• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA Working on Faster GeForce GTX 650 Ti

While I wonder if they did go with Boost what's the power section going to need?
Here's a GTX 650 Ti running a 16% overclock, that "costs" all of 1 to 6 watts more.
There's production build-up for PCB's by the AIB's, new printing of packaging, marketing… costs?
Yup. I heard board vendors hate bringing out new models...mainly from people who have never looked at the SKU line-up's available from vendors (e.g. MSI already lists six models of the card, while Asus lists seven)
For what its worth, most of the GTX 650 Ti's seem to already rely on proprietary cooling solutions to differentiate themselves in the marketplace - some, like Sparkle even use a (gasp!) non reference PCB
I think Nvidia just let loose more "higher" binned chips for less money. AIB will just stay with active Über PCB renditions and sticker the box’s as "advance clocks" and a new Sku's, while hopefully dropping the pricing. Or, is there some new GDDR memory that’s going to get above the 6000 MHz we see now on the GTX650?
Memory overclocking has already demonstrated itself to be less effective than pipeline (GPU) overclocking- both from power/heat and net gain- hence the number factory overclocked cards that sport stock clocked memory...or the fact that when AMD transformed the HD 7950 to the 7950 Boost, the memory speed was left untouched.
 
Last edited:
Memory overclocking has already demonstrated itself to be less effective than pipeline (GPU) overclocking- both from power/heat and net gain- hence the number factory overclocked cards that sport stock clocked memory...or the fact that when AMD transformed the HD 7950 to the 7950 Boost, the memory speed was left untouched.

Hardocp did a comparison of gpu and memory performance scaling on a 660 Ti and they proved equally effective. It's not surprising that now you can get bigger benefits from memory overclocking. When you think about it memory bandwidth has languished while cores keep getting faster. They're being under fed at this point.
 
Hardocp did a comparison of gpu and memory performance scaling on a 660 Ti and they proved equally effective. It's not surprising that now you can get bigger benefits from memory overclocking. When you think about it memory bandwidth has languished while cores keep getting faster. They're being under fed at this point.
I remember seeing the article - with that Galaxy card they were in love with I believe. Similar performance gains, although from what I read (and subsequently), the memory overclocked around 10% higher than the core in the benchmarks- I think they stated that the card was core clocking @ ~1.3GHz...whereas that frequency was peak/transient, and not sustained boost.

Given that Kyle and Co specialize in 2560x1600 testing + MSAA/SSAA, I'm not overly surprised that the 660 Ti's lack of stock bandwidth couldn't be exploited by comparison.
 
It's everywhere in the world except the US. Don't complain when it's your country's tax system at work causing cost push price inflation.

MSRP in US$ inherently favors the US. Mostly because for US customers MSRP is already tax inclusive, but distributors for other countries simply take the MSRP and slap additional tax and shipping on top of it, instead of on top of their actual acquisition cost.
 
that $150 turns into $170 for me with 13% tax

for the... less 'mainstream' countries, i also wonder if it's just a worse distribution chain, less stock, less customers, maybe import from other stores to resell

well even in noth america there are plenty of stores that jack up the price, or worse, pretend something is on sale (tigerdirect recently did geforce titan at $1,250 or so)
 
I can guarantee it's not only Europe, but Asia as well

Australia will put you all to shame lol :p

I already own a GTX 650 Ti GS OC model so meh, and we still have no idea how the 7790 will perform yet do we?
 
Australia will put you all to shame lol :p

I already own a GTX 650 Ti GS OC model so meh, and we still have no idea how the 7790 will perform yet do we?

yes we do! between 7770 and 7850. But that's such a wide gap. If I was to describe that gap it'll like "throwing a hotdog into a gymnasium"
 
yes we do! between 7770 and 7850. But that's such a wide gap. If I was to describe that gap it'll like "throwing a hotdog into a gymnasium"

Thats not realy helpful, thats more of a dir moment as everyone knows it be between those two cards, but we want to know how it performs against the GTX 650Ti and price, and so far no benchmarks so its all guess work at this stage
 
I can guarantee it's not only Europe, but Asia as well
That really depends on which part of Asia.
For example since Hong Kong enjoy very low tax rates, and no sales tax hardware prices are quite very good here.
This also have to do how close we are to mainland China.
 
