Who's buying Skylake-X? (You may now change your vote)

[wildhorse2k]

To those considering X299, X370 or X399 there may be one important advantage of X299 over AMD: wide range of CPUs available, from low cost to very high cost. You get one platform that supports multiple M.2, lot of high speed RAM, quad memory channel, lot of PCIe lanes. In the beginning due to limited budget you get a "cheap" 6-8 core but later if really needed (games really start using multiple cores or need more for work) you can swap in a 10-18 core and unlock its full potential.

With AMD you either have:

1. X370 - max 8 core, although cheap, very limited in M.2, RAM, PCIe lanes
2. X399 - the lowest will probably be 12 core, best 16 core. Lot of M.2, RAM, PCIe lanes, but also high cost. Perhaps you don't need all that at the beginning!

I think it would have been better if AMD had just one platform like Intel instead of 2.

 

[scannall]

I think it would have been better if AMD had just one platform like Intel instead of 2.

Um, Intel has 2 consumer facing platforms. And will continue to. The 270 and (soon to be) 299.

 

[MrTeal]

I'm actually shocked he was able to delid them without melting the solder. I would have thought you would crack the die.

 

[wildhorse2k]

Um, Intel has 2 consumer facing platforms. And will continue to. The 270 and (soon to be) 299.

I wasn't counting total number of customer facing platforms. Only pointed out that Intel covers 2 AMD platforms with 1. After arrival of Ryzen the Z270 is irrelevant, except perhaps for budget boards + CPUs.

 

[TheF34RChannel]

I wasn't counting total number of customer facing platforms. Only pointed out that Intel covers 2 AMD platforms with 1.

Exactly, or, the other way around; Intel offers a wide range of CPUs (bottom entry models to beasts) whereas AMD only offers a small range (beasts).

 

[Yeroon]

I wasn't counting total number of customer facing platforms. Only pointed out that Intel covers 2 AMD platforms with 1.

Thats not necessarily a good thing. You're also missing the fact that theres some serious segmentation from top to bottom with the x299 platform when saying lots of pcie, quad channel, etc. As well, running the low end x299 chips still require an expensive mb, as they all need to meet specs to support the 18c model.
http://www.anandtech.com/show/11461...new-highend-desktop-platform-and-x299-chipset

In reference to the pice/memory channel differences:

This is arguably a big mess. In order to compensate, boards might not implement the high-end configurations, meaning that a top of the line processor might not use all the features it comes with.

 

[Wilfred86]

I have to correct you here somewhat: people assume that it was a cost saving decision. Yet people simply don't know.

Yeah ok, it was an assumption of mine, but what else could it have been? It is the easiest explanation.

There is a good chance that at the packaging stage Intel doesn't even know if it's an "extreme" chip or not. Enthusiast SKUs only exist because it's trivial to relabel a regular SKU and overclock it.

You mean those chips are just repurposed Xeons that were at the last minute relegated to the consumer realm? By the way, are Broadwell-Xeons also soldered like their Extreme-series counterparts?

 

[TheF34RChannel]

This is arguably a big mess. In order to compensate, boards might not implement the high-end configurations, meaning that a top of the line processor might not use all the features it comes with.

I have a different view on that; boards may be aimed at certain segments of the lineup with the higher end boards packed with all features.

 

[LTC8K6]

I'm actually shocked he was able to delid them without melting the solder. I would have thought you would crack the die.

I was expecting some heat to soften the solder.

 

[ddogg]

To those considering X299, X370 or X399 there may be one important advantage of X299 over AMD: wide range of CPUs available, from low cost to very high cost. You get one platform that supports multiple M.2, lot of high speed RAM, quad memory channel, lot of PCIe lanes. In the beginning due to limited budget you get a "cheap" 6-8 core but later if really needed (games really start using multiple cores or need more for work) you can swap in a 10-18 core and unlock its full potential.

With AMD you either have:

1. X370 - max 8 core, although cheap, very limited in M.2, RAM, PCIe lanes
2. X399 - the lowest will probably be 12 core, best 16 core. Lot of M.2, RAM, PCIe lanes, but also high cost. Perhaps you don't need all that at the beginning!

I think it would have been better if AMD had just one platform like Intel instead of 2.

Those considering thread-ripper like myself are generally the enthusiast crowd and are willing to shell the money to get it. If someone was more budget minded and still wanted to get an AMD CPU they would have a 1800 or lower.

Sent from my Nexus 6P using Tapatalk

 

[Rifter]

Those considering thread-ripper like myself are generally the enthusiast crowd and are willing to shell the money to get it. If someone was more budget minded and still wanted to get an AMD CPU they would have a 1800 or lower.

Sent from my Nexus 6P using Tapatalk

I agree. Much better to have a mainstream and a HEDT platform so the mainstream buyers dont need to carry the cost of the HEDT mobos for features they will never use.

 

[Zucker2k]

I'm actually shocked he was able to delid them without melting the solder. I would have thought you would crack the die.

I think it offers a clue as to how well the solder bonds with the die. 10 years ago, this method would've certainly ripped the die from it's roots.

 

[MrTeal]

I think it offers a clue as to how well the solder bonds with the die. 10 years ago, this method would've certainly ripped the die from it's roots.

Sounds like an excuse to grab some P4s and do a little testing.

 

[scannall]

I think it offers a clue as to how well the solder bonds with the die. 10 years ago, this method would've certainly ripped the die from it's roots.

That's pretty much what happened when this guy tried to delid a Ryzen. After a couple tries he got one delidded without destroying the chip. After all that, couldn't really improve the heat dissapation, since it's soldered really well from the factory. Ryzen delid

 

[MrTeal]

That's pretty much what happened when this guy tried to delid a Ryzen. After a couple tries he got one delidded without destroying the chip. After all that, couldn't really improve the heat dissapation, since it's soldered really well from the factory. Ryzen delid

That's pretty much how I would have approached delidding the soldered processors.

I have to say though watching that guy use box knife blades and razor blades in that way makes my skin crawl.

 

[idGhost]

I would like to get the i7-7820X because it's ~$400 less than year, but it's still a lot of money for last gen tech. I was hoping Kabylake-x processors would've had more core options. I wonder how fast 6-core Coffeelake, (or was it cannonlake?) cpu would be.

 

[Yeroon]

I have a different view on that; boards may be aimed at certain segments of the lineup with the higher end boards packed with all features.

So having a feature segments among the same chipset & socket kinda defeats the point though? Like, theres a reason theres like 3 different chipsets for mainstream sockets, and none of those sockets even have things like dual/quad mem channel to worry about.

 

[TheF34RChannel]

I would like to get the i7-7820X because it's ~$400 less than year, but it's still a lot of money for last gen tech. I was hoping Kabylake-x processors would've had more core options. I wonder how fast 6-core Coffeelake, (or was it cannonlake?) cpu would be.

HEDT always lags behind so it's not really last gen. There's two platforms with their own cycle and release windows. And yeah, having both last 2 gens is an odd choice.

 

[LTC8K6]

I would like to get the i7-7820X because it's ~$400 less than year, but it's still a lot of money for last gen tech. I was hoping Kabylake-x processors would've had more core options. I wonder how fast 6-core Coffeelake, (or was it cannonlake?) cpu would be.

If the 7820X is on the 14nm+ process, then I guess it's actually Kaby Lake and not Skylake.

 

[wildhorse2k]

If the 7820X is on the 14nm+ process, then I guess it's actually Kaby Lake and not Skylake.

Manufacturing process does not equal CPU architecture. You could produce Intel 386DX on 14m+ if you wanted and that wouldnt make it Kaby Lake.

 

[Edrick]

Manufacturing process does not equal CPU architecture. You could produce Intel 386DX on 14m+ if you wanted and that wouldnt make it Kaby Lake.

When the process is the only thing that is different between skylake and kabylake, then LTC8K6 is correct.

 

[wildhorse2k]

When the process is the only thing that is different between skylake and kabylake, then LTC8K6 is correct.

Why are you convinced Skylake-X and Skylake/Kabylake have the same architecture?

 

[Edrick]

Why are you convinced Skylake-X and Skylake/Kabylake have the same architecture?

Wow, did you just pull that out of thin air? I said in my 1 line post that skylake and kabylake are the same architecture sans the process node (14nm vs 14+nm). Nowhere did I say anything about comparing Skylake-X and consumer skylake/kabylake.

In fact, just the cache structure changes coupled with the AVX-512 instructions, would suggest that Skylake-X is a different architecture over the consumer line.

So, if Skylake-X is not on 14nm process, then it does share a process with kabylake (14+nm).

 

[wildhorse2k]

Wow, did you just pull that out of thin air? I said in my 1 line post that skylake and kabylake are the same architecture sans the process node (14nm vs 14+nm). Nowhere did I say anything about comparing Skylake-X and consumer skylake/kabylake.

So, if Skylake-X is not on 14nm process, then it does share a process with kabylake.

Ok so lets recapitulate:

Original statement that I disputed:

If the 7820X is on the 14nm+ process, then I guess it's actually Kaby Lake and not Skylake.

The 7820X is Skylake-X.

You procceeded to claim LTC8K6 is correct. If you agree with his statement then you agree that Skylake-X is Kaby Lake.

 

[Edrick]

If Skylake-X is in fact on 14nm+, then has more in common with KabyLake-X than it does with Skylake-S. Not sure how you can dispute that.

But I never said Skylake-X has the same architecture as the consumer class skylake/kabylake.

You are simply mixing terminology to suit your argument.