[Videocardz] 1650 details

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

ozzy702

Golden Member
Nov 1, 2011
1,151
530
136
It inherited a lot from previous cores (~the entire BPU is pretty much from BD), and used a lot of things Intel did.

You never go for a ground-up redesign in GPUs.
They never work.

Everything GPU since the dawn of unified shaders to Turing is incremental.

That would be basically mobile 1650 (full TU117) clocked higher.

NVIDIA has made much greater leaps with it's incremental steps than AMD has with GCN. Nothing points to GCN making a magical leap in efficiency without a major overhaul (fundamentally a new architecture) on a level we haven't seen thus far from AMD, who knows, maybe it will happen.

You keep saying this, but where has it been confirmed that there will be no 1650TI with GDDR6? I know you like to pretend you're the god of GPU information, so please, enlighten us. From where I'm sitting that would leave a gaping hole in NVIDIA's lineup and I just don't see them making that move.
 

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
NVIDIA has made much greater leaps with it's incremental steps than AMD has with GCN
Their biggest step was probably Maxwell, consisted of, like, making a good SM for once.
Worked wonders.
Nothing points to GCN making a magical leap in efficiency without a major overhaul
It can be magical leaps of whatever.
Perfomance density?
Whatever you wish for.
fundamentally a new architecture
It's gonna be painfully incremental like all things GPU are since the dawn of unified shader.
but where has it been confirmed that there will be no 1650TI with GDDR6?
Right when we learned than full TU117 is a mobile SKU.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
On the other hand if it's just another incremental revision like we've seen so often since the 7000 series with GCN, at best they will catch NVIDIA's 12nm efficiency and when the 3000 series is dropped it will be the same story all over again with NVIDIA taking a massive lead and AMD forced to compete as the "budget" gpu with rock bottom prices.

You are talking like AMD will not release any other GPU after NAVI, yes NVDIAs 7nm 3000 series will be better than NAVI but it will come more than 6 months after NAVI and then when NV 3000 arrives, AMD will release Next-Gen at 7nm+.
I believe from NAVI forward we could see one company leapfrog the other with a new series every 6-12 months.
 

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
GCN was bad for gaming since day1.AMD lost their perf/mm2 and perf/watt lead.And since then Nvidia is better in that.
Biggest change was probably polaris-390(250w) performance at 150w TDP(rx480).But how much was that due new node shrink?(28 to 14nm)
Vega was just crap with no ipc gains.

Nvidia did much better.Biggest change was probably maxwell-780TI(250w) performance at 150w(gtx970) and at same node.Also GTX970- 1664SP and 224bit card matching 780TI- 2880/384bit card on same node was amazing.
Turing is also good and gained again another 15-20% "IPC" over pascal.
 
Reactions: ozzy702

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
Yes, a definite yes at that.
I still wonder how did they manage to nail their SIMD design on day1.
How it is good when amd lost all they lead they have with terascale?
They had the lead because nVidia was doing a terribad job at making GPUs back then.
VLIW itself was ass.
You call GOOD when you release worse product than competition?
Tahiti, both the og and the 1GHz refresh were better than 680.
I won't even talk of 7850/7870 duo, or 7870XT.
 
Last edited:

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
Tahiti, both the og and the 1GHz refresh were better than 680.
Better in what?
7970-352mm2 and 384bit 250w and 300w for Ghz edition
680-294mm2 and 256bit 195w
In games 680 was 10% faster at release than 7970 and on par with ghz edition.
 

ozzy702

Golden Member
Nov 1, 2011
1,151
530
136
GCN was bad for gaming since day1.AMD lost their perf/mm2 and perf/watt lead.And since then Nvidia is better in that.
Biggest change was probably polaris-390(250w) performance at 150w TDP(rx480).But how much was that due new node shrink?(28 to 14nm)
Vega was just crap with no ipc gains.

Nvidia did much better.Biggest change was probably maxwell-780TI(250w) performance at 150w(gtx970) and at same node.Also GTX970- 1664SP and 224bit card matching 780TI- 2880/384bit card on same node was amazing.
Turing is also good and gained again another 15-20% "IPC" over pascal.

Exactly my point. NVIDIA has fairly consistently executed sizable architecture upgrades which have led to gains in performance/watt. AMD has lagged far far behind with the progress of GCN and forgive me if I have a hard time believing we'll see GCN suddenly make huge IPC gains without a massive rework that's essentially a new architecture altogether. The efficiency gains AMD has had since the 7000 series have almost exclusively come from node shrinks in contrast with what we've seen from NVIDIA.

