080Ti, then that is a win in my books.

 

[KazeoHin]

I'd say it's more than likely the clocks have been misread. I can't see the 7nm Vega being 40% faster than the 1080Ti. Don't get me wrong, I for one would absolutely love that to happen. But I can't believe that is even possible with Vega, even with a large die shrink. IMO if it matches the 1080Ti, then that is a win in my books.

Matching the 1080 Ti this year is like matching the 1080 last year, in other words, it means nothing.


If Vega could clock at the same speeds as a 1080, it would wipe the floor with the Ti, so they don't really need to improve IPC much, just improve clockspeed and power draw.

Combine a 1900GHz Vega 64 with a ~10% IPC bump, wrap it all up in a 250w package and you have a proper high-end card.

Regardless, I really hope 7nm Vega unlocks some sort of hidden potential of the architecture, AMD needs it, and badly so.

 

[Maddness]

GCN has never been able to clock close to what Nvidia can do. I wouldn't hold out hope for a 1900mhz Vega chip. It would be pretty damn sweet, but not a reality I wouldn't think. I also can't see a die shrink and some IPC improvements netting a 70% increase. I also think that's not a possibility. I'm more interested to see what Navi is going to bring to the table.

 

[Pieter3dnow]

That's not really a true statement. Vega 64/56 are high-end cards, they soundly compete with the 1070/Ti -> 1080. It's just the 1080 Ti outclasses everything. It was late but they most certainly compete in the high-end space and will continue to do so for the forseable future, albeit hopefully much more aggressively. The only time they really haven't was the lul between Fury X -> Vega which wasn't the initial plan either. A number of factors resulted in said lul/gap and caused Vega to be very late.

I don't think they intentionally misled either, as a response to your first statement, they just under delivered and a lot of that was Raja's overly-enthusiastic presentations. Look at their most recent CPU roll-outs as an example of the opposite.

Navi is not a pipe-cleaner either, not too sure why the person you replied to said/thinks that, they have specifically stated that the 7nm Vega Instinct is gonna be the pipe cleaner of their GPU roadmap. I do think Navi will be something special, it's going to further expand upon the IF-minded design of Vega and I believe we will see something akin to R700's design philosophy w/o all of the baggage and bullshit that has plagued both camps mGPU setups since their inception. DX12/Vulkan mGPU lends itself to my line of thinking here as well.

Most of the people here have a misconception of what high end means. It means the top of the silicon the fastest around (what you seen Nvidia do with Titan). Not the mid range parts that Nvidia sells as high end nor what AMD considers high end (for desktop consumers).
Look at Vega why is it not high end , run Turbo mode check the amounts of Watt used then you will think twice about saying anything regarding that Vega is high end. You get 1 fps more from a boost of 100+ Watt from the previous driver mode, that is how badly things are.

This problem for AMD where you crank up the power and it fizzles is so sad when you see the competition still using GDDR based solution and use less power. This why there is no reason to have any of the thoughts you are having.

If Navi is special then what better market to put it on then the desktop consumer space? Without solving the really bad power requirements Navi will miss the target for the consumer space it will be to slow and use to much power to keep up with (by then over 2 year old) Nvidia midrange products.

Navi would have to according to fudzilla run at 1080 levels and thus would have to beat that power requirement levels from memory the Vega 56 ran 40 watt more then the GTX 1070 and that is as close as it came. Lower clocks better performance higher clocks a nightmare on power.

And then realize how long ago your competition did this (May 2016) . I am still wondering how any one in their right mind can think that Navi will be anything then a filler until AMD finally starts to solve problems (they kicked RTG personnel around so it will only take 3 years?).

 

[NKD]

Most of the people here have a misconception of what high end means. It means the top of the silicon the fastest around (what you seen Nvidia do with Titan). Not the mid range parts that Nvidia sells as high end nor what AMD considers high end (for desktop consumers).
Look at Vega why is it not high end , run Turbo mode check the amounts of Watt used then you will think twice about saying anything regarding that Vega is high end. You get 1 fps more from a boost of 100+ Watt from the previous driver mode, that is how badly things are.

