• 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 7nm GPU News and Rumours 2018/2019

Why do we get ourselves so in a twist over this stuff? We absolutely don't know, but I have already started to guess so waffle for waffles sake...

Instead of concentrating on a couple of mentions on a slide released 6 months ago about a card that will probably be released in almost a year - what is the design brief going to be on this architecture?

AMD have publicly declared Navi is a Mid Tier card - exactly how they define Polaris. They already have a price bracket in mind for that probably at the $200-$300 price point.

There has been significant noise about Navi going into the PS5. I believe that one of Sony's must haves will be the ability to game at 4k 60Hz at medium to high settings. I think that will mean hitting close to 1080 levels of performance.

They are also going to have to be able to create the chip to hit a console defined price point which I think also backs up the $2-300 number.

Therefore based purely on what I think the Engineering design brief has to be I think we are going to be looking at performance sitting somewhere around 1070-1080 levels for something like $300.

Just to chuck a guess on record so that I can come back and gloat if I get close... Anyone else fancy a guess?

The top Navi card will be 140mm2 die on a 7nm process with 52 CU's clocked at 2Ghz matched with 8 gigs of GDDR6 a launch price of $289 and performance within 5% of a 1080.
 
Panos said:
The screenshot is TOO old.
Last we heard from AMD was that Instinct cards run at that, the V20 is 20.9Tflop+ card.
Now V64 at boost (1530) is ~12.5Tflop, and Instinct Mi25 is 12.3Tflop maximum.

20.9Tflop is more than 35%. I wouldn't be surprised if a well tuned Vega 20 (downvolt, overclock), wouldn't hit 24Tflop. Already we do 15Tflop with the cards we have atm.


What AMD needs to fix is the Pixel Rate (aka core speed). A Vega 64 has 98.30 GPixel/s and 12.5Tflop while a GTX1080Ti has 139.2 GPixel/s and 11.4Tflop.
That difference is down to one is working at 2000Mhz (1080Ti) the other at 1530(tops 1720 with tuning).

If Vega was running at same speed at the Pascal cards, it would have given it a hell of a kicking to oblivion.
However that could be solved next gen, as we know the AMD 7nm is on HPP for all their product, no more LPP (low power process). That is why the Vega 64 (Power Save mode) and Mi25 just sip the power (170W 1550Mhz) while overclocked by just +10% needs +90% power.
Same applies to Ryzen CPUs. At 35W running fine, but to gain the extra 15% perf, they jump three times the power consumption.
Click to expand...

I think that's all fair points.
I don't think it's going to be difficult for them to get a mid tier card surpassing 1080 using the node shrink. A 7nm 580 at 2ghz would do it. For a top tier card, other tweaks could be added on top. Like you say Vega 64 at 2000mhz would handily beat the 1080ti, then you need another 30% more to get to the 2080ti. Most of the reason the 2080ti is so fast is just because it's massive, but Nvidia have a handle on power consumption.
I think for AMD it's mostly it's about getting those clock speeds up and power consumption down. Fix that and the gap gets drastically reduced.

With a bigger die, high clock speed and some more shaders or just higher memory bandwidth, it's not that difficult to get up to or at least near 2080ti level performance, but power can only come down so much on a node shrink.

I'm going to guess this time the high end part will be closer to Nvidias top end than people think, it might even beat it, but it will be brief, once Nvidia get on 7nm too it's going to get ugly again unless there is a new arcitecture for AMD.
 
Last edited:
Lokken86 said:
I think that's all fair points.
I don't think it's going to be difficult for them to get a mid tier card surpassing 1080 using the node shrink. A 7nm 580 at 2ghz would do it. For a top tier card, other tweaks could be added on top. Like you say Vega 64 at 2000mhz would handily beat the 1080ti, then you need another 30% more to get to the 2080ti. Most of the reason the 2080ti is so fast is just because it's massive, but Nvidia have a handle on power consumption.
I think for AMD it's mostly it's about getting those clock speeds up and power consumption down. Fix that and the gap gets drastically reduced.

With a bigger die, high clock speed and some more shaders or just higher memory bandwidth, it's not that difficult to get up to or at least near 2080ti level performance, but power can only come down so much on a node shrink.

I'm going to guess this time the high end part will be closer to Nvidias top end than people think, it might even beat it, but it will be brief, once Nvidia get on 7nm too it's going to get ugly again unless there is a new arcitecture for AMD.
Click to expand...

AMD need to match or exceed NVidia for clockspeed and efficiency on the same die size.

