Overclock.net banner

[Hexus] Review: Intel Core i9-7900X (14nm Skylake-X)

87K views 2K replies 160 participants last post by  Darklyric 
#1 ·
#2 ·
Wow...no NDA issue?

Also:

Our chip seems to be a good one and had no qualms about running at 4.7GHz across all 10 cores. Heck, it needed only 1.25V to make it happen.
 
#8 ·
#10 ·
It seems good except this part:
Quote:
Could frequencies go higher with more voltage? Probably, but putting 1.3V through the Core i9-7900X veins resulted in temperature soaring beyond 100ºC and automatic throttling. We swapped out our favoured Noctua NH-D15S in favour of an EVGA CLC 280 liquid cooler but even that couldn't cope with the increase in voltage.
Yikes
 
#12 ·
Quote:
Originally Posted by jelome1989 View Post

http://www.bit-tech.net/hardware/2017/06/16/intel-core-i9-7900x-and-x299-chipset-revie/6

4.6ghz 7900x slower than 4.4ghz 6950x on AotS? And by 15-20%? It's also lagging behind other benchmarks.
You are looking at games benchmarks only (which is still better than ryzen and the upcoming threadripper)

Non-gaming benchmarks showing some decent boost

If gaming is all what you care about then there is no reason to buy threadripper or 1800X anway because it has the worst performance
 
#13 ·
Quote:
Originally Posted by Lipos View Post




Both handbrake. What the hell? Don't know, from the hexus review it doesn't look impressive at all. But let's wait and see.
This is what i'm saying. Its like 25-30% faster, but it has 25% more cores. So yeah. Threadripper might end up being a great product.
 
#14 ·
Quote:
Originally Posted by ThePath View Post

Performance look better than what I expected

I think 10 core skylake-X could even challenge or even beat 12 core Ryzen
Are you reading the same charts? This is not much faster than the 1800X in most of the tests. It even loses to the 6950X in some. It's only looking like a big boost in hand brake in both reviews. Otherwise to be at 4.7ghz, it's actually the opposite of impressive. Also that temperature and power draw. Intel should sell delid kits themselves.
 
#15 ·
Quote:
Originally Posted by FLCLimax View Post

Are you reading the same charts? This is not much faster than the 1800X in most of the tests. It even loses to the 6950X in some.
I think what he meant to say was the 18 core i9 might be faster than 16 core thread ripper.
 
#19 ·
Quote:
Originally Posted by jelome1989 View Post

Ah.

It should be... it seems that single core IPC is about equal to Kaby so it was definitely underperforming in those tests
Yeah, check out what they said about warhammer,
Quote:
Total War is well off the pace. The game appears to suffer from unexplained stuttering, and once again, the latest BIOS saw average FPS climb from 48.2 to 74.9
http://hexus.net/tech/reviews/cpu/107017-intel-core-i9-7900x-14nm-skylake-x/?page=6

Looks like there's room for improvement.
 
  • Rep+
Reactions: BiG StroOnZ
#20 ·
It looks pretty good but I don't really agree with the "all out assault" part. It's 10 cores for 999. Okay. That's good, but not all-out-assault good. I think the 10 core for $500 or something, that would have been all out. This is Intel responding in a calm measured way. They don't want to try mess up their margins much. I don't blame them at all. This is probably what I would have done. This way there is plenty of downward price adjustment possibility later when the higher end parts come out.

The unsoldered part is a bit concerning though. The average person can delid (I mean average american here, income wise) an i5 part without much worry. Delidding a $1000 part will make people very nervous.

I had an idea about just taking the chip and leaving it sitting in fairly pure >90% MEK over a weekend or something. Do you folks know if that would work? Anyone try it? I'm VERY sure it would not damage the chip but I'd try on a cheaper one first.
 
#21 ·
Quote:
Originally Posted by ThePath View Post

You are looking at games benchmarks only (which is still better than ryzen and the upcoming threadripper)

Non-gaming benchmarks showing some decent boost

If gaming is all what you care about then there is no reason to buy threadripper or 1800X anway because it has the worst performance
I know that. You don't get my point. I'm questioning the fact why it was slower in games while it has about the same single core IPC as Kaby based on other tests. It was already answered though
 
#22 ·
Quote:
Originally Posted by FLCLimax View Post

Are you reading the same charts? This is not much faster than the 1800X in most of the tests. It even loses to the 6950X in some. It's only looking like a big boost in hand brake in both reviews. Otherwise to be at 4.7ghz, it's actually the opposite of impressive. Also that temperature and power draw. Intel should sell delid kits themselves.
Against 1800X

30% faster in handbrake

28% faster in cinebench

Faster in gaming benchmarks ( cheaper 6 and 8 cores skylake-X will be faster also)

8 cores skylake-X (i7 7820X) has 300Mhz higher base clock than 7900X. So, it should be easily faster than 1800X in almost all benchmarks (whether it is multi-cores or single core)
 
#23 ·
It seems like 4.6 GHz is the 24-7 clockspeed that we can expect.

Also the people who were expecting major IPC improvements over 4 core Skylake were not correct. There won't be much in the way of IPC until the next architecture.

From Bit-Tech: https://www.bit-tech.net/hardware/2017/06/16/intel-core-i9-7900x-and-x299-chipset-revie/8
There seems to be much more headroom with Skylake-X than its predecessor, and the main limiting factor is temperature if our CPU is anything to go by. We plumbed in 1.3V as a starting point and crept up from 4GHz all the way to an astounding 4.7GHz, which is 300MHz higher than we managed with the Core i7-6950X. Even more impressive was the fact that it was still completely stable with just 1.28V - far lower than the 1.44V we needed with the older CPU.

