X299 motherboards

[mackerel]

The same as on previous X platforms, cache frequency affects maximum memory bandwidth. It doesn't really matter if in use is DDR4-3000 or 3600 memory, bandwidth without cache overclocking is about the same. My CPU couldn't make more than 3300MHz on auto voltages. I will check later what affects cache.

Did you look further into cache? I'm about to see how far I can push mine as there are reports suggesting Skylake-X isn't scaling for high FPS gaming when overclocked, and I have to wonder if the cache is limiting it. I haven't looked at bios options, but I do note XTU allows you to set a fixed or offset cache voltage. It doesn't actually show what the stock voltage is though so maybe offset is safer... still, I think even if I go to ball park 3300 like you, that's still a significant boost over stock 2000. Let's see what bandwidth I can get

 

[Woomack]

After yesterday's tests I can say that there is bigger difference between memory frequency than I thought. Motherboard is sometimes acting weird when there are auto, C states and ASUS additional options enabled. But still cache affects maximum memory bandwidth while memory timings affect memory latency more than the bandwidth.

Maximum stable cache on my setup is 3200MHz. I can set it at auto voltages. 3300MHz is passing most tests but I have a feeling it's not fully stable. 3400MHz is sometimes booting and sometimes not. I'm not sure why but cache voltage is not helping at all in cache overclocking. The same VCCIO and SA at auto and at manual settings are not showing any improvements in overclocking. Or it's not working or something else is required at the same time.

 

[mackerel]

Motherboard is sometimes acting weird when there are auto, C states and ASUS additional options enabled.

This.

I haven't managed to put my finger on it, but I get really bad Prime95 throughput benchmark results when changing some setting. It happened yesterday when I first turned on XMP, and it happened again just now when I turned it off again to get some baseline data. I don't know how I fixed it yesterday, I think it was through XTU cache overclocking and undoing, but that didn't work today. Today a cold boot seemed to help - a warm boot didn't reset it yesterday. I suspect something isn't getting initialised correctly without some help.

On my 7800X sample, 3000 cache seems stable with 1 hour P95 blend (with CPU OC and 3200 ram). 3100 crashes on load (once BSOD, once locked up). Haven't tried playing with voltages on it yet.

 

[Woomack]

4-6 cores have one stage of turbo boost less than 8c+. I don't know if it affects performance and stability but you see about the same weird behaviour on 6 cores as I see on 10. It shouldn't really matter but hard to say if all is working correctly. I've noticed that now cache and memory clock go down when there is low PC load. Previous series were focusing mainly on CPU.

Btw. overclocked 7900X +1080Ti under load = 620W

 

[mackerel]

When I OC I tend to fix all cores to same clock, not worry about turbo-ing fewer. Even on Skylake-S I think I've seen cache downclock on idle for power saving, but not ram. I haven't seen memory clock change on Skylake-X.

Not looked at power consumption at all yet. Still think the 1000W was overkill but it does mean I never have to worry about power on that system. Also I'm kicking myself a little now... I found my dual Xeon server has a PSU with two EPS connectors I could have recycled in short term. My outstanding credit card bill is looking rather fat at the moment...

Right now I've set my 7800X fixed at 4.3 GHz, ram fixed at 3200, cache fixed at 3000, and it worked fine first time without anything funny going on. I do suspect the funny low results are due to some power saving maybe being too aggressive. I want to try some real multi-thread AVX2 workload on it later. On paper, it should still be slower than my 14 core Haswell Xeon, but it should be close and might even catch up due to inefficient scaling at higher thread counts.

 

[Woomack]

Intel's biggest fail on this platform ... VROC.
Yesterday I was trying to install RAID on CPU using 2x PCIE SSD in PCIE x16 slots. I found out it's impossible to make it work without ... hardware key which is unlocking RAID or dedicated Intel SSD Of course on the website is a lot of info how great it works, Intel is saying a lot of things about huge performance gain and how X299 is amazing ... but info how to make it work is in really small font in the motherboard's manual.
So what Intel did ... gave us VROC hardware key ... which is not available anywhere and of course have to buy it separately ... or forces us to buy their SSD to be able to set RAID0, for RAID1 is required key, for RAID5 is required even more expensive key ( of course not available )

Setting RAID is possible without a key but via Windows dynamic volumes. It's not possible to have booting PCIE SSD in RAID without the key or mentioned dedicated Intel SSD.
I was able to set 2x PCIE SSD ( Alphacool controller/cooler + Patriot Hellfire M.2 NVMe SSD ) as stripped dynamic volume on which it runs pretty good. Comparing to RAID on M.2 sockets I was able to achieve stable ~5GB/s sequential read with spikes up to 5.4GB/s while before, maximum was ~3.5GB/s with spikes up to 3.9GB/s ( so about max DMI ).