Just moving to 7nm with the same old inefficient designs is just a total waste of time and money as the benefits as you highlighted above will be very shortlived once NVidia also produce SKUs on the same die size.
 
Lokken86 said:
I think that's all fair points.
I don't think it's going to be difficult for them to get a mid tier card surpassing 1080 using the node shrink. A 7nm 580 at 2ghz would do it. For a top tier card, other tweaks could be added on top. Like you say Vega 64 at 2000mhz would handily beat the 1080ti, then you need another 30% more to get to the 2080ti. Most of the reason the 2080ti is so fast is just because it's massive, but Nvidia have a handle on power consumption.
I think for AMD it's mostly it's about getting those clock speeds up and power consumption down. Fix that and the gap gets drastically reduced.

With a bigger die, high clock speed and some more shaders or just higher memory bandwidth, it's not that difficult to get up to or at least near 2080ti level performance, but power can only come down so much on a node shrink.

I'm going to guess this time the high end part will be closer to Nvidias top end than people think, it might even beat it, but it will be brief, once Nvidia get on 7nm too it's going to get ugly again unless there is a new arcitecture for AMD.
Click to expand...

Have you seen the power consumption of the RTX2080ti when overclocked???

@Kaapstad the Vega is made using low power process, which is working at 1250 at 60W and 1550 at 176W. That is why the moment you try to go 1630 all cards are set to 330W from factory (doesn't need it can be done at 280W). It doesn't scale well because it is designed that way, since is a compute card to burn little power on its primary function. Computing.
 
Panos said:
Have you seen the power consumption of the RTX2080ti when overclocked???

@Kaapstad the Vega is made using low power process, which is working at 1250 at 60W and 1550 at 176W. That is why the moment you try to go 1630 all cards are set to 330W from factory (doesn't need it can be done at 280W). It doesn't scale well because it is designed that way, since is a compute card to burn little power on its primary function. Computing.
Click to expand...

I use a Computing card myself for gaming and it overclocks like a beast and is easily as good or better than Pascal even with the stock air cooler.

Vega needs to be improved on as it is a couple of generations behind what NVidia are offering at the moment in terms of efficiency and clockspeed.

A Titan V with 21 billion transistors can overclock to clockspeeds that no Vega card can get close to and uses less power to do it.

I am pretty sure that AMD are spending their resources working on efficiency as there have been no attempts by them to rush out make fix cards to compete with the 1080 Ti ot Titan Xp, far better for them to take their time and produce future cards that will compete with or beat the next generation of NVidia cards.
 
Eurasian Economic Union certification office has just listed PowerColor’s upcoming Radeon RX 590 graphics card.

PowerColor Radeon RX 590 Red Devil
Yet another proof of AMD’s yet unreleased Radeon RX 590 has been found. PowerColor is set to launch new Red Devil graphics card based on RX 590 SKU. The 8GB GDDR5 memory configuration appears to be confirmed by product code listed at EEC.

The certification is the last process before new cards reach the market. Based on what we have been seeing in the past is very likely that RX 590 will indeed launch in a few weeks.

  • AXRX 580 8GBD5-3DH/OC: RX 580 Red Devil 8GB GDDR5
  • AXRX 590 8GBD5-3DH/OC: ?
Click to expand...

https://videocardz.com/newz/powercolors-radeon-rx-590-red-devil-listing-found
 
https://wccftech.com/amds-vega-20-gpu-early-engineering-sample-benchmarked-in-final-fantasy-xv/

AMD’s upcoming Vega 20 GPU was recently spotted (courtesy of Videocardz) in Final Fantasy XV benchmark db and represents one of the first data points we have in how it will fare in gaming. It goes without saying that this is a very very early preview and will almost certainly not be indicative of the final product. This is in all likelihood an early engineering sample so clocks and drivers are expected to improve further as well.

AMD’s Vega 20 GPU benchmarked in Final Fantasy XV – trades blows with a GTX 1080
Before we begin, it’s worth pointing out that we have no information on whether this is the RX version or not. In fact, as we have reported earlier, we believe AMD will not be rolling out an RX version of their 7nm Vega 20 at all and will instead move directly to the next generation architecture. This is most likely the Instinct variant – which isn’t really designed for gaming of course. The PCI-E ID is 66AF:C1 and performs roughly equivalent to a Pascal GeForce GTX 1080.
Click to expand...

Not great news if this is actually indicative of node improvements for 7nm... although obviously massive bags of salt to be used here.
 
