FluffmeisterHehe, post success confirmed.

:slap:

XzibitThere is also another issue that might present itself with DX12 and Nvidia cards

Reference models are reaching 80C-81C in open bench test. DX12 is suppose to keep the GPU feed more consistently with information and will have less time for GPU Boost to down clock it in games leading to higher voltage/temps and possibly throttling.

thats interesting that they would set a aggressive limit like that especially seeing as how this nv engineer does not mention it and says a temp over 90c.
he does also say to use dx12 in valley witch is also a bit interesting.
www.pcgamer.com/sli-overclocking-guide-nvidias-tom-petersen-on-maximizing-multi-gpu-performance/#!

sure its 4gb or 3.5 + .5? lol

But seriously? a 750ti without SLI support?

I enjoyed my 750ti. It clocks really good and its quite. but an insane 4gb? for what?

xfia:slap:



thats interesting that they would set a aggressive limit like that especially seeing as how this nv engineer does not mention it and says a temp over 90c.
he does also say to use dx12 in valley witch is also a bit interesting.
www.pcgamer.com/sli-overclocking-guide-nvidias-tom-petersen-on-maximizing-multi-gpu-performance/#!

It's obviously not really an issue if DX12 is extracting more performance, I made this clear in another thread, but that went silent. :P

xfiathats interesting that they would set a aggressive limit like that especially seeing as how this nv engineer does not mention it and says a temp over 90c.
he does also say to use dx12 in valley witch is also a bit interesting.
www.pcgamer.com/sli-overclocking-guide-nvidias-tom-petersen-on-maximizing-multi-gpu-performance/#!

Firstly, the 80°C is a target that the Boost algorithm aims for. It doesn't mean that the 80°C is a hard limit. Secondly, the article you linked to is basically approving third party OC tools. Precision and Afterburner and the like allow the target temp to be raised in excess of 90°C.
As for the DX12 thermal throttling of reference cards on an open air bench, I guess that might be a problem for those people that use a test bench but don't know how to use an OC utility, and don't have the common sense to provide airflow. Reference (blower) cards are known to be less effective in cooling than AIB designs, and the cooling is largely predicated upon chassis airflow through the length of the card.
The only real issues I can see is that reviewers using reference cards and an open air stand might under-report the cards ability due to throttling, but anyone using a better cooling solution, and/or an OC utility to either raise Tjmax and/or alter the stock fan profile if required isn't likely to be greatly inconvenienced. I'm also pretty sure that the fan profile could be altered at a driver level for those with a complete absence of tech know-how...i.e. the exact same solution AMD came up with for their thermally throttling cards.

Much ado about nothing.

Doesn't stop people from being hypocrites about reference coolers.

xfia:slap:
thats interesting that they would set a aggressive limit like that especially seeing as how this nv engineer does not mention it and says a temp over 90c.
he does also say to use dx12 in valley witch is also a bit interesting.
www.pcgamer.com/sli-overclocking-guide-nvidias-tom-petersen-on-maximizing-multi-gpu-performance/#!

That's a water cooled system so it wont exhibit those issues. I believe the hard limit for Maxwell is 93-95C.

nothing? I thought that engineer did a decent job but since he didnt even mention it.. thats weak. why 80c?

xzibit had every right to think there might be a issue when toms says thermal ceiling and tpu says temp limit but I guess maybe smoke should be writing the articles if he thinks he is the all knowing technical guru. problem with that though is you have to be a people person.

xfiaxzibit had every right to think there might be a issue when toms says thermal ceiling and tpu says temp limit

80°C is the default target thermal limit. You asked how 90°C+ is obtained on a Maxwell card - the answer is by raising the Tjmax limit.

xfiabut I guess maybe smoke should be writing the articles if he thinks he is the all knowing technical guru. problem with that though is you have to be a people person.

I'm wounded!....oops, wrong verb. I meant "I'm yawning". Hyperbolic personal attack, and we barely know each other.:eek:
Firstly, the solutions are pretty self evident as I outlined, and as millions of people put into practice every day when they use their PC's.
Secondly, I don't claim to be a any kind of all-knowing technical guru, although I have written articles, so maybe the problem isn't as large as you envisage.
Memorable Overclocking Friendly CPUs
History of the Microprocessor and the Personal Computer
Iconic Hardware
The Rise and Fall of AMD
The History of the Modern Graphics Processor
Famous and Infamous Tech Quotations & Predictions
and an informal but reasonably comprehensive Watercooling Guide (along with some other smaller help-type articles spread around other sites, and two more historical articles I'm presently working on).

XzibitThere is also another issue that might present itself with DX12 and Nvidia cards





Reference models are reaching 80C-81C in open bench test. DX12 is suppose to keep the GPU feed more consistently with information and will have less time for GPU Boost to down clock it in games leading to higher voltage/temps and possibly throttling.

And thanks to TSMC's eternal delays with the 16 nm process/FinFET's, we ain't going to see any progress on that front for years. I hope Samsung gets its 14 nm process mass-production ready soon, so we can get some real competition going again.

I guess you wrote those or your being an ass.
If you want to be involved with a conversation you could actually read it man.
I didnt say it was some huge issue.. I said it was interesting and seemed aggressive due to the target temp I know amd set for 290's.
the engineers must have thought 80c was a good temp for a reason and to explore what the reason might be could be something nice to read for people instead of half way insulting comments.

edit-I just wanted to add I dont think anything bad of you and have respected what you have said several times but maybe its just not your best day

2,5 GB VRAM again, LMAO

xfiaI guess you wrote those or your being an ass.