This problem for AMD where you crank up the power and it fizzles is so sad when you see the competition still using GDDR based solution and use less power. This why there is no reason to have any of the thoughts you are having.

If Navi is special then what better market to put it on then the desktop consumer space? Without solving the really bad power requirements Navi will miss the target for the consumer space it will be to slow and use to much power to keep up with (by then over 2 year old) Nvidia midrange products.

Navi would have to according to fudzilla run at 1080 levels and thus would have to beat that power requirement levels from memory the Vega 56 ran 40 watt more then the GTX 1070 and that is as close as it came. Lower clocks better performance higher clocks a nightmare on power.

And then realize how long ago your competition did this (May 2016) . I am still wondering how any one in their right mind can think that Navi will be anything then a filler until AMD finally starts to solve problems (they kicked RTG personal around so it will only take 3 years?).

You seem to taking rumors and basing your opinion on them as they are facts. You don’t know what Navi is going to bring until it does. That’s the truth rest us speculation.

 

[Algrim]

You seem to taking rumors and basing your opinion on them as they are facts. You don’t know what Navi is going to bring until it does. That’s the truth rest us speculation.

I, for one, would love RTG to produce a non-sucky Navi-based GPU product. We need competition in the GPU space, just as much as we needed a good Ryzen/Threadripper/Epyc lineup to compete against Intel. Navi could surprise us but I think expecting anything great out of Navi is setting yourself up for disappointment. I would be delighted to be proven wrong but I won't hold my breath on it. Lisa Su pretty much just started managing RTG and it takes time for those results to be seen as production silicon.

 

[NKD]

I, for one, would love RTG to produce a non-sucky Navi-based GPU product. We need competition in the GPU space, just as much as we needed a good Ryzen/Threadripper/Epyc lineup to compete against Intel. Navi could surprise us but I think expecting anything great out of Navi is setting yourself up for disappointment. I would be delighted to be proven wrong but I won't hold my breath on it. Lisa Su pretty much just started managing RTG and it takes time for those results to be seen as production silicon.

Never expected anything for Navi! I just said don’t base assumption on rumors to the original poster. Other than rumor from fudzilla there is nothing we have heard from any credible source. Until Kyle says Navi is bust I won’t pass any judgement on it.

 

[Pieter3dnow]

You seem to taking rumors and basing your opinion on them as they are facts. You don’t know what Navi is going to bring until it does. That’s the truth rest us speculation.

I'm not taking any rumours anywhere. I'm saying based on the time from the R9 290X the last videocard I purchased the same routine takes place.
AMD releases card , no high clock when high clock massive power needed.
If you checked Buildzoid analysis of Vega he guessed that the same Vega card would draw near 500 Watt with GDDR5
It takes some gifted people to deal with these problems, you seen the personal changes in RTG haven't you they did not do that for show.

Unless AMD hired someone with magical pixie dust and sprinkled it around the gpu engineers there won't be anything out for a good while. At least it would be highly unlikely to happen within 2 years.

 

[luminousone]

Well just from looking at store listings getting dropped, availability(on average) declining, all the while the price is also declining, makes we wonder if they are clearing the supply chain. egg only has 4 vega 64 card listings atm(excluding open box) including third party sellers. Vega fe is listed as being no longer available now as well.

Even if the chip doesn't change perhaps a refresh using faster hbm2 or maybe moving all cards to 16gb. or perhaps lc hbm has quietly gotten ready(they talked about this last year). kinda like 290->390 "upgrade".

 

[ccityinstaller]

Well just from looking at store listings getting dropped, availability(on average) declining, all the while the price is also declining, makes we wonder if they are clearing the supply chain. egg only has 4 vega 64 card listings atm(excluding open box) including third party sellers. Vega fe is listed as being no longer available now as well.

