Hardware Intel Core i9-7900X (Skylake-X) Review

please Note RYZEN is Compatible with ECC modules but it Does Not have ECC Features enabled (1Bit correct and Stop on uncorrectable error) at the moment on RYZEN you're just using a ECC with one extra chip disabled at the moment (and slower ram speed)

unsure if AMD is going to release a AGESA update to allow ECC features to be used, Really depends if they want to artificially disabled it like intel for profit (as Intel dam make sure you can't use ECC to "Protect its pocket" and force you to buy a 2 core i3 or XEON) AMD has allowed ECC on AM2 before if you had a ASROCK or ASUS mobo (but was disabled on all other future CPU sockets as far as i know)

Threadripper Might come with ECC Feature support/Validated out of the box as it may be marketed as a Workstation CPU (but it is just 2 RYZEN cores under the lid), where as RYZEN is a consumer CPU and Epic is a Server CPU

 

the Threadripper post about it been an issue i cant see why as RYZEN is 2 Quad core CPUs connected via a fabric the Threadripper will be the same just its connecting 2 RYZEN cores together (i be more interested on how well the Quad channel works across the fabric but with 64 lanes between the RYZEN cores i cant see that really been a problem)

me does like the 10 Core intel CPU (more for games that are still bit stuck on single threaded side) but it's a lot of extra £ for 10% improvement for single threaded performance,, something like 70% more price for 10% more performance over RYZEN (maybe less price Gap on the threadripper 10 core vs intel 10 core but single threaded performance will not be any better on ThreadRipper vs RYZEN)

these new intel CPUs was a reaction to RYZEN and later Threadripper (and Intel should've Never made the quad core part for a extreme CPU product as it makes it over complicated for mobo makers or even buyers, as features are disabled if you don't have the higher end CPU)

 

I heard that a lot of OEMs are refusing to even work with X299 at this point. Do you blame them? I certainly don't.

Also, it's not the PCI-E lanes that connect the cores. It's the memory frequency that is most critical.

Threadripper will be four quad core modules for the 16c models that are on one substrate. They are all interconnected via the infinity fabric. AMD has really made these chips super scalable, bloody brilliant.

 

This. I saw the technical explanation due to the specific expansion co-efficient. If you always keep it heated there's not problem. It's the heat-cool cycling that causes micro-cracks.

HOWEVER I'm still not fully convinced Intel couldn't use a better TIM as evidenced by all those replacing it with better compounds, and older Intel CPUs used for years are still going strong so is it REALLY that serious issue?

Still got my 5960X! I'm more concerned about how long the board will last. Most don't survive beyond 3 years and it's been 2 now.

Leaves the door open for AMD to do Zen2 with 32 lanes on premium 8 core models, and 16 on 4-6 core.

 

That's the problem with Intel changing chipset with nearly every chip, when you have a failing board there is no replacements to be found

 

I would love this to delid and play around with and see what clocks are doable even on a standard all in one. 10 cores at 5ghz stable seems crazy. Way outta my price budget though.

Threadripper base clocks need to be similar to Ryzen for it to show scalability. Which I think is the reason it's geekbench scores suck. Even assuming the 3.39 is the base clock they could be very far behind.

Be fun to play and bench both from the geek perspective, I am never likely to own either lol

 

Either that or 2nd hand board prices on ebay and the like go crazy

 

The Geekbench score that has been floating around for Threadripper is BS.

http://hexus.net/tech/news/cpu/107005-amd-ryzen-threadripper-1950x-16c32t-cpu-gets-geekbenched/

Near identical score in single core result to 8 core Ryzen despite clock deficit?
Only such a tiny gain in multi core score from doubling core count?

Smells very fishy.

 

I was aware of them saying that solder cracked the dies. However something just doesn't add up. I guess the only way to really know would be to delid a Broadwell E Xeon and find out.

They may be using crap paste to limit overclocking. Just a thought, likes. If you limit your CPUs by 400mhz they will live longer surely?

So yeah I agree that their reason was straightforward however I can't get around the ming of fish.

 

It's the same core silicon underneath. TR is just multiples of Ryzen put together on a huge package. The limitation will be power provision, and apparently TR (samples, at least) is guzzling monster when it comes to Amp draw.

I wouldn't trust ANY TR benchmark at this stage. Platforms are early. God knows what that Alienware will launch like.

 

Of course it doesn't add up, what would it cost them to find a solution to the cracking problem? 1 million? 5 million? 50 million? Peanuts for a company with an annual R&D budget in the billions.

And the benefit isn't just for the handful of people who oc, think oems being able to get away with smaller heatsinks, less noise due to lower fan speeds etc...

 

Yeah I totally agree man. It's BS of the highest order.

 

Just seen this on another forum - it raised a smile or two

 

haha it's the girl that's so funny in that

 

That's his wife.

I saw that video a few days ago, and it definitely got a few laughs out of me too.

 

I have no clue how the "cheap TIM" meme continues to propagate. The issue is 100% down to the distance between the surface of the die and the underside of the heatspreader. Reducing this gap contributes 100% of the benefit of delidding/relidding.

So why is there a gap in the first place? Because the presence of TIM allows the IHS to move independently of the substrate the die is bonded to, and Intel cannot design a CPU that will crush the die if the user does not take adequate care in installing it (too much pressure in one corner, IHS deforms and concentrates load on a single point on the die, die corner is crushed). When the die is mechanically bonded to the IHS, load is distributed evenly.

 

But surely you eliminate that gap when you solder, right? so how come Intel (AFAIK) still solder the Broadwell E cores? Solder (especially modern tin solder) is pretty hard.

The more I think about this the more I am starting to feel that they are using that paste to deliberately keep the clocks on their CPUs lower.

 

I believe that Intel were going to use tinfoil on the IHS but some posters have used it all up for their hats

 

It's certainly not 100% because people have replaced the TIM and got better results.
And frankly the IHS should sit perfectly on the core. It's a known and and exact distance. TIM isnt polyfiller. It's meant to fill the tiny gaps only. If Intel can't make an exact fitting IHS then they need to sort out their manuf. process. That's shocking.