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

Page 125 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

Shivansps

Diamond Member
Sep 11, 2013
3,873
1,527
136
Also why this rather large FPS gap from the 4GB version to the 8GB version? i belived this never happened on Polaris.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,448
10,117
126
Also why this rather large FPS gap from the 4GB version to the 8GB version? i belived this never happened on Polaris.
That's a good question. Is RDNA's architecture more like NVidia's traditional arches., where things like ROP counts, are tied to memory bus sizes? Could the 8GB RX 5500 XT have more ROPs or something?

Or are we finally seeing AAA games REQUIRE more than 4GB of VRAM, @ 1080P (High/Ultra settings)?

If the latter is true, than any 4GB or less VRAM card these days (very late 2019) is completely DOA in my eyes. (I've been saying that for nearly a year already.)
 

DeathReborn

Platinum Member
Oct 11, 2005
2,755
751
136
Tragic ,
If they would just port Polaris 30 (RX590) to 7nm they would have had a smaller, faster and cheaper card than this.

I dont know about you but with NAVI 14 its looks to me that RDNA is not scaling down very well.

Doesn't bode well for Samsung on mobile. The voltage still high for 7nm on Navi 14 so hopefully undervolting can o some serious wonders.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,448
10,117
126
Honestly, thought, even though the GTX 1650 (and Super) are only 4GB, and only offer mediocre performance, and the RX 5500 XT essentially matches them for price/performance, with a power efficiency advantage... the one thing to note is, the GTX 1650 Super cards are being sold out everywhere, that's not a bad thing. I predict that AMD's RX 5500 XT cards are going to sell out too, there's a lot of pent-up demand in that space, I believe.
 

DeathReborn

Platinum Member
Oct 11, 2005
2,755
751
136
First products showing up in the UK at a hideous price.

5500XT 4GB - £208
5500XT 8GB - £238-£259
570 8GB - £120-£230
580 8GB - £140-£275
590 8GB - £165-£314

1650S 4GB - £149-£212 (ASUS overpriced garbage)
1660 6GB - £180-£248 (ASUS again)
1660S 6GB - £200-£289 (does ASUS even bother trying?)
1660 Ti 6GB - £235-£363 (Gainward pips ASUS to the absurd price crown this time)

I'd expect the 4GB to drop to £160 max & the 8GB to £190 max, even then it's overspriced but to a more tolerable level. VAT is 20% on these by the way.
 

amrnuke

Golden Member
Apr 24, 2019
1,181
1,772
136
Yeah its because amd doubled prices like nvidia with kepler.They moved whole stack one tier higher.5500XT should be priced like rx560 successor.But AMD doubled prices so we have rx580 performance 2 years later for same price.ITS DOA.
That's an interesting proposal. If the 5500XT were $130 like the RX560 that would have been killer. But they wouldn't have been able to keep them in stock!

As for "doubled prices" I think it's at least reasonable to consider that perhaps AMD doubled prices on Navi10 because 7nm costs almost twice (1.75-2x) as much as 14nm. I think AMD were focusing solely on utilizing the new process for performance, not for better efficiency or smaller die size, when it comes to 5700/XT.

For 5500XT it seems they used the new process to gain better performance per mm2, while lowering price and improving power draw. (RX480 was $199 on release. 5500XT is $169).

15% lower price, ~25% less power consumption under load, 25% better performance per mm2. Certainly not anything to write home about. But also not terrible.
 
Last edited:
Reactions: guachi

mohit9206

Golden Member
Jul 2, 2013
1,381
511
136
That's an interesting proposal. If the 5500XT were $130 like the RX560 that would have been killer. But they wouldn't have been able to keep them in stock!

As for "doubled prices" I think it's at least reasonable to consider that perhaps AMD doubled prices on Navi10 because 7nm costs almost twice (1.75-2x) as much as 14nm. I think AMD were focusing solely on utilizing the new process for performance, not for better efficiency or smaller die size, when it comes to 5700/XT.

