[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[Olikan]

Yes against VEGA its fine because VEGA Gaming ipc was atrocious.
NAVI is not that much better if better at all against Polaris though and you can clearly see this by comparing Polaris 30 (5.7B transistors) against NAVI 14 (6.4B transistors). Polaris 30 (RX590) with less transistors and a process disadvantage its faster.
As i have said before, if you port Polaris 30 to 7nm it would be smaller and faster than NAVI 14.

Yes, you can claim as a product polaris 30 is better for games... not as a whole

While navi have 0.7B transistor more, polaris lacks VP9, PCIE4.0, GDDR6, hardware virtualization, and many others features... Oh, and 5500xt also have 2 Cu disabled

 

[amrnuke]

16% performance increase in 2 years is pretty laughable. Should have been 40%, either that or price should have been $149.

Well with processes advancing you get choices:

- more performance, same power draw
- less power draw, same performance
- a combination of the two

It appears AMD chose a combination, 16% performance increase, 27% power draw decrease.

I do wonder, with a top-tier cooling solution, if one could pump more power in (since they appear to have solid VRMs on the 5500XT) to transfer some of the efficiency back into performance gains.

I do want to look at this more closely and do a per-square-millimeter performance comparison between RX570-590, 5500XT up to 5700XT, and perhaps even Vega, and chart it out to see what exactly the gains are from a power draw and performance standpoint.

Also curious: Anandtech's article states: "RX 5500 XT and Navi 14 are not meant to be upgrades for the RX 500 series. That is a job for the RX 5700 series. Instead, the RX 5500 XT is closer to an upgrade for the Radeon RX 560; Navi 14 is by and large the 7nm successor to Polaris 11." I find that interesting. They don't even bench the 5500XT against the RX560 despite stating that in the intro.

 

[fleshconsumed]

Yes, you can claim as a product polaris 30 is better for games... not as a whole

While navi have 0.7B transistor more, polaris lacks VP9, PCIE4.0, GDDR6, hardware virtualization, and many others features... Oh, and 5500xt also have 2 Cu disabled

LOL who cares about PCIE4.0 or GDDR6 or disabled CU units? Those are the design decisions that could allow for faster card, except the final product isn't faster than the old polaris. VP9 hardware support sure, that's a plus over polaris, the rest is meaningless.

 

[zinfamous]

so, an XT with 2 disabled CUs.

Does this mean we will finally see another XTX...TXTXXXT version???

 

[RetroZombie]

And? We are talking about TSMC 7nm here which has a proven track record in mobile but more importantly also in high performance x86 cpus. It's for sure not the process that's the issue here.

There never was an high performance x86 cpu made by tsmc I think (amd cats and via designs doesn't count), so it's not easy to compare, and even graphics there wasn't any radeon gcn done at tsmc 16nm, so no previous comparison can be made.
And if I use the 3 years old RX480(polaris) vs Nvidia new gtx 16xx line it also doesn't look good to Nvidia.

 

[linkgoron]

Very mediocre launch from AMD. However, it's not very surprising given the early 5500 benchmarks. It is at best equal, or gets beat, by the 1650S in all fronts. Equal perf, worse perf/watt and worse price/perf.

The 5700/5700XT easily beat the 2060/2060S and even the 2070, in the most important metrics - perf and price/perf. They actually improved things across the board (while still being quite expensive, historically). The 5500 is very late, and vs. the 1650 the 5500 would've been killer. However, Nvidia has already launched their "super" card (what used to be the regular card) more than a month ago, after the terrible 1650, and the 5500 is just not enough.

 

[amrnuke]

so, an XT with 2 disabled CUs.

Does this mean we will finally see another XTX...TXTXXXT version???

XFX 5500XT XTXTXTXXXT CHONKBOI
Featuring a cardboard heatsink and a battery-operated propeller hat for all the cool you'd ever need*
* do not actually plug in, this card is a literal fire hazard

 

[Shivansps]

so, an XT with 2 disabled CUs.

Does this mean we will finally see another XTX...TXTXXXT version???

Knowing AMD they probaly going to call it "RX5500 XT SUPER"

 

[Olikan]

LOL who cares about PCIE4.0 or GDDR6 or disabled CU units? Those are the design decisions that could allow for faster card, except the final product isn't faster than the old polaris. VP9 hardware support sure, that's a plus over polaris, the rest is meaningless.

The discussion started by this quote: "RDNA isn't any step forward compared to GCN."

 

[SteveGrabowski]

For AMD the last saving grace in the midrange market is RX 5600 series right now. And the last resort for PC gamers, that want a GPU that is not Nvidia's...

