Can someone explain Maxwell 750 / 750ti pricing to me?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Mondozei

Golden Member
Jul 7, 2013
1,043
41
86
You need them? Haven't you been paying any attention to the 750 series launch? If you had, and I'm sure as an enthusiast you have, then you would not need any links. Why you require them is beyond me, but as you can see, Tviceman delivered. He is the man.

OP was incorrect about r7-260, but r7-265 is a lot better on a price/perf ratio and it costs the same. It's out of stock at Newegg, but too bad, because it isn't where most people live. (Newsflash: the world is larger than 'Murica).

It is a far less efficient card, of course, which is to be expected when compared to a much newer architecture, but the point stands.
On a price/perf ratio, the card is too expensive.

Some people will not care about that. But many, perhaps most, do.
 

wand3r3r

Diamond Member
May 16, 2008
3,180
0
0
You need them? Haven't you been paying any attention to the 750 series launch? If you had, and I'm sure as an enthusiast you have, then you would not need any links. Why you require them is beyond me, but as you can see, Tviceman delivered. He is the man.

No I actually haven't been following it beyond the mining speculation. It's not a card I would consider (atm, unless for mining) and I've been far too busy lately. As an enthusiast I've been slacking lately since I've been too busy. ;)

Thing is, you knew all this already and still tried to cast doubt asking for them.
Why?

No need to try tarnish me for asking for simple data to back up a claim. :whiste:

If I "knew" as you claim, why would I be happy tviceman provided a great explanation?
 

Keysplayr

Elite Member
Jan 16, 2003
21,219
55
91
No I actually haven't been following it beyond the mining speculation. It's not a card I would consider (atm, unless for mining) and I've been far too busy lately. As an enthusiast I've been slacking lately since I've been too busy. ;)



No need to try tarnish me for asking for simple data to back up a claim. :whiste:

If I "knew" as you claim, why would I be happy tviceman provided a great explanation?

Thanks for your reply.
 

Gloomy

Golden Member
Oct 12, 2010
1,469
21
81
Yes certainly if big Maxwell maintains this efficiency, AMD's next gen is going to look hungry, hot and pathetic in comparison! Not to mention Maxwell is a mining monster too.

Scary times ahead for AMD.

Do we even know anything about Pirate Islands? :|
 

KingFatty

Diamond Member
Dec 29, 2010
3,034
1
81
The price should reflect the demand for the product. I'm sure Nvidia knows how to research their intended market and price accordingly.

The wildcard will be if miners seize on this card, and wreck the expected supply/demand. I am very happy that Nvidia engineers are working at improving mining... can't wait to see the big daddy Nvidia cards come out and how they mine...
 

Keysplayr

Elite Member
Jan 16, 2003
21,219
55
91
If a 750Ti at stock speeds is getting around 280 for a hash rate it's pretty easy to extrapolate what the big boys will do when they arrive. But mining is fleeting for GPUs and becoming highly unstable and becoming infiltrated by scams. Unless it becomes regulated.
Anyway. Another topic for another thread.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
If you want to build - diy style - a passive high perf gpu this is a god send piece of kit. And for low power its also without compettion. If you have that needs its downright cheap a a no brainer. Some people is very happy here :)

I assume this is not passive yet? Or is there anyone in the pipeline?
 

OCGuy

Lifer
Jul 12, 2000
27,224
37
91
The thing overclocks like crazy. At 1080p it can beat a stock 7850 in games.

You can have a profile for perf/watt and a profile for gaming, with a fairly noticeable difference.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
I don't know if this was mentioned, in fact I highly doubt it, but it should be mentioned that excellent performance per watt of low end parts directly translates into higher performing high end parts when the architecture is scaled upwards with more shaders or CUDA cores. GPUs are a scaling architecture, they are not like CPUs in this respect. GK107 (which was the first released Kepler) scaled directly with higher CUDA core counts which resulted in the high end GK110 parts performing very well.