However, temperatures were definitely a concern with Cinebench and Terragen pushing 100°C with our 240mm AIO liquid cooler. As a result, while stable and potentially tameable under custom water-cooling, we decided to go for 4.6GHz for benchmarking, which required a super-low 1.22V. Interestingly our Core i7-6950X ran much cooler despite using a significantly higher voltage, albeit at 4.4GHz. This could well be due to thermal paste having been used between the heatspreader and CPU core with the new Skylake-X CPUs, in which case delidding could potentially yield significant benefits given the high heat density.

That's disappointing. A 5960X could do 4.4 - 4.6 GHz as well. A few golden samples could do 4.7 GHz.

Now granted, the 5960X has 8 cores versus the 10 cores on this 7900X on a 22nm process rather than this 14nm+ process for Skylake E, but it's still disappointing. At least 4.6 GHz is the lower end of what you can expect with a 6700K.

Overall though it looks like Intel will be perhaps 20-25% faster on single threaded than Threadripper. That's assuming this is stuck at 4.6 GHz, with many samples at 4.5 GHz. A few golden samples might push to 4.7 GHz 24-7.

So basically no improvements apart from the extra cores, and the slightly faster IPC. Also, I don't think delidding will get more than a couple of hundred MHz (it didn't on the 3770k, 4790k, or the 7700K).

Let's see if the HCC CPUs are soldered.

Quote:
Originally Posted by Lipos View Post




Both handbrake. What the hell? Don't know, from the hexus review it doesn't look impressive at all. But let's wait and see.
I'm thinking that it's not as big as you think. With 25% more cores, it's just slightly faster per core.

Basically take the Ryzen scores at 4 GHz, and for apps that scale perfectly with cores, multiply then by 2. Then you should have a good educated estimate of Threadripper 16 core performance. Not all benchmarks of course scale linearly with cores (games especially). Also, we don't know what the penalty is for off die CCXs. We also don't know how quad channel RAM will affect Ryzen. Too many variables - so multiply by 2 is just a ballpark.

There are a few other weird parts. The Bit Tech review showed Ashes a lot slower. With the Ryzen optimizations, that should no longer be the case.

Quote:
Originally Posted by extracrunchy View Post

It looks pretty good but I don't really agree with the "all out assault" part. It's 10 cores for 999. Okay. That's good, but not all-out-assault good. I think the 10 core for $500 or something, that would have been all out. This is Intel responding in a calm measured way. They don't want to try mess up their margins much. I don't blame them at all. This is probably what I would have done. This way there is plenty of downward price adjustment possibility later when the higher end parts come out.

The unsoldered part is a bit concerning though. The average person can delid (I mean average american here, income wise) an i5 part without much worry. Delidding a $1000 part will make people very nervous.

I had an idea about just taking the chip and leaving it sitting in fairly pure >90% MEK over a weekend or something. Do you folks know if that would work? Anyone try it? I'm VERY sure it would not damage the chip but I'd try on a cheaper one first.
They are clearly hoping that people will keep buying because of the single threaded performance over Ryzen and because "it's Intel".

Let's face it, the Ryzen 1700, and 1600 seem to be the best as far as price:performance.

I think that this might be a big boon for silicon lottery as most people will want the CPU to be delidded professionally. But even then ... you lose your warranty.

There is also the matter that next year, AMD will release it's Zen+ CPUs. All Intel has is 14nm++ and 10nm, which is according to their own charts, slower than 14nm++ , although more power efficient (10nm+ will be faster though).
 
#26 ·
From Bit Tech:
Quote:
There seems to be much more headroom with Skylake-X than its predecessor, and the main limiting factor is temperature if our CPU is anything to go by. We plumbed in 1.3V as a starting point and crept up from 4GHz all the way to an astounding 4.7GHz, which is 300MHz higher than we managed with the Core i7-6950X. Even more impressive was the fact that it was still completely stable with just 1.28V - far lower than the 1.44V we needed with the older CPU.

However, temperatures were definitely a concern with Cinebench and Terragen pushing 100°C with our 240mm AIO liquid cooler. As a result, while stable and potentially tameable under custom water-cooling, we decided to go for 4.6GHz for benchmarking, which required a super-low 1.22V. Interestingly our Core i7-6950X ran much cooler despite using a significantly higher voltage, albeit at 4.4GHz. This could well be due to thermal paste having been used between the heatspreader and CPU core with the new Skylake-X CPUs, in which case delidding could potentially yield significant benefits given the high heat density.
From Hexus:
Quote:
We'll start with the usual proviso: your overclocking mileage may vary and discussions with various partners lead us to believe that frequency headroom fluctuates significantly from one sample to the next. Our chip seems to be a good one and had no qualms about running at 4.7GHz across all 10 cores. Heck, it needed only 1.25V to make it happen.

Could frequencies go higher with more voltage? Probably, but putting 1.3V through the Core i9-7900X veins resulted in temperature soaring beyond 100ºC and automatic throttling. We swapped out our favoured Noctua NH-D15S in favour of an EVGA CLC 280 liquid cooler but even that couldn't cope with the increase in voltage. The good news for users who have invested in a high-quality cooler for LGA2011v3 is that compatibility with LGA2066 has been retained with identical mounting-hole positions.
So it would appear that 1.3V/4.7 GHz might be the hard limit if you're running an AIO.

Man that pigeon poop TIM is really hampering things. And this is just for the 10C. To get any kinds of good OC on the 12-18C parts delidding is a must.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top