Meh it'll probably be 1660 Super performance for 1660 Super price three months later. And then Nvidia will release a 1660 Super Black to just edge it out a couple of weeks early.

 

[Stuka87]

It doesn't cost 1.75 to 2x as much compared to the early cost of 14nm finfet. 14mm finfet wafers were in the 8-10k range initially, while 7nm wafers cost somewhere around 12k. It costs 2x as much when compared to current wafer pricing for 14nm finfet but not the initial pricing of 14 finfet.

FinFET Rollout Slower Than Expected

FinFET Rollout Slower Than Expected It costs nearly three times more to design a finFET-based chip than a 28nm planar chip, and it takes more than twice as long to get working silicon.

semiengineering.com

This quoted wafer cost is the price for TSMC with the selling price for buyers being 2x.

Considering both companies could afford to give 200-232mm2 chips on 14nm finfet when it was still new, AMD can't use that excuse particularly because AMD writes off it's IP so slowly because of the long product lifecycles. Prices for wafers fall overtime which is why Nvidia's chips although much larger with Turing, can create really high margins similar to their Pascal.

The pricing of a consumer product is not based on the price to manufacture alone. R&D has to be considered, as well as marketing, etc. And it has been said by numerous sources that the initial R&D for 7nm is significantly more than it was for 14nm.

 

[CHADBOGA]

Such a disappointing product.

Now I am left with choosing between a 1660 Super and a 5700XT.

Nothing else makes any sense to me.

 

[GodisanAtheist]

While there is definitely a gaping hole between the 5700 series and the 5500, TPU has the 5700XT ~86% faster than the 5500xt while having 81% more CUs, 2x ROPs and 2x the bus width/Bandwidth, so the scaling here is about as good as one could expect.

The price could definitely be better, but it's weird that people were expecting 1660ti performance levels, even if the card shipped with it's full 24 CU compliment. That would have required regressive scaling for the 5700XT and spelled certain doom for a theoretical 5900xt.

 

[mohit9206]

5600 and 5600XT are the 2 models rumored to be announced next month, if i have to guess 5600 will be $249 with 6gb memory and 5600XT $279 with 8gb memory. 6gb in 2020 for $249 seems very regressive when AMD was already offering 8gb for $240 in 2016 with RX480. 5600XT will be the card to keep an eye on ,hopefully this time around it beats 1660Ti by a decent margin.

 

[eek2121]

5600 and 5600XT are the 2 models rumored to be announced next month, if i have to guess 5600 will be $249 with 6gb memory and 5600XT $279 with 8gb memory. 6gb in 2020 for $249 seems very regressive when AMD was already offering 8gb for $240 in 2016 with RX480. 5600XT will be the card to keep an eye on ,hopefully this time around it beats 1660Ti by a decent margin.

I haven't heard anything about the 5600 series existing, let alone being announced.

What we DO know is that AMD has a big Navi variant. This big Navi has 3 different skus and supports hardware raytracing, 4k60 gaming, 8k support, 4k120 for some titles, a new 3D sound codec, and more. How do we know this? Microsoft spilled the beans a few hours ago.

Given all the info that has been leaked out so far, I am willing to bet 2020 will be the year of AMD.

 

[mohit9206]

Actually 2020 will probably be the year of Nvidia just like 2019 but you're free to believe that AMD will make a comeback next year.

 

[ItsAlive]

I got tired of waiting so I just got a used 1660 ti for $160......I would seriously hope that 5500 cards are on the heels of 1660 ti at least.

 

[tajoh111]

Nvidia destroying PC gaming because they keep increasing prices.
GTX970-330usd
GTX1070-450usd
RTX2070-600usd

Don't forget AMD's 7870's role in this which allowed mid range dies to come in at near flagship prices.

The 7870 was a mere 212mm2 die on cheap planar 28nm wafers that launched at $350 which is equivalent to almost 400 dollars today when you account for inflation. Considering how cheap 28nm wafers are compared any version of finfet, it was even greedier than Turing pricing. AMD high initial 7xxx series pricing could be considered the catalyst for all these price increases.

If AMD had priced the 7870 at 199 and the 7970 at 350, the gtx 680 would have likely been priced at 299 and the gtx 670 at 250. But because the 7970 was priced at $550 which is over 600 dollars for a moderately sized 354mm2 and $350 for a 212mm2 mainstream die, the price of Nvidia cards rocketed up. Considering AMD was still considered the value brand at the time and was Nvidia's only competition, it would have been foolish for Nvidia not to raise their prices from a business standpoint considering kepler was initially superior to AMD's offerings.