This will also apply to Maxwell as an architecture. GM107 for desktop has it's place, but the primary design is for efficiency which is important for mobile *and* desktop - GM107 is already appearing in ultrabooks while not being "cut down", while GK107 had to be cut down from what I remember to fit in a 60W TDP. Correct me if i'm wrong on that, but i'm nearly certain that GK107 had to be cut down to be fit for ultrabook usage - typically, you want 60W TDP or less for that type of form factor. Anyway - This is huge for mobile. This is ALSO huge for desktop. GM107 is double the performance per watt over GK107. This means that the high end Maxwell parts will also be double the performance per watt over Kepler - it isn't inconceivable that Maxwell will be able to achieve GTX 770 level performance in a 110W TDP. The GM107 is doubling GK107s performance in the same TDP window. The same will apply to the higher end parts - GPUs scale linearly when more shaders and cores are added. Maxwell will be no exception, so while GM107 doesn't set benchmark records; the high end Maxwell will be fairly amazing in terms of performance. Performance per watt for GPUs directly translates into high performing high end parts. That doesn't necessarily apply to CPUs because CPUs are software dependent for multi core operation - so while Haswell has amazing performance per watt, that doesn't translate as much in terms of higher straight up desktop performance over Ivy Bridge. This is not true with GPUs- GPUs don't have such a crutch. When they scale from PPW, the double performance per watt also scales with the high end parts.

Also, keep in mind that the R7-265 is a 150W TDP part. The R7-260X is...110W TDP I believe. So it (R7-265) is slightly faster than the 750ti at stock clocks, yes, and it's more performance per dollar for the desktop with no power supply limitations. Yet you will never ever see a R7-265 in an ultrabook. Not unless it is severely gimped or put in a full size laptop that has no TDP considerations. You will see GM107s in ultrabooks and they will not be cut down - the nearest competing AMD part is the 65W TDP R7-250? I think? And that performs a lot slower than the 60W TDP GM107. While AMD can compete meaningfully on the desktop in terms of performance per dollar (ignoring US / Canada prices of course), AMD won't be able to really meaningfully compete in mobile. Maxwell is simply far better than anything else on the market in terms of both performance and performance per watt for mobile - intel ultrabooks with a mobile dGPU will be using a ton of those chips later this year. And like I said...that efficiency also translates into desktop performance when scaled upwards. So it's really a win-win for both mobile and desktop, but we just don't know when the big Maxwell chips for desktop will hit the market. It was 2H 2014 from various rumors i've seen, but who knows. Shall be interesting. I personally think high end Maxwell will perform insanely well given the excellent PPW, so i'm anxious to see what the big die Maxwell brings. And it'll be interesting to see, of course, how AMD responds.
 
Last edited:

Insomniator

Diamond Member
Oct 23, 2002
6,294
171
106
Either way its $10 more or less for 5% more or less performance... who cares. Its not like companies are going to perfectly calculate the price based on the exact performance delta with the current drivers/games etc. Plus rebates are changing every day.

Pick a card, neither are a 'bad' deal IMO, just slightly less optimal. If you regret your choice because of performance it won't be because the other $140 card is so much better, but because you only spent $140 in the first place.
 

FalseChristian

Diamond Member
Jan 7, 2002
3,322
0
71
I like the GTX 750 Ti 2GB. It's 25-30% faster than that good 'ol workhorse the GTX 460 1GB.
The GPU overclocks way better than my EVGA GTX 760 2GB needs no 750w power supply and is priced right.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
I like the GTX 750 Ti 2GB. It's 25-30% faster than that good 'ol workhorse the GTX 460 1GB.
The GPU overclocks way better than my EVGA GTX 760 2GB needs no 750w power supply and is priced right.

Lets keep this realistic. I would gladly run a 780ti on a 400w high quality ps. For the desktop that power demand doesnt make a damn difference unless you want to run passive and have an external ps. Then we can talk. Low cost Mobile is what its for and where it will make a huge difference.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Low cost Mobile is what its for and where it will make a huge difference.

edit: maybe i'm mis-reading here but i'm not sure if you're referring to the 750ti specifically or the architecture in general. You made mention of the 780ti as some basis of comparison so i'm not sure. Anyway, If you're talking Maxwell in general:

Don't confuse the purpose of the chip versus the purpose of the architecture as a whole. Kepler as a whole was designed for efficiency as well, but the purpose of the chips varied. The architecture is for efficiency, the chip (GM107) is designed primarily for mobile. What you're saying is skewing things a bit. Kepler was also designed for efficiency with GK107 being designed for mobile (as is GM107). So let's review. The ARCHITECTURE is designed for efficiency, this helps *both* the desktop GPU and mobile. The GM107 chip itself is designed for mobile as a direct result of that efficiency. Low end chips being efficient directly translates into the high end chips of the same family having high performance.