For 5500XT it seems they used the new process to gain better performance per mm2, while lowering price and improving power draw. (RX480 was $199 on release. 5500XT is $169).

15% lower price, ~25% less power consumption under load, 25% better performance per mm2. Certainly not anything to write home about. But also not terrible.

That is a wrong comparison, the comparison should be with RX470 as 5500XT is also a cut down chip.
 

amrnuke

Golden Member
Apr 24, 2019
1,181
1,772
136
Thst is a wrong comparison, the comparison should be with RX570 as 5500XT is also a cut down Navi 14.
Sure. Though RX570 isn't a cut-down Polaris. It just has shaders disabled. But whatever, let's roll with it.

RX570 retailed $169 on release.
5500XT retails for $169

- 5500XT beats 570 by 16% in 1080p gaming.
- Beats it by 41% in performance per watt by reducing power consumption at peak gaming by 27% (47W).
- Die size 232mm2 for 570, 158mm2 for 5500XT (32% smaller die size).
- At release yields for 14nm were 78% for 570/580 die size. Current 7nm yields for 5500XT die size are 88%. So yields are 12% higher. 7nm costs almost twice as much though. In the end the silicon probably costs about the same.

Overall I think the price isn't far off if you just look at the tech. But a 13% performance gain and 27% efficiency gain isn't far off from what is promised, AMD just chose to focus on less power draw instead of more performance for the same price, perhaps.
 

mohit9206

Golden Member
Jul 2, 2013
1,381
511
136
Sure. Though RX570 isn't a cut-down Polaris. It just has shaders disabled. But whatever, let's roll with it.

RX570 retailed $169 on release.
5500XT retails for $169

- 5500XT beats 570 by 16% in 1080p gaming.
- Beats it by 41% in performance per watt by reducing power consumption at peak gaming by 27% (47W).
- Die size 232mm2 for 570, 158mm2 for 5500XT (32% smaller die size).
- At release yields for 14nm were 78% for 570/580 die size. Current 7nm yields for 5500XT die size are 88%. So yields are 12% higher. 7nm costs almost twice as much though. In the end the silicon probably costs about the same.

Overall I think the price isn't far off if you just look at the tech. But a 13% performance gain and 27% efficiency gain isn't far off from what is promised, AMD just chose to focus on less power draw instead of more performance for the same price, perhaps.
16% performance increase in 2 years is pretty laughable. Should have been 40%, either that or price should have been $149.
 
Reactions: psolord

RetroZombie

Senior member
Nov 5, 2019
464
386
96
They decreased prices with their Super-lineup and forced the competition to a pre-release price cut and a two months delay.

Interesting I always saw the other way around, Nvidia (and Intel too) when AMD is about to release a new product they always do Refreshes, Rebrands, Relaunches. Nvidia for example just launched the 1650 in May 2019.

AMD normally always counter competition new products with price cuts. Now just imagine all the RDNA reviews without the RTX/GTX Super cards how good would they have looked.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
Tragic ,
If they would just port Polaris 30 (RX590) to 7nm they would have had a smaller, faster and cheaper card than this.

7nm costs 2x more vs 14nm. GDDR6 probably 2x too. For the remaining pieces of the cards the costs stay the same for both.
I'm not sure but the RX580 8GB release price was around 250$ and the RX570 199$ (real price), if today I can buy them new at about half the price great! I don't mind if Nvidia or Intel do the same.
 

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
Yep, don't expect 7 nm Nvidia consumer GPUs anytime soon. We haven't even got all Turing refresh from top to bottom.

Yeah why would NV pay for 7nm when their competior can just barley match their 16nm offerings. RDNA isn't any step forward compared to GCN. Instead of transistors for actual functional units they used a ton of them for higher clocks while loosing a lot of compute in the process.
Only matching your competitor in performance/watt and performance/$ on a node advantage is utter failure. We can see in CPU space (especially server) how a node advantage should lead to complete slaughter.
 
