Jump to content

Recommended Posts

Was ready to upgrade to a GTX980 and now Nvidia plops this in our laps. At stock speed it's 30+% faster and it even overclocks.

 

So far my one concern in reviews has been the thermal imaging videos. Under full load and with Nvidia's reference cooler, the Titan X immediately rises to its 84C auto-throttling temp, and because Nvidia is not selling the GM200 to card manufacturers the choices for alternative coolers might be very limited. I'm not sure what the point is of having the speed if you can't make full use of it, especially on a $1K card. We know heat is the limiting factor, because with nitrogen cooling the GM200 has been proven stable to over 2GHz. With air cooling, aside from limited o/c headroom (150-200MHz), thermal stress on the card (30C idle/84C load) is tremendous and might cause durability problems.

 

I'm sure Nvidia knows 6GB is just as good as 12 for virtually all games in existence today, and will eventually release a less expensive, cooler (and hopefully 20nm) 6GB model of the Titan. In the meantime I can't justify spending $1K on an inadequately cooled video card that throttles itself under normal use.

Edited by TheMastersSon
Link to comment
Share on other sites

I'm waiting for AMD to release their next round of cards which will trigger Nvidia to release their 980 Ti, or equivalent card. They did the same in the 700 range, and the 780 Ti completely blew the ridiculously expensive Titan of the time out of the water for a fraction of the price. Really want to upgrade the graphics but the 980 isn't a big enough upgrade from the 780 Ti when I know the 980 Ti is just around the corner if I wait a few more months.

Link to comment
Share on other sites

It might not even be that long. One story on Guru3D is already claiming specs for the 980Ti:

 

http://www.guru3d.com/news-story/geforce-gtx-980-ti-has-3072-cuda-cores-and-6gb-of-gddr5.html

 

Unlike the Titan, it will be made by Nvidia's usual partners, so a good selection of coolers should be available. Also gotta love how it's expected to be clocked 10% higher than Titan. The bleeding edge continues being bitchslapped. :)

Link to comment
Share on other sites

Might not call it 980Ti (it's bad marketing - 780 and 780Ti were physically the same card, 980 and Titanx aren't, and you usually want to convey the image of your better product).

Doubt they'll sell it for $699, I'd probably expect $799 or at least $749 - Nvidia has been keen to raise the prices each generation lately.

As to the expectations of how high it's to be clocked... maybe, but then it will instantly throttle down - modern GPU are governed by heat, not clock rate.

Link to comment
Share on other sites

"modern GPU are governed by heat, not clock rate"

 

Reports say they've reached the die size limits for 28nm and I can't wait to see a 20nm version of Titan. Another post in the same discussion said it's coming but not until September. Meanwhile Newegg is already sold out of the Titan X (took under a week) so it looks like I'm back to a 980 for an upgrade.

Link to comment
Share on other sites

For what it's worth my other upgrade interest, M.2 x4 SSDs, have the same thermal problems, e.g.

 

http://www.anandtech.com/Show/Index/8979?cPage=13&all=False&sort=0&page=11&slug=samsung-sm951-512-gb-review

 

It takes less than two minutes of sustained writing for the SM951 to start throttling itself to a whopping 75MB/s. It's not a big deal for the devices these drives were designed for (laptops), where apps rarely if ever write this much data at once, it's still very disappointing for video editing and other desktop use. I wonder how long until we start seeing heatpipes and water cooling for M.2? :)

Link to comment
Share on other sites

What makes you think there will be any sort of a 20nm version of the current cards? Or are you referring to "Titan" as the brand Nvidia now uses for $1,000+ video cards?

 

SSD aren't going to get water cooling (LOL), and well-built laptops already get heatpipe type spreaders. The issue here is that the SSD with its M.2 interface lacks an aluminum case to serve as a heatsink. The actual thermal load is still minimal, but there's nowhere for the heat to go.

Link to comment
Share on other sites

Currently it's strongly suspected that NV and AMD will skip the 20nm process, since the benefits over 28nm are very marginal. The 20nm process doesn't appear to be useful for reducing the consumption of very powerful and complex devices due to its leaks.

http://www.fudzilla.com/news/processors/36968-tmsc-to-invest-16-billion-in-advanced-chip-factory

 

That means the next half-generation would stay at 28nm, and maybe the next generation, with possibly another process further away for a further generation.

 

 

Remember also that 28nm, 22nm FinFET, 20nm, 14nm FinFET, etc, they're all little more than names. Yes, this is the distance between certain features in a certain pattern, but there is some flexibility in how to measure it, what features "count" and "don't count".

The outcome is not linear at all. Each tech node is an entire industrial chain with a lot of constants and variables. It's like a 14ft car, a 17ft car and a 20ft car - unless it's the same model, all differ in so much more, so you can't talk about them in terms of buying X feet of car.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...