GPUs scale with higher shader counts and higher CUDA core counts. Kepler scaled linearly, Maxwell will also scale in a similar fashion. Maxwell is not "just" for mobile. Once Maxwell is scaled up with 3000-5000 CUDA cores, the performance per watt will become realized even further; GM107 is getting double the performance per watt of the prior GK107. This means that with scaling, that bigger Maxwell chips will be able to achieve 780ti performance with half the TDP. The excellent performance per watt of GM107 is a definite foreshadowing of performance to come with GM200/204. If GM107 can get near GTX 660 performance in a 60W TDP window, that means the big Maxwell scaled up for 250-300W operation will just be insanely fast. Should be interesting whenever that happens. And it will be interesting to see how AMD responds of course. I'm guessing Hawaii price cuts depending on how aggressive NV is, since multiple sources at hardware websites have indicated that Hawaii is here for the long haul (desktop GPU wise) with no new architecture and no new 20nm parts slated for this year by AMD. I don't know though. We'll see, it'll be interesting. Price cuts and new tech will be good for consumers either way, regardless of which brand one leans towards.
 
Last edited:

Gloomy

Golden Member
Oct 12, 2010
1,469
21
81
All this bickering over a card that's only 10-15 dollars too expensive is hilarious. :whiste:
 
Aug 11, 2008
10,451
642
126
Lets keep this realistic. I would gladly run a 780ti on a 400w high quality ps. For the desktop that power demand doesnt make a damn difference unless you want to run passive and have an external ps. Then we can talk. Low cost Mobile is what its for and where it will make a huge difference.

I agree the best application is for mobile. Especially when shrunk to 20 nm, and if broadwell brings even more power savings, the results could be outstanding.

However, the ability to run without an external six pin adapter is a huge advantage for the desktop as well. It means now that one can buy a $400.00ish i3 desktop on sale from a big box store, dell outlet, etc., drop in a 750 Ti, and have 7790 level performance without screwing around with the power supply. Or to save money, add in the GTX750 and still get, what, around 7770 level of performance.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
I agree the best application is for mobile. Especially when shrunk to 20 nm, and if broadwell brings even more power savings, the results could be outstanding.

However, the ability to run without an external six pin adapter is a huge advantage for the desktop as well. It means now that one can buy a $400.00ish i3 desktop on sale from a big box store, dell outlet, etc., drop in a 750 Ti, and have 7790 level performance without screwing around with the power supply. Or to save money, add in the GTX750 and still get, what, around 7770 level of performance.

Agree its an advantage. But look couldnt some 7750 do that also 2 years ago? I mean its not exactly an revolution here.
 

Keysplayr

Elite Member
Jan 16, 2003
21,219
55
91
Lets keep this realistic. I would gladly run a 780ti on a 400w high quality ps. For the desktop that power demand doesnt make a damn difference unless you want to run passive and have an external ps. Then we can talk. Low cost Mobile is what its for and where it will make a huge difference.

How can you say "Lets be realistic" and then say "I would gladly run a 780Ti on a 400W high quality ps." ?

There isn't anything realistic about that. Not at all. Even if you could do it on one of the best 400WPSU in the world, you wouldn't do it gladly. ;)

If you're going to say lets be realistic, then at least be realistic.
 
Last edited:

OCGuy

Lifer
Jul 12, 2000
27,224
37
91
Absolutely no problem with a high quality ps and say a non oc 4670 and a single ssd.

Wow...

You might want to look @ AT's review.

Also a lesson on amperage as opposed to only wattage would help.
 
Last edited:

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
Wow...

You might want to look @ AT's review.

Also a lesson on amperage as opposed to only wattage would help.

Yes please lecture me. I build my own dac, pre- poweramp and speakers. And especially was very proud of my own low jitter clock design. Tell me about amperage?