Preview: Intel Core i9 Skylake-X processors

No, that's not how it works. You mount the drives via network, the SAN/NAS is located in a centralized server room. For most workstations, Gigabit performance is all that is needed. For higher performance, you just need 10GbE add in cards, which are relatively cheap (<$300) and just need 4-8 PCIe lanes for 2 ports.

This means you still have more than enough lanes for a dGPU, if required for the specific workstation.

Unless you are running a rendering farm, mining operation, or have a very unique usage scenario needing lots of AICs, you generally don't need that many PCIe lanes. Yes it looks annoying on paper, but even for most enthusiasts, the difference is actually quite trivial.

 

If a centralized server [room] is used, what's the point of getting an i9 rig? If you need the raw processing power, it's best to have a dedicated server for that. Kind of the point of having something like an i9 or TR build is being able to do serious production work without needing a server to offload the workload to.

Think of it like this:
Let's say you're a CAD designer for very high-tech machinery. Generally speaking, you buy more processing power in order to get things done quicker. Whether you have 4 cores or 20, if all your threads are maxed out while you're rendering, you (as the user) are not really going to be able to do anything else (at least not without slowing something else down). That's why so many high-end applications allow you to offload the work to a separate server. That way, you get the sheer processing power of the server without slowing down your workflow.

To me, no workstation ever needs more than 12 threads. If your work is serious to warrant more, you should have a dedicated server to process it.

 

For most companies, this is correct. The PCs are generally not very powerful, or so called "thin-clients".

Not everything can be migrated to a centralized server though, especially since it's generally pretty costly too. Some pro programs also require a powerful front-end prior to the rendering (imagine operating on a CAD file with 1M+ parts). Even if you do not utilize distributed computing, you'll still find iSCSI and centralized storage used. This is because this can safeguard everyone's data while providing good enough performance (or great performance if you upgrade your network infrastructure). An iSCSI storage unit mounted on a PC behaves as if it is just another partition. Your IT can however, make sure it is properly backed up.

In this usage scenario, you can for example, keep all of your work files, such as... CAD, images, etc. all on this "partition" and remount it on another computer in the event the original is trashed. My personal usage is like this, we have workstations and all flash equipped SANs serving a chunk of storage to each of them. There is also a central NAS for files that should be shared.

My point here was just to point out that hardware RAID is largely not utilized in any SMBs and up. For NAS, even the DIY crowd has largely moved to software based RAID that is in many ways, much more resilient and tested vs built-in "hybrid" RAID systems integrated on chipsets.

 

I know that my next PC related item is going to be a something Ryzen related. Its probably going to be a laptop if there is a ryzen based laptop with an Nvidia graphics card. Right now as far as desktops go I have no reason to upgrade any time soon. Laptop related I have a reason because of it having a workstation Quadro Graphics card rather than having a gaming card.

Also I really see no reason for any gamer purchasing a CPU with this many cores and threads. To me this would be used for a Virtualization style machine or one of them scientific PCs that do those advanced things.

 

I'd like to add to that:

TR: no quad channel (CPU possibly will be strangled because of it); = 2 1800X on a die so not a true native 16C (= likely high latencies); very likely very low frequencies. Sounds awesome! True winner. I know, best price/performance blah blah and that's all AMD usually has, because they surely have always lacked the raw power and pure performance the therefore more expensive Intel has. Power and performance play my games and run my apps, pennies saved do not.

To quote someone else: you know Skylake-X is possibly very good when you have the entire AMD army in this thread desperately trying to downplay it.

Enough messing around; I truly hope TR will kick Intel between the legs but from the little we know I remain skeptical for now.

 

Wow, wrong right off the bat. FYI, Threadripper will be quad-channel for all SKUs.

I think you may be thinking of Kaby Lake-X, which will be limited to dual-channel on X299 boards (note that you will still be paying for quad-channel, it just won't be enabled for KB-X chips).

X399 will also have 64 PCIe lanes for all SKUs, instead of the 44, 28 or 16 lanes for X299 (talk about strangled, LOL).

 

Uh... TR does support quad channel, and there are many sources to back that up (guru3d censors links). If you haven't noticed, i9's frequencies are pretty low once you get to the same core counts as TR. Besides, as has been established several times in this thread, you're not going to want to get beyond 4GHz (on either TR or S-X) because thermals are going to be unmanageable.
As for gaming... really? Not only are neither of these platforms for gamers, but do you not understand how games work? Adding more cores is not going to make your games run faster. 6-core i7s have existed for years, and yet many hardcore gamers (where expense isn't an issue) have never bought one - because they never needed to. You're delusional if you think an i9 setup is going to definitively give you a better gaming experience than an overclocked 7700K.

There's nothing to downplay - it is objectively downplayed already. There are 6-10 core i7s that have more PCIe lanes. There are Xeons slightly more expensive that offer more cores, PCIe lanes, and RAID features than Skylake-X. Most of the people "downplaying" S-X here are currently rocking Intel rigs. Is there something I'm missing here, or are you just trolling?

No, you don't. If you actually hoped for the success of TR you wouldn't make statements like "Power and performance play my games" when you own an i5. Kind of hypocritical, eh?

 

Yeah I thought it was 100% confirmed TR is quad.. pretty sure all the motherboards at Computex all said quad memory support.

 

Actually they are saying 4.3Ghz on SLx 10core (this is likely heat limited with the TIM). As far as I know there have been no 4.3Ghz R7's (I have seen a few 4.14-4.2Ghz R7's though).

I will be willing to bet 4.5Ghz is fairly common on a de-lidded SLx and 4.7Ghz possible.

 

I'm sure 4.3Ghz on a 10-core will probably be the most common high-speed overclock for i9s, but I suspect most people will be limited to around 4GHz. I think 4.5 could be achieved with some real beefy cooling, and anything greater than that will be very rare; it'd probably be cheaper to buy more cores than to cool something that fast. But as you get more cores, your realistic maximum speed drops. Once to get to 14 cores, I don't think any i9 users are going to want to go above 4.1Ghz.

If getting high frequencies is an objective, neither i9 nor TR are good choices. The point of these is to have many fast (but not "very fast") cores in a single PC for a modest price. If strictly high-speed single-threaded performance is your goal, i7 is really the only option. If you just want a crapload of cores with no expense to spare, Xeon is the only option. This is where TR becomes a very happy medium.

 

Motherboard manufacturers said 4.3 was the magic number on 7920x that's the 10 core. This is without de-lidding and the CPUz screen I saw showed 1.256v. That tells me with a de-lid 200Mhz is almost a giving.