Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Intel Core i9-7900X review: The fastest chip in the world

arstechnica.com

Post reply on HN
Print view View original

21–30 of 119 posts

Re: Intel Core i9-7900X review: The fastest chip in the world

#21

Post by FiReaNG3L »

paulmd wrote: post #20

The real problems with Skylake-X are chipset cost, power consumption, shitty partner boards, and TIM. All of these are forgivable given the performance - except the TIM. Chipset cost will come down in 6-12 months after launch like it always does. This is part for the course, at launch X370 boards for Ryzen were going for well over $250 as well. Power consumption is a consequence of AVX512 and the mesh interconnect al…

I had to google it so I don't think im alone here:

TIM = Thermal Interface Material

Re: Intel Core i9-7900X review: The fastest chip in the world

#23

Post by brigade »

rkv wrote: post #9

Out of the loop when it comes to the latest processor technology. How does this chip maintain the same TDP as a 6850K but with 4 more core? Same size (lithography), roughly the same freq, memory size/types.

Power/frequency scaling is quadratic since voltage can decrease with frequency, so even the 10% difference in base clocks could explain it, depending on where in the voltage/frequency curve they are.

Plus, this is Intel's 3rd iteration on the same process; even if the feature size doesn't change you can extract a bit more power efficiency with 2 years of feedback.

Re: Intel Core i9-7900X review: The fastest chip in the world

#24

Post by paulmd »

FiReaNG3L wrote: post #21
paulmd wrote: post #20

The real problems with Skylake-X are chipset cost, power consumption, shitty partner boards, and TIM. All of these are forgivable given the performance - except the TIM. Chipset cost will come down in 6-12 months after launch like it always does. This is part for the course, at launch X370 boards for Ryzen were going for well over $250 as well. Power consumption is a consequence of AVX512 and the mesh interconnect al…

I had to google it so I don't think im alone here: TIM = Thermal Interface Material

Yup. TIM goes on the die to transfer heat to the Integrated Heat Spreader (which is the lid you normally see). Intel has an issue with their thermal paste, it doesn't properly fill the void and doesn't contact the IHS well, so heat just builds up in the die. Check out this amazing chart from the Tom's Hardware review:

http://i.imgur.com/7BIJmxS.png

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skyl...

Heat just is not getting to the IHS properly on these chips, and heavy overclocking just makes the whole thing worse.

It's really been a slow-burning problem since Ivy Bridge, where Intel switched from soldering the lid to TIM + an adhesive. Solder has long been preferred for its superior heat transmission, but Intel says say smaller dies have problems with cracking over time due to thermal cycling. However, AMD has been happily soldering the lid on much smaller Ryzen dies, so apparently it's not all much of an issue in practice.

Well, you can get away with that on a processor that puts out 50 watts during normal operation. It's been an issue for a while on the unlocked/overclockable SKUs, particularly on the latest 7700Ks, but even an OC'd 7700K only puts out ~100W, so it was relatively manageable. Extreme overclockers could delid and replace the thermal paste with something better (often liquid metal like Conductonaut), which does help performance quite a bit.

But, with the higher TDP of Skylake-X, this has become a pressing issue just for normal operation. Things change when we're talking about a $1000 processor that needs to be delidded to sustain boost at stock settings. That's just not acceptable.

I almost would rather have a bare die at this point. Mounting pressures are no longer insane so it wouldn't be as terribe an ordeal to mount as as Athlon XPs were back in the day (god forbid your screwdriver slip on that bracket, with 50+ pounds of force you are guaranteed to gouge something).

Re: Intel Core i9-7900X review: The fastest chip in the world

#25

Post by vosper »

paulmd wrote: post #20

The real problems with Skylake-X are chipset cost, power consumption, shitty partner boards, and TIM. All of these are forgivable given the performance - except the TIM. Chipset cost will come down in 6-12 months after launch like it always does. This is part for the course, at launch X370 boards for Ryzen were going for well over $250 as well. Power consumption is a consequence of AVX512 and the mesh interconnect al…

At the desktop level I don't get why people care that much about power consumption? It means you have to dissipate more heat, okay, that means you can't use a cheap cooler. But it AFAIK even an extra 100w is cheap in the most expensive areas, especially when contrasted against productivity or cigarette breaks, or people sometimes being 20 minutes late to work...