 

[EarthDog]

I don't think many care about VROC... kind of tempers the 'fail' a bit. Otherwise, I agree.

 

[Woomack]

Maybe it's more like a disappointment when you buy high end platform with features advertised as one of the best reasons why to move to x299 and later you get a slap in the face when you find out that you have to spend next $100+ to make something quite simple to work because Intel is just greedy ... not to mention it's not even available in stores.
I agree that not so many care and most don't even know what is VROC but while googling VROC for some answers I almost only found angry users.

 

[mackerel]

Having to unlock VROC and its Intel SSD requirement was one of the points negatively viewed by many when it was announced, but it is pretty much a niche. For performance terms, how many are limited by a single high end NVMe drive anyway? IOPS wont improve from raid, do you need even more sequential? It could be handy if you really need one big drive but we're getting really niche here...

 

[Woomack]

If you are using it for larger projects with rendering etc or databases then you may want to use RAID1 or 10 on SSD. Then VROC can be handy. This whole platform is niche but many of those who actually decide to buy it, think about something more than only gaming.
IOPS go up in RAID but random bandwidth not so much and it's the most important during work. High queue calculations are going up but it's related almost only with servers and databases. So of course in typical home/office software you won't see any difference. Actually there is barely any difference between SATA and PCIE SSD as what counts most are random transfers and these are not much worse on higher series SATA than PCIE SSD.
However X series chipsets already cost premium and it's the main reason why these motherboards are so expensive so paying additionally for unlocking functionality is a bit too much. I could understand if that key would add whole controller ... but it's only unlocking option in board's BIOS and expected price is at least like 1/4 of this board ( hard to say what will be in real ).

btw. there is new BIOS for TUF Mark 2 http://dlcdnet.asus.com/pub/ASUS/mb...25.759329787.1501262947-1017150005.1499368898
I had no time to check it yet. I'm installing Win7 since yesterday ... constant issues with some updates, USB devices or combination of AHCI and NVMe drivers.

 

[mackerel]

I may be making assumptions, but logically raid including 0 or 5 are the ones that could benefit from the connection to CPU. If raid 1, bandwidth is less of an issue. Still, I struggle to see how random performance could be improved through raid 0 or 5. Only raid 1 possibly. Maybe 10 but how many would run that?

Will maybe check out the bios later. Fooding and gaming time tonight, will give tinkering a bit of a break.

 

[Woomack]

New BIOS seems worse than the old one. I see no differences except that memory has problems to boot at 3600 now and it was fully stable on older BIOS. I also can't back to older BIOS, flasher says that file is corrupted.

 

[mackerel]

I haven't updated yet. If there's a way to backup the old one I could try, but I can't make any guarantees of course.

 

[Woomack]

Looks like it's protected against flashing older version and there is no flash recovery option like on more expensive boards. One file was corrupted but I was trying with other flash drive and once more downloaded BIOS and I got info that file isn't correct BIOS file.

 

[wingman99]

Looks like it's protected against flashing older version and there is no flash recovery option like on more expensive boards. One file was corrupted but I was trying with other flash drive and once more downloaded BIOS and I got info that file isn't correct BIOS file.

What are you going to do?

 

[Woomack]

I'm trying to figure out what is affecting worse memory OC on new BIOS and I think I found it. So far I made it work at 3600 and tighter timings but to make it work at CR1 and/or tighter timings, I have to set higher IO/SA when on older BIOS it could stay at auto. Also when XMP or manual option is enabled then memory can't boot at 3600. I have to leave it at auto and set everything else manually. On older BIOS it was working on all 3 settings.

 

[wingman99]

Well at least you have it working at 3600 with tighter timings. What did you set the Vccio and Vccsa at?

 

[kitchu]

Hi, I need help to boot my newly bought ASUS X299 E motherboard with 7740 processor & Corsair DDR4 ECC 32 GB RAM with ASUS Graphics card GEFORCE 1080. While booting, nothing appears in my 4 k monitor & the code display stops after A0. Can you pls help me to go to the boot screen to activate use of M.2. I have not installed any HDD/SSD on the board other than M.2 slot.

 

[mackerel]

7740 doesn't support ECC. Dunno if it will run with ECC in non-ECC mode or refuse. Having said that, a quick search suggests that A0 is relating to IDE... if you have non-ECC DDR4 ram you can borrow from elsewhere I'd still try that. Otherwise try a bios reset.

 

[wingman99]

Hi, I need help to boot my newly bought ASUS X299 E motherboard with 7740 processor & Corsair DDR4 ECC 32 GB RAM with ASUS Graphics card GEFORCE 1080. While booting, nothing appears in my 4 k monitor & the code display stops after A0. Can you pls help me to go to the boot screen to activate use of M.2. I have not installed any HDD/SSD on the board other than M.2 slot.

Do you have a speaker connected to the front panel so you can hear a beep when it post?