Even if the chip doesn't change perhaps a refresh using faster hbm2 or maybe moving all cards to 16gb. or perhaps lc hbm has quietly gotten ready(they talked about this last year). kinda like 290->390 "upgrade".

That's is interesting. All of my 56s (flashed with the bios of a LC 64) will do 1100Mhz on their HBM mining/gaming stable aside from one which will happily do 1090Mhz all day everyday.

The single 64 I have will only do 1050Mhz. Anything over that is stable mining but it will cause minor artifacts if a display is attached. It's worth noting all my cards have Samsung ram. I've never tried a card with any other vendor of HBM.

The VEGAs love that fast ram though. The difference in min frame rates is insane, along with the avg fame times in regards to gaming.


As far as mining goes, going from 950 to 1100 nets me 20+% for less then 10W more power usage.


I would love to see a fully WC'd VEGA 56 on the 7nm node. If AMD could squeeze out 1950Mhz with good yields I would be happy. I'd love to see a 1080TI killet simply because the competition is great for us consumers, but a 1950Mhz 56/64 with 12GB of HBM would be perfect for a 4K FreeSync'd setup.

 

[Pieter3dnow]

That's is interesting. All of my 56s (flashed with the bios of a LC 64) will do 1100Mhz on their HBM mining/gaming stable aside from one which will happily do 1090Mhz all day everyday.

The single 64 I have will only do 1050Mhz. Anything over that is stable mining but it will cause minor artifacts if a display is attached. It's worth noting all my cards have Samsung ram. I've never tried a card with any other vendor of HBM.

The VEGAs love that fast ram though. The difference in min frame rates is insane, along with the avg fame times in regards to gaming.


As far as mining goes, going from 950 to 1100 nets me 20+% for less then 10W more power usage.


I would love to see a fully WC'd VEGA 56 on the 7nm node. If AMD could squeeze out 1950Mhz with good yields I would be happy. I'd love to see a 1080TI killet simply because the competition is great for us consumers, but a 1950Mhz 56/64 with 12GB of HBM would be perfect for a 4K FreeSync'd setup.

We all would love to see something new and exciting from AMD. Imagine how simple things can be if they only could manage it what you write here, something the competition just did but AMD did not manage this ever wonder why?

If the approach towards the GPU would be as simple as limitations of the production process AMD could get much better scaling just do it on lower nm process and you are done.

So something is stopping AMD from doing the obvious, what is it?

 

[Shop_vac]

We all would love to see something new and exciting from AMD.

So something is stopping AMD from doing the obvious, what is it?

They can't trust people to buy it.

When your profit margin in gaming is propped up by a mining craze that could vanish into thin air at any moment, you might second guess the intended market too. Also, a lot of AMD fans seemed to sit Vega out, even when it was affordable.

Since Vega was potentially going to sell at a loss until rebates gave partners breathing room, it might be more expensive than they think it's worth (strictly as a gaming part).

 

[ccityinstaller]

We all would love to see something new and exciting from AMD. Imagine how simple things can be if they only could manage it what you write here, something the competition just did but AMD did not manage this ever wonder why?

If the approach towards the GPU would be as simple as limitations of the production process AMD could get much better scaling just do it on lower nm process and you are done.

So something is stopping AMD from doing the obvious, what is it?

I wouldn't buy into the "we have abandoned the gaming market" nonsense. If AMD can make a profit, even a small one, then they will be there.


Chasing the HPC market is what they need to do in order to rebuilt their R&D budgets. Lisa Su likes to compete. More importantly, she likes to win. Given the shakeups she has overseen, coupled with the wonderful success Ryzen has been, all points toward AMD coming out swinging. It might take 18-24 months but I believe it will happen.

I'm actually putting my money where my mouth is abd picking up more AMD stock. It has no where to go but up. Way up.