Beren said:
https://wccftech.com/amds-vega-20-gpu-early-engineering-sample-benchmarked-in-final-fantasy-xv/



Not great news if this is actually indicative of node improvements for 7nm... although obviously massive bags of salt to be used here.
Click to expand...

You do understand any card could have been taken, flashed to the ID and used yes? Also what we see are some bar charts and no further evidence.

An tuned Vega 64 is faster than this. Which is a 12tflop card. The vega 20 Instinct is 21tflop according to AMD own words.
 
Hence bags of salt. However, just because a compute card can compute at a vast tflop level does not mean it can game. I suspect the first instinct card will have the same 64 Vega CUs it will just have a lot of extra stuff on top as well.
 
Panos said:
You do understand any card could have been taken, flashed to the ID and used yes? Also what we see are some bar charts and no further evidence.

An tuned Vega 64 is faster than this. Which is a 12tflop card. The vega 20 Instinct is 21tflop according to AMD own words.
Click to expand...


Where has AMd said Vega 20 is 21tflop?

All I have seen is laughably inaccurate rumors and speculation, e.g. assuming that by magic the 7nm is 2.8x as dense at 14/16nm, and that every extra transistor goes in to compute, and that clock speeds will be maximized regardless of power consumption.


For starters, Vega 20 has to have a lot of the compute logic added back in, especially in regards to FP64 getting back to 1:2.
And then there is the fact that AMD would be screwed if they don't add a lot of additional logic to accelerate deep learning and mixed precision. They may not have dedicated tensor cores but they is still some things they can do with their stream processor to help.

Vega 20 will then need a bigger memory interface to access 4 HBM stacks, and the memory interface is quite a large part of the Vega 10 chip. There are new interface standards and likely improvements to HBCC etc that will eat of the transistor budget.


And then all of this will be highly variable in relation to power usage and clocks. AMD will want to stick between 250 and 300w I imagine.


You will probably find 15-16tflp is the most likely range. Even then I wouldn't be at all surprised if it was still only 13tflp.
 
Beren said:
https://wccftech.com/amds-vega-20-gpu-early-engineering-sample-benchmarked-in-final-fantasy-xv/



Not great news if this is actually indicative of node improvements for 7nm... although obviously massive bags of salt to be used here.
Click to expand...

Seems a bit late in the day for this to be appearing now.

Surely if Vega 20 has been sampling for months and due to launch anytime now, they would be benching something 'newer' than a very, very early ES chip?

What's the chances of this being Navi? Or RX590? :p
 
Last edited:
Vega 20 going by current info and "leaked" numbers, AMD would be wasting their time releasing it to the public, The performance increase over a standard Vega 64 would just be a decent overclock.
Vega 20 will probably be for severs and compute stuff only.
 
I’m very keen to support AMD. If they can release something say 10% better than the 1080ti, I’d happily pay £550 for it. As long as it OCs ok and I can pop it under water.
 
AMD chief Lisa Su confirms new high-end GPU; “we will be competitive in high-end graphics”
AMD-Ruby-feature2-672x372.jpg

https://www.dsogaming.com/news/amd-...-we-will-be-competitive-in-high-end-graphics/
 
LoadsaMoney said:
Believe it when i see it.
Click to expand...

Intel already feels it and the nail goes into their coffin next year.
its not like amd is out of it,
amd 590 king with 1080p.
vega king with 1440p
the rest 4k and so on is like less than 10% of a user base that will overpay for stuff anyhow.
The majority of users isn't buying 1500 euro cards.

amd will ryzen again with gpus to :)
 
Intel rested on their laurels and Ryzen gave them a scare. Ultimately it's a dumb decision for any company to underestimate their competitors as has been shown with Intel's scrambling response to Ryzen with a 90c 8 core cpu.
 
LoadsaMoney said:
Im not getting on the hype train, thats for sure :p
Click to expand...

Problem is the hype train is created by random sites pulling numbers and specs out of their asses for the most part. AMD generally say nothing about performance before the launch. Yet people act like it was AMD giving out the info.

Launch after launch we see this, websites looking for clicks putting up "info" that could be interpreted either way so they can ignore what they were wrong on and point to the info they got vaguely right.

Literally any rumour about new cards is almost treated like gospel these days for whatever reason. Then when the real numbers come in and if they're under the hype meter people act like they were lied to, not by the websites but the GPU manufacturer who to that point said virtually nothing. Said it a few times, rumours somehow turn into "AMD said" this or that over the course of a thread somehow. This thread will likely be no different.
 
You must log in or register to reply here.
Back
Top Bottom