Skylake-X, Kaby-X, and X299

[Charles Wirth]

http://x299.weebly.com/

Formatted Review

[Buckeye]

Sorry for no screens, but trust me they are awesome screens, really nice, I am told they are really nice too, you gotta see them.

I bet they are yugggeee !!!!

[Charles Wirth]

After working with the Asus and the MSI boards trying to figure out the best raid configurations it has been mostly fail. With the MSI board, I was happy that it had a U.2 port as I can connect my 750 SSD but I want the boot and run my RIP software the Optane drives in RAID 0. The M.2_2 on this board shares bandwidth with the U.2 port totally killing an ideas of using both together. When you run this configuration performance is reduced dramatically and you never get that performance back even if you remove the U.2 drive and reload drivers.

Switching over the the Asus Tuf X299, the M.2_2 seems to be shared with the lowest PCIe slot and the performance was exact to the MSI, I moved the 750 SSD into the 16X PCIe slot and performance was reporting correctly and this configuration worked.

I am testing with 15.5 and 15.7 f6 drivers, the newest Intel NVMe f6 driver and the corresponding exe's after OS install.

Some of the issues I had on the MSI were due to the Optane Genie, this enables RAID and RAID must be enabled to setup the single Optane drive as a cache drive. Optane Genie would fail after a reboot and a menu would appear "press F1 setup or press F2 defaults". Going into BIOS you will find that it disabled RAID. Pressing defaults it will try to boot and fail, you will need to go back to BIOS and enable RAID. This may be fixed in future BIOS.

I also tried to setup the Optane as a cache using the 15.7 driver but I kept doing the full install, run the Optane exe and have it error out with the message "Boot drive not unsupported" when using a U.2 750 SSD, PCIe AIC 750 SSD, or a M.2 960 EVO. Another reviewer suggested I try again with the 15.5 drivers as Intel may only want this used with a drive connected to the SATA port.

I hope and expect a lot of the issues I ran into will be worked out in BIOS or drivers soon, we are upgrading our RIP servers the the X299 platform.

The reason I/we need the fastest drives possible and you see me with some silly setups is that time is money in the print industry. I work for Advanced Technologies, we own Pictographics. Picto is one of the largest grand format printing companies in the world based in Las Vegas and in business 20+ years with clients like Intel, Disney, Samsung, LoL, NFL, Nascar, and so much more. When you see a huge print on the side of a building, a double decker city buss wrapped, the monster US flag being spread out on the infield of a pro bowl, the entire stadium and stage at League of Legends, and when you are watching Taylor Swift singing at a I Heart Radio show, everything in the stadium and what you see is done by us. We do a lot of booth stuff as we have conventions here every week of the year.

The images for these massive prints are scaled down and manageable on workstations but when it comes time to print one of the files on the grand format printers, the file needs to be scaled up many times and this is called RIPPING, Raster Image Processing. The RIP server will sample every pixel in the image and divide it based on the adjacent pixel, the algorithm changes based on how much you are scaling up and the number of adjacent pixels it needs to sample to make new pixel in the process of scaling an image up for print. This has to be done for every print and each printer has a RIP server feeding it. The RIP can scale up a 10MB PDF into a 100GB+ low raster image to be printed, during this process it is extremely processor and drive intensive with the drive being the bottleneck in the ripping process.

Now that I spent many hours on these boards using an AIO cooler and the cascade I can touch on the VRM and the expectations on the 7900X. I am using a Thermaltake Water 3.0 Ultimate AIO cooler with two Enermax Twister fans, my chip is not delided and I idle around 36C at 4.5Ghz SVID and full loaded I did not hit 90C in XTU. I could not complete XTU at 4.7Ghz ~1.35v but everything else ran fine. Due the the placement of the fans and radiator it blows air onto the VRM, no extra fans added. Using the Hall of Fame XMP profile I had no problem running 4Ghz in quad channel and improving upon that was easy. Hitting a 100K on Aida64 has never been easier. If you plan on going with the 7900X I highly suggest getting a premium AIO cooler for the best ambient results.

Running under that cascade I hit the cold bug at -114C on my chip so I needed to boot in at 5.5Ghz 1.3v minimum to prevent CB at idle and I could run everything fine, kicking up to 5.7Ghz with 1.35v the cascade could not hold the load but everything else ran fine and 5.8Ghz seems to be the limit for my chip, I could boot in but stability was an issue. Something to watch out for is the Asus BIOS defaulted the cache speed to 1Ghz, manually kicking to 2Ghz is needed and higher for overclocking. On the AIO cooler going over 2.8Ghz became unstable so I ran at 2.5Ghz with auto voltage. Over the years SVID has gotten better, most people blindly trust auto values to work perfect and others hard set the values. For me, I rely on SVID until I find the edge and adjust accordingly and for AIO cooling the only values I might touch are CPU and RAM multipliers, timing and vmem.

I will get back to the 7740K soon, but first I wanted to get some FLIR images of the VRM under stress, my FLIRone has the newest firmware for highest accuracy of temp sensor. My next update will have screens and images.

[JPQY]

Is it possible for you to run also a chess-bench on it, Thanks..

JP.

[Revv23]

Fantastic! Thanks for the info! Who would have thought, airflow cools the VRM!

[Charles Wirth]

I can run chess on it

[Charles Wirth]

I forgot to run chess but I will get it.

Updated first post.

[JPQY]

No problem..their are already a few people who tell me they will run Chess-bench..so i'm looking forward to see & get them !

JP.

[Buckeye]

I used to do Flexo Printing back in the day. Most of what we did was simple labels.
But also things like when you get a 12 pack of beer, the cardboard box it comes in. These were huge plates, but simple step and repeat of the same image. Often these would be 4-5 images wide. The printing plate was etched in copper on a large cylinder. Big enough that they had to use a crane to lift the die's on to a press for printing. They would print hundreds a minute in 4 color process.

We also had RIP software for what ever printer type were used. Its all pretty neat stuff really. Images were worked with were generally 133 lines per inch resolution and kept on a RIP server. We would use a OPI (Open Prepress Interface), low res image for working with and then the RIP would swap out the real image for printing. We often used Quark-Express for printing as it ripped much faster than other apps, Adobe Illustrator for layout and building the art-work... import into Quark to print.

My daughter still does this, she is much more advanced that what I did back in the day. So much new stuff compared to how I did it back then.
We did a lot of stuff on art tables before Mac's with Adobe Illustrator came along. So much dark room stuff, step and repeat cameras, striping negatives together. That was real fun with 4 color process stuff. I was hired at a company called MEPCO to get the new computer setups working, man back in the 80's I think it was. Two people in the art department were let go not long after... a type setter and camera operator.

Of course Charles stuff is much bigger, would love to see that setup some day.