 

[Dribble]

Given all the info that has been leaked out so far, I am willing to bet 2020 will be the year of AMD.

The info points very much to next year being the year of Nvidia - they move onto 7nm so AMD looses it's node advantage which being as AMD still can't match Nvidia for efficiency with the node advantage is pretty bleak. In addition it'll be all about ray tracing and Nvidia will have second gen ray tracing cards out, where as AMD will have just released their first cards so are likely to get destroyed in ray tracing performance.

I still get the feeling AMD aren't massively interested in discrete gpu's - they'll take some console and apple wins, keep some presence in the descrete market and be happy with that. In the mean time all the money and focus goes on the cpu market, which tbh is understandable - it's the larger market and they are much better placed.

 

[linkgoron]

Given all the info that has been leaked out so far, I am willing to bet 2020 will be the year of AMD.

Unless RDNA2 is a huge leap forward (doubtful IMO) and appears quickly, Nvidia's 7nm should easily beat AMD. AMD are just barely competitive in perf/watt with a node advantage vs. the 12nm cards as it is.

 

[Mopetar]

Actually 2020 will probably be the year of Nvidia just like 2019 but you're free to believe that AMD will make a comeback next year.

Come on now, we all know that next year is going to be AMD's comeback year. Just like it was last year, and the year before that, and the one before that too.

Kidding aside, we can complain about NVidia continually raising prices, but I don't think anyone can accuse them of sitting on the same architecture for half a decade and refusing to try to push things forward.

 

[exquisitechar]

The info points very much to next year being the year of Nvidia - they move onto 7nm so AMD looses it's node advantage which being as AMD still can't match Nvidia for efficiency with the node advantage is pretty bleak. In addition it'll be all about ray tracing and Nvidia will have second gen ray tracing cards out, where as AMD will have just released their first cards so are likely to get destroyed in ray tracing performance.

That is, if Nvidia moves to 7nm GeForce next year...lol.

I still get the feeling AMD aren't massively interested in discrete gpu's - they'll take some console and apple wins, keep some presence in the descrete market and be happy with that. In the mean time all the money and focus goes on the cpu market, which tbh is understandable - it's the larger market and they are much better placed.

They are interested, you will see that in their execution in the coming years.

 

[Head1985]

Don't forget AMD's 7870's role in this which allowed mid range dies to come in at near flagship prices.

The 7870 was a mere 212mm2 die on cheap planar 28nm wafers that launched at $350 which is equivalent to almost 400 dollars today when you account for inflation. Considering how cheap 28nm wafers are compared any version of finfet, it was even greedier than Turing pricing. AMD high initial 7xxx series pricing could be considered the catalyst for all these price increases.

If AMD had priced the 7870 at 199 and the 7970 at 350, the gtx 680 would have likely been priced at 299 and the gtx 670 at 250. But because the 7970 was priced at $550 which is over 600 dollars for a moderately sized 354mm2 and $350 for a 212mm2 mainstream die, the price of Nvidia cards rocketed up. Considering AMD was still considered the value brand at the time and was Nvidia's only competition, it would have been foolish for Nvidia not to raise their prices from a business standpoint considering kepler was initially superior to AMD's offerings.

Yep they both have guilt.

 

[tajoh111]

The pricing of a consumer product is not based on the price to manufacture alone. R&D has to be considered, as well as marketing, etc. And it has been said by numerous sources that the initial R&D for 7nm is significantly more than it was for 14nm.

AMD mechanism for fighting this has been moving most of it's R and D for graphics to China and extending the lifetime of their products to 3 to 4 years meaning the amortization rate of new IP is not written off not nearly as quick. As a result, they don't have as much of an excuse to raise prices due to R and D compared to Nvidia who still use primarily and North American based workforce for R and D and make a new lineup of cards every 24 months or less typically. An engineer at AMD Shanghai makes 1/4 the wage compared to their North American staff.

While Nvidia has kept the same cadence or product release on new nodes, AMD has slowed down drastically and cut down the number of cards they produce so if people do not allow the R and D excuse to be used on Nvidia cards, they have even less reason for this excuse to be used with AMD.

 

[beginner99]

There never was an high performance x86 cpu made by tsmc I think (amd cats and via designs doesn't count), so it's not easy to compare, and even graphics there wasn't any radeon gcn done at tsmc 16nm, so no previous comparison can be made.

lol. you kidding right? Where do you think all the Zen 2 chiplets are coming from on 7nm? Hint: it's not GF.