I really do hope that massive architecture changes are made and that AMD can execute properly (because I'm tired of NVIDIA competing against their prior generation and screwing the consumer). The 7970 was the last AMD GPU I rocked in my main box and it was really a fantastic card that held up well over time (and mined BTC like a champ), I'd love to be able to recommend NAVI like I'm currently pushing people towards dirt cheap 570s and Ryzen CPUs without any reservations. AMD has certainly proven with Ryzen that it's possible to make enormous gains, but until I see proof I'll remain skeptical.
 
Last edited:

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
NVIDIA has fairly consistently executed sizable architecture upgrades which have led to gains in performance/watt
Well Maxwell was that, that's about it.
Good SIMDs just work.
I have a hard time believing we'll see GCN suddenly make huge IPC gains
One should never apply "IPC" to GPUs, for these are SIMD machines.
that's essentially a new architecture altogether.
It's still gonna be GCN-compliant.
 

ozzy702

Golden Member
Nov 1, 2011
1,151
530
136
You are talking like AMD will not release any other GPU after NAVI, yes NVDIAs 7nm 3000 series will be better than NAVI but it will come more than 6 months after NAVI and then when NV 3000 arrives, AMD will release Next-Gen at 7nm+.
I believe from NAVI forward we could see one company leapfrog the other with a new series every 6-12 months.

Leapfrog? I have a hard time believing that. I think we'll see NAVI come close for a few months and then be left in the dust, then the iteration maybe come close again, then be left in the dust, and that's best case scenario until 2021-2022 when some of that sweet CPU cash flow into the GPU division materializes in massive gains. This is also with the assumption that NVIDIA doesn't step up their game and redeem themselves after the 2000 series debacle. Time will tell.
 

tajoh111

Senior member
Mar 28, 2005
306
326
136
About the bolded: LOL! Two of the biggest names in YouTube hardware reviewing, said that the card is "pointless".

And if you have info on Navi performance vis-a-via NVidia's recent card releases, and clocks, and power consumption numbers, let's have the info! Don't keep it to yourself.

Edit: And it's kind of funny that you should say that (only way that they can match Nvidia in performance), because the GTX 1650, even when fully overclocked, can't even match a cheaper AMD RX 570 card.

I would say during any other launch, the 1650 would be a so so launch. Not the bad launch it is at the moment.

The problem is Polaris, particularly the rx 570 variant is largely underpriced at the moment. This is from a combination of over supply in the used market, the price drops due to competition from GTX 1660 and a general oversupply of new Polaris in general. What this has caused is the price of the RX 570 and to a lesser extent 580s to undergo firesale pricing where AMD isn't making money, hence the underpricing. Polaris was a lower margin videocard for AMD in the first place, so a 30-40% decrease in price(even considering the amount of time that has passed), is probably break even to a loss for AMD.

You are getting alot of hardware in terms of components, die size and complexity of the PCB. Far more than any point in the past really. Taking into account inflation, a $130 rx 570, is equal to about 120 dollar r7 270, a 116 dollar 6850, $112 4850s. These predecessors never reached pricing this low(definitely not sustained)because they didn't stay on the market for the same length of time. They were discontinued before the price got too low. Polaris on the other hand has been on the market forever and as a result we have seen the price fall over time(without mining influence). AMD has repaired this largely by rebranding, discontinuing old stock and repricing to levels where it makes profits largely better than the launch pricing(better yields).

With so much Polaris chips and Navi coming somewhat soon, AMD cannot rebrand this chip again and the price has been allowed to fall to the point the rx 570 has hit floor pricing which is basically the cost of the videocard.

This low pricing is painful for AMD because they are likely selling these cards closer to cost(perhaps even a loss) at this point and they were potentially losing big money with the game bundle(atleast with the rx570).

Normally new card do not have to compete with firesale/floor pricing cards because the remaining stock are sufficiently managed on both sides. Nvidia will have cleaned out most of it's old stock and AMD will have a rebrand prepared to phase out the old SKU and raise the pricing of cards again.

This time, its different because there is huge stock of old cards available because neither company managed their inventory well because of the mining craze. AMD tried to maintain pricing of their cards and prevent this price freefall from happening by releasing their gaming bundle as a substitute which mostly worked until RTX 2060 and lower were released.

Prices dropped to historical lows for Polaris and for the most part are selling at or below cost.

When these cards were selling for around 130 new(for 8gb versions), it is difficult to see AMD making money anymore. The significant cost of 8gb of memory cost is some where around 50 dollars. Add board partner profit, retailer, distributer cost and your taking another 25%. That leaves you around 50 dollars to build the rest of the card like the PCB, power circuitry, cooler etc(the cooler alone is 15-20$), the packaging, logistics, and there is simply no profit left for AMD.

Overall the point is all cards are going to have a very bad price to performance vs the rx570 because it is being sold at or below cost and was already a pretty good value in the first place. Normally cards are not compared to the closeout pricing of cards but that reality exists for the 1650 series due to the sizable quantity of rx 570's available.