And why would I claim credit for something I didn't write? If your comment is just a means to deliver an insult you can skip down to the next quote
I guess if you don't believe me you couldclick my name (Graham Singer) on the byline> click "Techspot profile" under my name on the page it takes you to, and read the profile post I just put up just for you.

xfiaI didnt say it was some huge issue..

I thought it was obvious I was referring to the post you quoted, not what you said, given that you weren't the one that mentioned reference coolers and an open test bench. What you queried (re: the 80C thermal limiter and the 90C+ mentioned in the article/vid) I attempted to answer in a civil manner. The matter brought up by the other poster you quoted is largely inconsequential as they are fully aware. That got the response it warranted.

xfiaI said it was interesting and seemed aggressive due to the target temp I know amd set for 290's.

That was explored at Hawaii's launch. The 290/290X limits are somewhat higher than the norm because of the limitations of power/heat production/noise consideration and the reference cooling solution AMD chose to employ (see basically any launch review for confirmation). The 95C limit ensures that the cards using the reference cooler don't throttle themselves excessively. The limit is still in place as part of AMD's specification, but largely negligible, for AIB designed cooling solutions.

xfiathe engineers must have thought 80c was a good temp for a reason and to explore what the reason might be could be something nice to read for people instead of half way insulting comments.

It is exactly the same paradigm as AMD used - power/heat/noise vs temp vs performance. Why Nvidia chose 80C is best explained when that temp was set when GPU Boost 2.0 intro'd with Titan. If you're still aggrieved please feel free to put me on ignore.

Common response: You don't need 4gb!!!! Slow GPU!!! Silly!!!! Can't use!!!

Correct response: You probably won't need the extra memory, so it probably won't make any difference, but if you do, then you'll be glad you have it.

I'm not salty and I dont think your a liar. If I thought that I would not have expected more from you and ignored what you said.
people dont come read reviews for health.. they are just bored or looking for direction.. I mean one great post could help thousands of people over time or inspire a teenager to build a first gaming rig ya know..
I know ya got that love for great literature and what your words can do.

thank you for the titan article.. I have never seen a better explanation with compressive graphs for the subject.

xfiawhat is the ram going to put the price up to? near the price of a 960?

Who knows what it will really be, but the 2GB GTX 750s are typically retailed at $10 more than the same card with 1GB. So $20 more for 2GB more?

MikeMurphyCorrect response: You probably won't need the extra memory, so it probably won't make any difference, but if you do, then you'll be glad you have it.

Only if it was free.

.........baaaah this is card is too punny.............i'll just wait for the 8 gig tripple slot super supper clocked version:roll:. Gotta love it.......

MikeMurphyCommon response: You don't need 4gb!!!! Slow GPU!!! Silly!!!! Can't use!!!

Correct response: You probably won't need the extra memory, so it probably won't make any difference, but if you do, then you'll be glad you have it.

It's always better to have too much, than not enough. (GTX 970 *cough*)

Asus Manufacturing Manager: Hey boss we have some extra vram chips lying around, what about we put it in our unsold low end gpu and then we ask for more $$$.

Asus CEO: Nice idea!

btarunrIn a bid to woo those who choose graphics cards by memory amounts (and cars with engine-displacement), ASUS rolled out a 4 GB variant of its GeForce GTX 750 Ti Strix lower mid-range graphics card. Pictured below, the card is built identical to the standard 2 GB model, but with double the GDDR5 memory amount. It features out of the box clock speeds of 1124 MHz core, 1202 MHz GPU Boost, and 5.40 GHz (GDDR5-effective) memory. Based on the 28 nm GM107 silicon, the GTX 750 Ti features 640 CUDA cores based on the "Maxwell" architecture, and a 128-bit wide GDDR5 memory interface. ASUS didn't disclose pricing.

How much is it

purplekayceeHow much is it

Did you not read the last four words in the post you just quoted.

LOL even this has more vram then the gtx 970...

dgwineLOL even this has more vram then the gtx 970...

Though the 970 vram clocks 29.6% faster

xfiawhat is the ram going to put the price up to? near the price of a 960? surely a amd gpu with more horsepower could be had and its not like having a full dx12 gpu is going to matter for a long ass time.
guess someone has to look blindly and try keep nv rep up since they don't seem to be to worried about it that much.

Why is everyone forgetting that Nvidia and AMD have made it very clear that AMD's GCN and Nvidia's Fermi and Kepler and early Maxwell are DX12 capable. So a 750 Ti is a DX12 GPU. Besides the older DX10 cards not being able to run DX11 was never a hardware deficiency anyway, It was Just Microsoft adding code to the updated DX code to keep DX10 cards from running DX11, (marketing) But thanks to Mantle threating Microsoft, They decided to not do this with DX12.

P-40EWhy is everyone forgetting that Nvidia and AMD have made it very clear that AMD's GCN and Nvidia's Fermi and Kepler and early Maxwell are DX12 capable. So a 750 Ti is a DX12 GPU. Besides the older DX10 cards not being able to run DX11 was never a hardware deficiency anyway, It was Just Microsoft adding code to the updated DX code to keep DX10 cards from running DX11, (marketing) But thanks to Mantle threating Microsoft, They decided to not do this with DX12.

:ohwell:

techy1you know what would be better card for upgrade VRAM to 4GB ... a GTX 970 ;)

No one needs more than 640K.