BSN - Nvidia Desperately Rushes the GTX 650 Ti Boost to Answer AMD Radeon HD 7790

We received preliminary specifications for the GeForce GTX 650 Ti Boost which might change by launch date, but they stand at:
•Codename: GTX 655
•Retail: GeForce GTX 650 Ti Boost
•GPU: GK106
•Cores: 768
•Base Clock: 980MHz
•Boost Clock: 1030MHz
•Memory Interface: 192-bit
•Memory Clock: 1500MHz QDR (6 billion transfers per second)
•Memory Bandwidth: 144.000 MB/s, 140.63GB/s
 
Memory overclocking has already demonstrated itself to be less effective than pipeline (GPU) overclocking

Well, no argument, but if you already told AIB's the chips they've already acquired, are copasetic to "X" bin core clocks can/would Nvidia say they're now good to something more? They're certainly getting more good "Top-Shelf binned" chips, so yes they’ll just release more that have ~1070Mhz at lower prices and give some AIB's a kick back on what they own so mid range clock card should come down.

While I'd say don't discount bandwidth... Have a look at this synopsis from XbitLabs. They show that a Gigabyte with memory OC’d at 7160Mhz and a Zotac AMP at 6200Mhz. Comparing very close core clocks 1078Mhz Gigabyte and 1072Mhz/5400mhz of the EVGA the Gigabyte is 8% faster. While Zotac vs. (stock) Gigabyte are both 1033Mhz core clock; the 5400Mhz memory on the Gigabyte against the Zotac at 6200Mhz memory, there's again almost another 8% improvement. Also, look at the 4xAA/AF16x between the EVGA and OC Gigabyte there's 10% differance. I think that's what Nvidia is hanging there hopes on.
http://www.xbitlabs.com/picture/?src=/images/graphics/geforce-gtx-650-ti-roundup/zfulltable.png
 
Last edited:
Codename: GTX 655
•Retail: GeForce GTX 650 Ti Boost
•GPU: GK106
•Cores: 768
•Base Clock: 980MHz
•Boost Clock: 1030MHz
•Memory Interface: 192-bit
•Memory Clock: 1500MHz QDR (6 billion transfers per second)
•Memory Bandwidth: 144.000 MB/s, 140.63GB/s

Ok now we wait to se if the run with the longer GTX660 PCB, I think they will. Appear they aren't seeing as many "Top-Shelf 768" as I would have thought. How cost effective can this be?
 
Ok now we wait to se if the run with the longer GTX660 PCB, I think they will. Appear they aren't seeing as many "Top-Shelf 768" as I would have thought. How cost effective can this be?
Boost clock is largely subjective with Kepler cards- it think that has already been amply demonstrated, so I'd guess that any stated boost number is going to be dependent upon GPU load efficiency- or, given Titan's Boost 2.0, possibly deterministic regarding thermals since the 650 Ti isn't TDP limited.
If the latter is the case, then it would stand to reason that dual fan (and longer PCB) designs would dominate from vendors...but then the longer PCB is exactly how it stands now. Not maxing out the clocks also allows vendors to market overclocked models of the "new card".

BTW: The present 650 Ti size approximates that of the HD 7750/7770/7850, and I'm pretty certain that the Nvidia based cards from the vendors don't carry a higher BoM than Radeons...so why point out cost effectiveness ? or is the lack of cost effectiveness to the end user? Don't you need pricing to complete that equation ?


/Someone should tell BSN that I was joking when I suggested using the name GTX 650 Ti Boost (in post #23)
 
Last edited:
Now why do I want a GPU that'll be slower than the PS4, let alone run poorly optimzed next-gen ports.
 
Now why do I want a GPU that'll be slower than the PS4, let alone run poorly optimzed next-gen ports.

You dont. This product is meant for someone else.
 
Now why do I want a GPU that'll be slower than the PS4, let alone run poorly optimzed next-gen ports.

By which metric is it slower, and by how much? No sarcasm here, i just don't know and i'm not going to learn how to convert AMD numbers to Nvidia numbers to performance, i got my plate full of things to learn.
 
I doubt just clocks are going to be enough to compete.

EDIT: missed the addition 192 bit bus.
 
Australia will put you all to shame lol :p

I already own a GTX 650 Ti GS OC model so meh, and we still have no idea how the 7790 will perform yet do we?

In Malaysia, we pay $500 for a GTX 670 from ASUS.

But for some reason, prices for Gigabyte and ASUS cards are inflated and yet cards from Sapphire are seem to be priced better.
 
In Malaysia, we pay $500 for a GTX 670 from ASUS.

But for some reason, prices for Gigabyte and ASUS cards are inflated and yet cards from Sapphire are seem to be priced better.

Same here in Indonesia, MSI cards are slightly priced better than Asus and Gigabyte. The best deals here are come from Sapphire and Zotac, which came from same company (PC Partner).
 
Last edited:
Back
Top