To be fair, that's probably true. If it cost the SAME as an RX 570, I think that it would be a lot more viable in the market, and not DOA like it kind of is now. I could see NVidia moving the price down to $110-120, to match the price of the cheaper RX 570 variants, while at the same time, slotting in a "full fat" GTX 1650 ti, with GDDR6, as faster than the current batch of RX 570 cards, for an equivalently higher price, matching RX 570's price/performance, and beating it on power-consumption. (But if the 1650 non-ti barely makes the cut for the lack of 6-pin PCI-E power, I don't see too many of these hypothetical GTX 1650 ti cards lacking PCI-E power either.)


Maybe I'm just sensitive to speculation, wholly painting AMD in a poor light. After all, they came out with Zen, and that does wonders for power-efficiency, they even have an 8C/16T CPU that's 65W, and going to get even better with 7nm Zen. And Navi is going to be on 7nm too.
 
Last edited:

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
Leapfrog? I have a hard time believing that. I think we'll see NAVI come close for a few months and then be left in the dust, then the iteration maybe come close again, then be left in the dust, and that's best case scenario until 2021-2022 when some of that sweet CPU cash flow into the GPU division materializes in massive gains. This is also with the assumption that NVIDIA doesn't step up their game and redeem themselves after the 2000 series debacle. Time will tell.

AMD already has almost the same perf/watt as NVIDIAs Turing cards with Polaris 30 and Vega 20. The reason you dont see this is because AMD uses those chips to compete at higher segments than their competitors.

For example,

Polaris 30, RX 590 (232mm2 at 12nm GloFo) is competing against TU116, GTX 1660 (284mm2 at 12nm TSMC) instead of competing against TU117, GTX 1650 (200mm2 at 12nm TSMC).
So with a smaller chip they are trying to compete against a bigger chip, the result is to increase power consumption to the roof. If they would compete against GTX 1650, then Polaris 30 would have very close perf/watt but they have to sell those RX 570/580 at that segment and put Polaris 30 at a higher segment.

Also, If AMD would release a 175W TDP VEGA 20 card, they would have almost the same performance as RTX 2070. But, instead they increased performance to catch RTX 2080 and bye bye efficiency.

I dont know what NAVI performance targets are, but even Polaris at 7nm would completely destroy Turing in perf/watt (at iso Perf).
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
If NVIDIA can sell a 200mm2 at 12nm (GTX1650 4GB) at $149 MSRP and make a profit (New product, they spend money for R&D), then AMD can make a profit with selling Polaris 10 RX 570 4GB (232mm2 at 14nm GloFo) at $130 today after 3 years in production.
 
Reactions: Ranulf

ozzy702

Golden Member
Nov 1, 2011
1,151
530
136
AMD already has almost the same perf/watt as NVIDIAs Turing cards with Polaris 30 and Vega 20. The reason you dont see this is because AMD uses those chips to compete at higher segments than their competitors.

For example,

Polaris 30, RX 590 (232mm2 at 12nm GloFo) is competing against TU116, GTX 1660 (284mm2 at 12nm TSMC) instead of competing against TU117, GTX 1650 (200mm2 at 12nm TSMC).
So with a smaller chip they are trying to compete against a bigger chip, the result is to increase power consumption to the roof. If they would compete against GTX 1650, then Polaris 30 would have very close perf/watt but they have to sell those RX 570/580 at that segment and put Polaris 30 at a higher segment.

Also, If AMD would release a 175W TDP VEGA 20 card, they would have almost the same performance as RTX 2070. But, instead they increased performance to catch RTX 2080 and bye bye efficiency.

I dont know what NAVI performance targets are, but even Polaris at 7nm would completely destroy Turing in perf/watt (at iso Perf).


You're hyper focused on the size of these chips, which in the end means nothing. NVIDIA's larger chips also boost higher, with significantly less power consumption so... everything you just wrote is completely wrong and an apologist excuse for AMD's lack of efficiency/competitiveness. NVIDIA is WAY ahead of AMD on performance/watt and that's what matters. NAVI will be lucky to come anywhere near Turing in performance/watt even with a vast 7nm advantage.
 

tajoh111

Senior member
Mar 28, 2005
306
326
136
If NVIDIA can sell a 200mm2 at 12nm (GTX1650 4GB) at $149 MSRP and make a profit (New product, they spend money for R&D), then AMD can make a profit with selling Polaris 10 RX 570 4GB (232mm2 at 14nm GloFo) at $130 today after 3 years in production.

It simply not just about the die size(on top of being larger), its also about the complexity of the PCB, the required powered circuitry, stronger cooler, the amount of memory(this is a biggy). 4gb vs 8gb of memory is 25 dollars in cost for the bill of material for the card.