Reactions: insertcarehere

Olikan

Platinum Member
Sep 23, 2011
2,023
275
126
RDNA isn't any step forward compared to GCN.
LOL... did you miss latest drivers right? Navi10 is now matching Radeon 7, using way less power, half bandwidth and at the same node... oh and smaller die
 
Last edited:
Reactions: guachi

lifeblood

Senior member
Oct 17, 2001
999
88
91
AMD normally always counter competition new products with price cuts. Now just imagine all the RDNA reviews without the RTX/GTX Super cards how good would they have looked.
If it was the 5500 XT vs 1650 vanilla with no Super model it would an absolute slaughter. Much higher FPS plus modern media engine? 5500 XT would look like it was worth the price they're charging for it.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
LOL... did you miss latest drivers right? Navi10 is now matching Radeon 7, using way less power, half bandwidth and at the same node... oh and smaller die

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.
 
Reactions: beginner99

SPBHM

Diamond Member
Sep 12, 2012
5,058
410
126
I guess they might drop the price once the stocks of RX 500s are gone, because it's pretty bad right now for the 5500xt
 
Reactions: CHADBOGA

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
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.

The RX590 had a 256bit bus, this is most likely the only reason its faster than Navi 14. Give Navi 14 a proper bus, and I think it would be the card that people want.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
As i have said before, if you port Polaris 30 to 7nm it would be smaller and faster than NAVI 14.

With that I agree, specially with higher clock speeds at 7nm.
But we don't know the uncore addons, there might be enterprise features that take lots of space and add nothing to gamers.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
Yeah why would NV pay for 7nm when their competior can just barley match their 16nm offerings. RDNA isn't any step forward compared to GCN. Instead of transistors for actual functional units they used a ton of them for higher clocks while loosing a lot of compute in the process.
Only matching your competitor in performance/watt and performance/$ on a node advantage is utter failure.

We don't know that, for example the Intel 10nm cpu is worst than the 14nm in all fronts and metrics.
What Intel gained with 10nm vs 14nm besides the smaller die? Name one.
 

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
We don't know that, for example the Intel 10nm cpu is worst than the 14nm in all fronts and metrics.
What Intel gained with 10nm vs 14nm besides the smaller die? Name one.

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.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
The RX590 had a 256bit bus, this is most likely the only reason its faster than Navi 14. Give Navi 14 a proper bus, and I think it would be the card that people want.

Although RX5500XT has half the memory controllers its memory bandwidth is not that much lower than RX590

RX 590 with 256bit memory and GDDR-5 has 256GB/s of memory bandwidth
RX5500Xt with 128bit memory and GDDR-6 has 224GB/s of memory bandwidth

Just imagine Polaris 30 with GDDR-6 and ported to 7nm, it would destroy NAVI 14 easily.
 

tajoh111

Senior member
Mar 28, 2005
305
320
136
Polaris die is 232mm2, 1 month after RX570/580 release (4/2017) yields were reported 80% on their 8-core Zen chips (~9.67x22mm), which was a defect rate of 0.105, and when you calculate Polaris yields (~17.5x13.25mm) based on that defect rate you get a 14nm yield for Polaris to be 78%.

Navi die is 251mm2, presently, 7nm defect rate reported as 0.09, hence we get a yield of 80% for a die of Navi dimension (~17.95x14mm).

AMD have said that 7nm costs 1.75-2x as much as 14nm.

And they were able to scale performance on the same die up by the same amount as the new process costs (1.75-2x).

Why shouldn't they sell it for almost twice as much if it performs almost twice as well and the process costs almost twice as much, and yields are about the same?

Edit: Now, sure, we can go back and examine the RX400 series to RX500 series price increase. But to claim that the intrinsic problem is that the Polaris successor is pricey completely ignores a lot of other issues surrounding it. Such as that it is cheaper per FPS than Nvidia, and AMD are a for-profit company, and that the pricing is in-line with Polaris with respect to performance.

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.


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.
 
Reactions: Head1985
Status
Not open for further replies.
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |