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

[Arachnotronic]

This is seriously bad. Very bad. I just don't know what to say about it. A few months ago I was certain Skylake-X was going to be incredible. But now, so many things have gone wrong, and these aren't little things like BIOS issues. These recent revelations actually do indicate a potential fire hazard. This is just crazy.

Seems like the board vendors messed up.

Fortunately, I'm not planning on overclocking my SKX parts, but yeah, these are great CPUs marred by some really annoying issues (package design on the CPU, mobo makers not cooling VRM properly, etc.).

 

[moonbogg]

In the EVGA forums someone mentioned a possibility that the PSU used may be the culprit regarding the cable temps going too high. Supposedly a bad PSU design where a single cable performs multiple functions or something. They mentioned a universal 9 pin design causing too much current to be pulled through that single cable. Hopefully its just specific to that PSU, but I don't see that having much to do with the bad VRM cooling though. A second round of boards can remedy the VRM temps or possibly even a recall. We'll find out soon enough. Its been a wild ride these last few months. So much drama in the CPU world.

 

[IEC]

This is seriously bad. Very bad. I just don't know what to say about it. A few months ago I was certain Skylake-X was going to be incredible. But now, so many things have gone wrong, and these aren't little things like BIOS issues. These recent revelations actually do indicate a potential fire hazard. This is just crazy.

Looks like I'm going to wait for X299v2...

 

[Topweasel]

Having some poor performance issues with NVMe drives on this x299 system. Getting half the speed I was on z170. Still working my way through the BIOS.

Are you sure you are using NVME from the CPU and not one from the chipset. The chipset uses a switch to turn 4 lanes into 24.

 

[moonbogg]

Looks like I'm going to wait for X299v2...

Seems there isn't much choice. I can't imagine building a rig and ending up with issues like this. Then you get to tear everything apart, return all the stuff etc etc.

 

[Edrick]

Are you sure you are using NVME from the CPU and not one from the chipset. The chipset uses a switch to turn 4 lanes into 24.

I know I am not using lanes from CPU (m.2 slots on MB). But I am comparing apples to apples as these drives were installed on my Z170 rig (using the same DMI bus). On the Z170 board, these drives were scoring between 2800 to 3200 MBps. Now they are scoring 1800 to 2000 MBps.

 

[Edrick]

Note that the CPU clock and NB Clock are being incorrectly reported. They are actually 4Ghz and 2.4Ghz respectively. But the numbers are not that bad otherwise.

 

[WhoBeDaPlaya]

Uh oh, looks like all/most X299 mobo VRMs are having issues, and Der8auer does not recommend getting ANY X299 mobo at this point unless you watercool the VRMs; says to wait 1-2 months for companies to get their act together?

Not sure whether "real world" MT loads like rendering / encoding would give the same issues as power virii like Prime95, but still annoying.

The only time I've run into this issue in the past was on a crappy ASRock Z77 Pro4 that would throttle when OCing a 2500K. Pointing a small fan at the VRM eliminated the throttling.

 

[moonbogg]

From overclock.net

Originally Posted by Silicon Lottery
I am having trouble with some of these X299 motherboards. I've bought a wide variety for this launch, and none of them are really handling the load of an overclocked 7900X as well as I'd expect. VRM temps through the roof and boards throttling.

This confirms that VRM issues are not limited to one person's experience. All of this is happening with a 10 core CPU. Even with the VRM cooling issues fixed by using better heatsinks, what will things look like with a 16 or 18 core CPU? If those chips are very similar to this 10 core, how much power will those pull at higher clocks? 4-500 watts for just the CPU?

 

[.vodka]

Something like that might handle the ridiculously high OC'd power needs of the 12-18C CPUs, that is, if the 8pin power input sockets on boards that have 2 of those don't melt first. >1000w PSUs are going to become relevant, again.


I hope Coffeelake doesn't run into these ridiculous issues, considering it's also been rushed a few months, too.

 

[ub4ty]

Something like that might handle the ridiculously high OC'd power needs of the 12-18C CPUs, that is, if the 8pin power input sockets on boards that have 2 of those don't melt first. >1000w PSUs are going to become relevant, again.

I hope Coffeelake doesn't run into these ridiculous issues, considering it's also been rushed a few months, too.

Classic form over function and cost cutting where it counts. Note how many older boards had heat-pipes going over the VRMs after some iterations of aesthetic VRM heatsink designs that failed. Even then, there were issues :
http://www.alphabetcityblog.com/2014/05/taming-gigabyte-ga-990fxa-ud3-voltage.html

Amazing that it's 2017 and we have RGB light shows, carbon aesthetics, and other goofy things slapped on the board to upsell it, yet these clowns can't manage to cool one of the most fundamental aspects of the board with proper heatsink/heat-pipes/air-flow and yet they're charging $300-400 for it. I see all this unecessary junk slapped on the boards and it looks a hellavuh lot like their causing heat insulation vs. actually allowing airflow to pass over their aesthetically pleasing non-thermally painted VRM shrines. This is what happens when consumer theatrics and lowest common denominator (Ohhhh flashy lights and carbon and _____ ) rule a market for too long. This platform is drawing far more wattage than previous boards and the cooling on crucial board components looks like a joke compared to the past. To be fair, the threadripper boards seem to be plagued with equal if not worse madness as there's even less space to play with and thus even smaller surface area on the VRM metal art. Mainboards used to even have heatpipes on the backside of the boards in these areas.... No more of that. Now everything has worthless tacky carbon shrouds and heat insulating aesthetics thrown atop. It's like the blinged out ricer scene of the 90s but for mainboards. All looks and no performance. With crap like this, I'm even thinking of giving 8+ core count platforms a year or so of refinement before giving them a look. What a clown show.

Just..

 

[WhoBeDaPlaya]

I miss X58 HEDT mobos. IIRC, ~$150 for a Gigabyte EX58-UD3R, and it ran an OCed i7 920 D0 @ 4GHz rock solid 24/7/365 for years.

 

[ub4ty]

I miss X58 HEDT mobos. IIRC, ~$150 for a Gigabyte EX58-UD3R, and it ran an OCed i7 920 D0 @ 4GHz rock solid 24/7/365 for years.

Dat heatpipe. Dem fins. T H I C C

 

[Edrick]

Ok, had only a few hours tonight to break in my new rig, but here are my initial impressions and test results.

Stock (4Ghz all cores) - voltages set to Auto
- Very stable, and efficient. No problem OCing RAM to 3200mhz with no required bump to IMC or NB voltages.
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.05v
Max Core Temp: 61C
Max VRM Temp: 60C
142 watts

Mild OC (4.2Ghz all cores) - voltages set to Auto
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.10v
Max Core Temp: 67C
Max VRM Temp: 64C
151 watts

Did not have time to go any higher tonight. But to be honest, I was happy with my 6600K at 4.2Ghz (which required 1.25v to achieve), so I think I may eventually land back here as my 24/7 speed.

The Good
- Stable out of the box. Much better experience than with my z170x MB on release day two years ago. It took 3 BIOS updates later to even OC my RAM to 3000mhz. With the X299, OCing RAM to 3200 was a snap (even with an immature BIOS).

- Very energy efficient at stock and low OC speeds. Less voltage and runs cooler than my 6600K at the same speeds. I think this is a large part due to 14nm+.

- Different offsets for AVX and AVX-512. This can really help achieve better OC results in my opinion.

The Bad
- Northbrige runs at 2.4Ghz regardless of CPU speed. And unlike all past platforms, this speed is not changeable in the BIOS. Maybe this is simply a Gigabyte BIOS issue, but there is no place to change that stock value. *Update - Resolved! Found where to OC the NB*

- You can not disable the AVX offset. You either have to choose a value from 1 to 31 (1 being 100mhz less than CPU frequency) or choose AUTO (which sets the offset to 300mhz). So you can never run AVX code at CPU speeds. Again, this may be fixed in later BIOS updates and not be a limitation of the platform itself.

- NVMe drives running slow. Same drives on a Z170 MB were running much faster (see my post above). Using the same test program (Aida64) and same drives, speed decreased by over 1000MBps. This may be an Aida issue or again a BIOS issue. I will try other tests tomorrow.

The Ugly
- None yet. But other people's reviews of the 7900X is concerning.

Thoughts
So far I am happy. I expected issues as an early adopter so I am not concerned with the few issues I have found. I am also not experiencing the same heat/power/vrm issues with the 7820X as others are having with the 7900X. I will do more testing tomorrow after work.

 

[jj109]

Welp, my interest for X299 has been put on ice for now. Waiting for v2 motherboards to reconsider, though at that point it might as well be "wait for Coffee Lake", "wait for Zen+", "wait for Godot".

Edrick, what's the uncore/mesh/ring frequency?

I'm wondering what's causing this 20ns spread for RAM latency Guru3D's seeing.

http://www.guru3d.com/articles_pages/msi_x299_gaming_pro_carbon_ac_review,19.html <- 64ns
http://www.guru3d.com/articles_pages/gigabyte_x299_aorus_gaming_3_motherboard_review,15.html <- 84ns

Edit:
Ah, I missed your post.

Yeah too bad you can't change the mesh frequency. It would be neat to see how the memory system performs with the mesh clocked 50% higher.

 

[multippp]

Ok, had only a few hours tonight to break in my new rig, but here are my initial impressions and test results.

Stock (4Ghz all cores) - voltages set to Auto
- Very stable, and efficient. No problem OCing RAM to 3200mhz with no required bump to IMC or NB voltages.
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.05v
Max Core Temp: 61C
Max VRM Temp: 60C
142 watts

Mild OC (4.2Ghz all cores) - voltages set to Auto
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.10v
Max Core Temp: 67C
Max VRM Temp: 64C
151 watts

Did not have time to go any higher tonight. But to be honest, I was happy with my 6600K at 4.2Ghz (which required 1.25v to achieve), so I think I may eventually land back here as my 24/7 speed.

The Good
- Stable out of the box. Much better experience than with my z170x MB on release day two years ago. It took 3 BIOS updates later to even OC my RAM to 3000mhz. With the X299, OCing RAM to 3200 was a snap (even with an immature BIOS).

- Very energy efficient at stock and low OC speeds. Less voltage and runs cooler than my 6600K at the same speeds. I think this is a large part due to 14nm+.

- Different offsets for AVX and AVX-512. This can really help achieve better OC results in my opinion.

The Bad
- Northbrige runs at 2.4Ghz regardless of CPU speed. And unlike all past platforms, this speed is not changeable in the BIOS. Maybe this is simply a Gigabyte BIOS issue, but there is no place to change that stock value.

- You can not disable the AVX offset. You either have to choose a value from 1 to 31 (1 being 100mhz less than CPU frequency) or choose AUTO (which sets the offset to 300mhz). So you can never run AVX code at CPU speeds. Again, this may be fixed in later BIOS updates and not be a limitation of the platform itself.

- NVMe drives running slow. Same drives on a Z170 MB were running much faster (see my post above). Using the same test program (Aida64) and same drives, speed decreased by over 1000MBps. This may be an Aida issue or again a BIOS issue. I will try other tests tomorrow.

The Ugly
- None yet. But other people's reviews of the 7900X is concerning.

Thoughts
So far I am happy. I expected issues as an early adopter so I am not concerned with the few issues I have found. I am also not experiencing the same heat/power/vrm issues with the 7820X as others are having with the 7900X. I will do more testing tomorrow after work.

Good information, thank you. Is power consumption the overall system power consumption?

 

[AdamK47]

I changed vote from a Yeah to a solid Nah.

Intel had to rush things and it bit them.

 

[Dave3000]

I right now have an i7-4930k but I think later next month I might upgrade to an i7-7820x. Does the 7820x throttle to base clocks when it runs the latest Prime 95?

 

[wildhorse2k]

Ok, had only a few hours tonight to break in my new rig, but here are my initial impressions and test results.

Stock (4Ghz all cores) - voltages set to Auto
- Very stable, and efficient. No problem OCing RAM to 3200mhz with no required bump to IMC or NB voltages.
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.05v
Max Core Temp: 61C
Max VRM Temp: 60C
142 watts

Mild OC (4.2Ghz all cores) - voltages set to Auto
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.10v
Max Core Temp: 67C
Max VRM Temp: 64C
151 watts

Did not have time to go any higher tonight. But to be honest, I was happy with my 6600K at 4.2Ghz (which required 1.25v to achieve), so I think I may eventually land back here as my 24/7 speed.

Please test read/write memory speed using AIDA64. In reviews Skylake-X appeared to have a bottleneck in write speed, it was much lower than read speed and not increasing with better RAM speed.

Uncore frequency should be configurable on better boards.

 

[Wyrm]

I'm bailing on it as well. Thanks to Der8auer for the no-BS review of motherboards and CPU. All these companies need to get their Scheiße in order if they want to sell it as HEDT hardware. They spend time on RGB crap and can't even attach a proper heatsink to their VRM.

 

[moonbogg]

Ok, had only a few hours tonight to break in my new rig, but here are my initial impressions and test results.

Stock (4Ghz all cores) - voltages set to Auto
- Very stable, and efficient. No problem OCing RAM to 3200mhz with no required bump to IMC or NB voltages.
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.05v
Max Core Temp: 61C
Max VRM Temp: 60C
142 watts

Mild OC (4.2Ghz all cores) - voltages set to Auto
- Ran Prime95 (FMA-AVX2) for 10 minutes using option that generates the most heat. Results are as follows:
VCore: 1.10v
Max Core Temp: 67C
Max VRM Temp: 64C
151 watts

Did not have time to go any higher tonight. But to be honest, I was happy with my 6600K at 4.2Ghz (which required 1.25v to achieve), so I think I may eventually land back here as my 24/7 speed.

The Good
- Stable out of the box. Much better experience than with my z170x MB on release day two years ago. It took 3 BIOS updates later to even OC my RAM to 3000mhz. With the X299, OCing RAM to 3200 was a snap (even with an immature BIOS).

- Very energy efficient at stock and low OC speeds. Less voltage and runs cooler than my 6600K at the same speeds. I think this is a large part due to 14nm+.

- Different offsets for AVX and AVX-512. This can really help achieve better OC results in my opinion.

The Bad
- Northbrige runs at 2.4Ghz regardless of CPU speed. And unlike all past platforms, this speed is not changeable in the BIOS. Maybe this is simply a Gigabyte BIOS issue, but there is no place to change that stock value.

- You can not disable the AVX offset. You either have to choose a value from 1 to 31 (1 being 100mhz less than CPU frequency) or choose AUTO (which sets the offset to 300mhz). So you can never run AVX code at CPU speeds. Again, this may be fixed in later BIOS updates and not be a limitation of the platform itself.

- NVMe drives running slow. Same drives on a Z170 MB were running much faster (see my post above). Using the same test program (Aida64) and same drives, speed decreased by over 1000MBps. This may be an Aida issue or again a BIOS issue. I will try other tests tomorrow.

The Ugly
- None yet. But other people's reviews of the 7900X is concerning.

Thoughts
So far I am happy. I expected issues as an early adopter so I am not concerned with the few issues I have found. I am also not experiencing the same heat/power/vrm issues with the 7820X as others are having with the 7900X. I will do more testing tomorrow after work.

It looks like your rig is running great. That's excellent. Nothing worse than having to rip apart a rig right after building it. Good luck moving forward and if you decide to push the clocks to the standard 4.6ghz expected from these chips or even beyond, please do share the details.
Also, those memory speeds are insane. Way up there in the stratosphere. That's just crazy.

 

[Zucker2k]

Note that the CPU clock and NB Clock are being incorrectly reported. They are actually 4Ghz and 2.4Ghz respectively. But the numbers are not that bad otherwise.

These look great. It may go even go up if you disable HT. Could you do a no HT overclock as well and report temps and power numbers? Thanks!

 

[krumme]

Classic form over function and cost cutting where it counts. Note how many older boards had heat-pipes going over the VRMs after some iterations of aesthetic VRM heatsink designs that failed. Even then, there were issues :
http://www.alphabetcityblog.com/2014/05/taming-gigabyte-ga-990fxa-ud3-voltage.html

Amazing that it's 2017 and we have RGB light shows, carbon aesthetics, and other goofy things slapped on the board to upsell it, yet these clowns can't manage to cool one of the most fundamental aspects of the board with proper heatsink/heat-pipes/air-flow and yet they're charging $300-400 for it. I see all this unecessary junk slapped on the boards and it looks a hellavuh lot like their causing heat insulation vs. actually allowing airflow to pass over their aesthetically pleasing non-thermally painted VRM shrines. This is what happens when consumer theatrics and lowest common denominator (Ohhhh flashy lights and carbon and _____ ) rule a market for too long. This platform is drawing far more wattage than previous boards and the cooling on crucial board components looks like a joke compared to the past. To be fair, the threadripper boards seem to be plagued with equal if not worse madness as there's even less space to play with and thus even smaller surface area on the VRM metal art. Mainboards used to even have heatpipes on the backside of the boards in these areas.... No more of that. Now everything has worthless tacky carbon shrouds and heat insulating aesthetics thrown atop. It's like the blinged out ricer scene of the 90s but for mainboards. All looks and no performance. With crap like this, I'm even thinking of giving 8+ core count platforms a year or so of refinement before giving them a look. What a clown show.

Just..

They know how to design vrm with cooling for a cpu. Some of the boards is really nice stuff eg the highend gigabyte stuff with ess9018 dac. Damn fine. And costly to make so why skimp on cooling?

Its imo just aparent Intel raised freq above what the mb manufacturers expected or they didnt get proper power numbers from Intel. Either way its using more power than they expected.

 

[tamz_msc]

Note that the CPU clock and NB Clock are being incorrectly reported. They are actually 4Ghz and 2.4Ghz respectively. But the numbers are not that bad otherwise.

Something is definitely not right with the L3 numbers.

 

[jj109]

According to jonnyguru, if der8auer used the SuperFlower modular PSU cables he showed in the video, it might just be terrible cable design causing the 12V rail to drop too far under load.

PSU problem causing this uproar would be hilarious.