A 128bit pcb vs 256bit is considerably more complex and when you add the additional voltage and power circuitry, the 20 dollar cheaper price, A new game bundle which is probably costing AMD 15-20 dollars, and 25 dollars less in memory, there is simply no profit left.

The GTX 1650 has that profit built in because it lacks the above even with the R and D that needs to be accounted for because the bill of material is so much less. Zero games + 128bit bus/pcb, 4gb of memory + $20 dollar additional retail price = profitable card.

8gb of memory + 256bit bus/pcb + 2 games evaporates the margin because your looking at something that 40-50 dollars more in terms of BOM while selling for 20 dollars less.

Yes Nvidia still has to cover that R and D expense which is why it needs a better margin at this point, but that 60-70 dollar difference in selling cost and bill of material more than covers it.
 

ozzy702

Golden Member
Nov 1, 2011
1,151
530
136
There's an area tax for that all, you know.

Yes, which is more than compensated for by vastly superior performance and efficiency which translates into more profit when all is said and done. AMD needs a quantum leap to catch up with NVIDIA and that will likely take a major rework of the architecture AND improved software, which all costs a significant amount of money. It's why I'm not hopeful with NAVI but down the road see AMD coming back with a vengeance with AMD being able to support sizable R&D budgets due to their fantastic execution on the CPU side. The money hasn't been there to develop NAVI, and this kind of undertaking takes years so I'm pretty skeptical that AMD will pull it off.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
It simply not just about the die size(on top of being larger), its also about the complexity of the PCB, the required powered circuitry, stronger cooler, the amount of memory(this is a biggy). 4gb vs 8gb of memory is 25 dollars in cost for the bill of material for the card.

Only the 4GB RX 570 is currently at $130 not the 8GB.

A 128bit pcb vs 256bit is considerably more complex and when you add the additional voltage and power circuitry, the 20 dollar cheaper price, A new game bundle which is probably costing AMD 15-20 dollars, and 25 dollars less in memory, there is simply no profit left.

There is no Game bundle for the RX 570 4GB at $130, also I higly doubt they give more than $5 for each game.

The GTX 1650 has that profit built in because it lacks the above even with the R and D that needs to be accounted for because the bill of material is so much less. Zero games + 128bit bus/pcb, 4gb of memory + $20 dollar additional retail price = profitable card.


8gb of memory + 256bit bus/pcb + 2 games evaporates the margin because your looking at something that 40-50 dollars more in terms of BOM while selling for 20 dollars less.

Yes Nvidia still has to cover that R and D expense which is why it needs a better margin at this point, but that 60-70 dollar difference in selling cost and bill of material more than covers it.

See above.

Also, 12nm TSMC wafers should be more expensive than 3 years old 14nm GloFo wafers.
Yes the RX 570 PCB is a little bit more complex vs GTX 1650 but that will not make the RX 570 sold at near loss. They are not making a fat profit either thats for sure but those two cards are entry level now. That means low margins high volume products.
 
Last edited:

Guru

Senior member
May 5, 2017
830
361
106
Exactly my point. NVIDIA has fairly consistently executed sizable architecture upgrades which have led to gains in performance/watt. AMD has lagged far far behind with the progress of GCN and forgive me if I have a hard time believing we'll see GCN suddenly make huge IPC gains without a massive rework that's essentially a new architecture altogether. The efficiency gains AMD has had since the 7000 series have almost exclusively come from node shrinks in contrast with what we've seen from NVIDIA.

I really do hope that massive architecture changes are made and that AMD can execute properly (because I'm tired of NVIDIA competing against their prior generation and screwing the consumer). The 7970 was the last AMD GPU I rocked in my main box and it was really a fantastic card that held up well over time (and mined BTC like a champ), I'd love to be able to recommend NAVI like I'm currently pushing people towards dirt cheap 570s and Ryzen CPUs without any reservations. AMD has certainly proven with Ryzen that it's possible to make enormous gains, but until I see proof I'll remain skeptical.
WTF are you talking about? Nvidia's current Turing architecture is literally the same one from their GTX 200 series, which was their foray into unified shaders. Ever since then its been iterations of the same architecture. Turing it literally GTX 200 iterated over 7 generations.

Whether AMD uses GCN or not, whatever they call it, doesn't matter. It's all about finding the most efficiency that you can while keeping the product profitable. We'll have to wait and see what Navi brings to the table, but I think it's a fully fledged series that will compete with the RTX 2080ti at a certain degree. Maybe they'll fall a bit short off of it, but they are probably going to sell it much cheaper for it.

I think AMD's performance leap with Navi over Polaris/Vega is going to be bigger than Nvidia's leap from Pascal to Turing. Most of Nvidia's gains were achieved by increasing the die size and using GDDR6.
 
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/    |