Re: Intel Core i9-7900X review: The fastest chip in the world

#26

Post by paulmd »

vosper wrote: post #25
paulmd wrote: post #20

The real problems with Skylake-X are chipset cost, power consumption, shitty partner boards, and TIM. All of these are forgivable given the performance - except the TIM. Chipset cost will come down in 6-12 months after launch like it always does. This is part for the course, at launch X370 boards for Ryzen were going for well over $250 as well. Power consumption is a consequence of AVX512 and the mesh interconnect al…

At the desktop level I don't get why people care that much about power consumption? It means you have to dissipate more heat, okay, that means you can't use a cheap cooler. But it AFAIK even an extra 100w is cheap in the most expensive areas, especially when contrasted against productivity or cigarette breaks, or people sometimes being 20 minutes late to work...

I fully agree, and what's more Intel has the performance to back it up. This chip pulls a lot but it's wicked fast, it's a massive step forward in framerates. It combines the minimum-framerate improvements of HEDT/Ryzen with the single-threaded performance of Kaby Lake. Oh yeah and AVX512 too.

For a sense of perspective here, going from a circa-2012 2600K to a current 7700K is a 40% jump in performance, so it's roughly equivalent to 4-5 years of gains at Intel's usual tempo - only you also have 10 cores on this platform. This thing is an absolute monster for gaming or other tasks that lean heavily on single-threaded performance.

But the power consumption is really the triggering issue for the problems with shitty partner-boards overheating and the TIM. The TIM is really the showstopper right now.

Re: Intel Core i9-7900X review: The fastest chip in the world

#27

Post by Sephr »

paulmd wrote: post #15
arcanus wrote: post #6

Earlier quoted context omitted.

I'll answer your subtle comment quite explicitly: never. Intel clearly rushed this out to prevent AMD from having the perception of leading the space, at least in terms of the largest core count. The article even references this. I just hope that it stays competitive, as this is clearly a win for consumers.

Skylake-X certainly would have existed regardless of Ryzen, Intel wants to sell these chips to Google/Amazon/Facebook as Xeons. The idea that this uarch is a reaction to AMD is absurd, this uarch has been on the roadmap for years and at most was pulled forward a few months. People will attribute anything one of AMD's competitors does to fear of AMD - 780 Ti, 980 Ti, 1080 Ti, Skylake-X, you name it. It's frankly a lit…

> Intel wants to sell these chips to Google/Amazon/Facebook as Xeons

They have already been selling vastly superior Skylake-SP (and -EP) Xeons to Google/Amazon/Facebook since 2015.

Re: Intel Core i9-7900X review: The fastest chip in the world

#28

Post by Nition »

NamTaf wrote: post #18
tracker1 wrote: post #5

Another issue beyond the fact that the X299 boards are problematic in terms of how many, and/or what slots you can and can't use seems to be excessive heat and stability in some of the reviews I've seen. I'm still running a i7-4790K at home, and though I'd like something with more cores... nothing is compelling enough to bring me to switch given the costs involved. If I were building new, would most likely go with an…

CPU gains for the last few gens have mainly been in perf/watt and, now, core count in lower-class chips thanks to AMD. The biggest benefit of a system upgrade for the last few gens, from a user perspective, has been for the chipset's gains - NVMe, USB3, USB-C, etc. more than raw performance. That's slowly trickled up but the real-world gains in many tasks haven't been enough to really justify it. I know many people s…

Similarly, I only just upgraded from an i7 860 (2009) to a 6700K at the end of last year. In performance benchmarks, it's only twice as fast as my old one, and that's over seven years. Big difference to how things felt in the past. Imagine if your 66Mhz Pentium still ran well enough that you weren't sure it was worth upgrading to a 500Mhz Pentium III. Same time span.

Re: Intel Core i9-7900X review: The fastest chip in the world

#30

Post by chx »

NamTaf wrote: post #18
tracker1 wrote: post #5

Another issue beyond the fact that the X299 boards are problematic in terms of how many, and/or what slots you can and can't use seems to be excessive heat and stability in some of the reviews I've seen. I'm still running a i7-4790K at home, and though I'd like something with more cores... nothing is compelling enough to bring me to switch given the costs involved. If I were building new, would most likely go with an…

CPU gains for the last few gens have mainly been in perf/watt and, now, core count in lower-class chips thanks to AMD. The biggest benefit of a system upgrade for the last few gens, from a user perspective, has been for the chipset's gains - NVMe, USB3, USB-C, etc. more than raw performance. That's slowly trickled up but the real-world gains in many tasks haven't been enough to really justify it. I know many people s…

> NVMe

While NVMe benchmarks are stunning I really am curious whether in an ordinary setting the difference to SATA can be felt at all. Not measured -- felt. When the era of swap ended because CPUs / chipsets finally could handle enough RAM and when we went from HDD to SDD, those could be felt for sure.

> This is marginally different for laptops

In laptop world, the problem is that with the spread of thin craze most laptops are now running 15W CPUs instead of 35W like in the old age and so there is little performance increase, no core count increase.

Post reply on HN