AKBrian

Some motherboards can be a little bit selective about their memory modules - typically this only becomes a problem with all four slots populated, but I've seen it happen with only two as well. If you're able to boot with each stick individually, it likely needs a little bit more juice to run all four sticks. I owned an ASRock P68 which was very particular about its DIMMs and what worked for that system was to simply bump the memory voltage one notch. To do so, boot with a single module and go into the BIOS. Tab over to the OC Tweaker menu, and scroll down to the voltage configuration. Change DRAM voltage to 1.55v or 1.6v (stock is 1.5v on those Ripjaws, but they don't mind extra voltage), save and reboot. Once it reboots successfully, you can shut the system down. Then, install the remaining DIMMs and power up. If all goes well, you should have all four sticks. If you're still unable to boot with four DIMMs, you may need further voltage or to bump the VCCIO/VCCSA voltage one notch - populating all four DIMMs puts an extra load on the memory controller and this can resolve the issue. If all four sticks are cooperating at this point, you can try lowering the voltage a little bit to reduce heat, but there's really no risk of running 1.5v or 1.6v on DDR3 modules. The Ripjaws series DDR3 memory is pretty good. I've been running 32GB (4x8GB) of the Ripjaws DDR3-1866 memory in my Z77 OC Formula for a loooong time. Mine is overclocked to 2133MHz and set at 1.65v.

You can use Xeon CPUs in that motherboard, but you won't be able to overclock them by changing the multiplier - only BCLK (there is limited ability to modify the all-core turbo setting, but it's still effectively a locked CPU series for that generation). You can use up to an i7-3770k, plus the CPU can be overclocked if you wish - make sure you have a good cooler, as well. The i7-2600k or 2700k are good alternatives. They generally overclock to a slightly higher frequency, but the 3770k is a bit faster per-clock so they tend to even out, and you may be limited in overclocking by your motherboard. If you go with a 3770k, be sure to check your BIOS version. They require version P2.10 to be recognized properly.

TL;DR: What is your goal with the system? Are you still wanting to utilize a four way NVMe M.2 PCI-E x16 card, or are you now just wanting to build a system using three separate PCI-E M.2 drives? If you've abandoned the quad NVMe M.2 card idea, and are specifically wanting to use an i7-9700K, choose a motherboard like Gigabyte's Z390 Aorus Master / Ultra or ASRock's Z390 Taichi / Taichi Master. They are solid motherboards which feature three NVMe M.2 slots. However, if you are going into this expecting to combine your drives into a RAID 0 array for high bandwidth, you'll be very disappointed. As the three M.2 slots will be partially sharing bandwidth across the chipset's DMI link, which operates at PCI Express 3.0 x4 (~4GB/s). You'll also likely lose a few SATA ports due to potential lane sharing with M.2 slots, although the specifics vary by board and for most people four SATA ports are still fine. If you are using the M.2 drives individually (eg, formatted separately and with their own access patterns) this DMI link limitation typically won't be an issue. If you're looking to combine (for example) three 3GB/s drives in RAID 0 and hope to achieve a 9GB/s read rate, though, you're not going to like the results. A ~4-4.5GB/s rate is what you're likely to see due to the DMI bottleneck between chipset and CPU. Consumer platforms simply don't offer enough CPU connected PCI Express lanes, and it is for that reason X399 (Threadripper) or X299 (Intel HEDT) platforms are essentially required. They support both the quantity of PCI Express lanes needed to achieve high transfer rates from multiple storage devices, and support the lane bifurcation necessary for four-way NVMe cards such as the Asus Hyper M.2 or ASRock's similar adapter. There are a few ways to try and sneak around this limitation, such as utilizing PCI Express adapter cards to plug an M.2 drive into a CPU-connected expansion slot. The downside here is that it will kick your GPU down to x8 mode. At least one motherboard, Gigabyte's Z390 Designare, offers a mode switch to split the CPU-connected slots into an x8/x4/x4 configuration. Some boards may have a modified BIOS available that allows slot bifurcation, which would let you use a quad M.2 card in the top x16 PCI Express slot, but would necessitate plugging your GPU into a slow (x4) chipset connected expansion slot. These aren't ideal solutions, obviously, and are more of a side effect of specific boards that happen to support odd operating configurations.

I've not personally used an X370 based motherboard, but the top two SATA ports you reference (ASATA) are designated as APU connected SATA Express lanes. There should be an option in your BIOS under the Chipset page labeled "APU SATA Port Enable" - ensure that this is set to Enabled. Then, you should see two options immediately below which read "APU SATA Port0" and "APU SATA Port1" - this should display drive information for any connected and detected devices. If you have toggled the APU SATA Port Enable option, you'll need to restart first for the devices to be listed. If you're looking at the PDF manual from Gigabyte's site, the information for this section is on page 31. Double check to ensure that the SATA Mode option is set to AHCI instead of RAID, as well. Some motherboards will limit active ports when a RAID mode is used, and on X370 it looks like RAID is only supported on the four chipset derived ports. Hopefully it turns out to be something simple like a BIOS toggle!

It should also be noted that to enable PCI-E 3.0 functionality you need to use a 3rd generation (Ivy Bridge) Core CPU, - the i5-2500 is a 2nd generation (Sandy Bridge) and will only allow PCI-E 2.0 mode. I only mention this because you listed it as a deciding factor. In reality, you're unlikely to run into the bandwidth limitation of even PCI-E 2.0 with that board and processor, and are likely to be using an equally modest GPU and standard SATA drives.

Very cool. I've always wondered what happens to the custom CPUs Intel produces for industries like High Frequency Trading, but I suspect most are simply destroyed rather than recycled and resold through liquidators like these. Looking forward to the followup, Linus!

Haha, cheers! ??

You'll love it, the 3700X is a big bump coming from an i5-3570K. If you feel the upgrade itch again next year, look into an RTX 2060 Super or Radeon HD5700XT around Black Friday. It's likely that both will be available on sale for about $300, making them a good buy. Both cards will offer approximately double the performance of your existing GTX 980, not accounting for features such as Freesync/G-Sync, screen recording, increased VRAM and such. Keep an eye out for used hardware, too. The next batch of nvidia GPUs will likely land in the spring and there are always early adopters looking to sell their old cards on classifieds. A $150 GTX 1080, for example, shouldn't be passed up.

This board has a bit of an odd party trick in that there is a physical switch that can toggle the last slot (the x4) to utilize the CPU's bandwidth rather than the PCH. In normal operation, the main x16 (physical) slots will be divided up as such: · x16 / x0 / x4 (PCH) · x8 / x8 / x4 (PCH) You can, by using the switch to route the lanes to the CPU, also choose one of these three modes for the board's main x16 (physical) slots: · x16 / x0 / x0 (CPU) · x8 / x8 / x0 (CPU) · x8 / x4 / x4 (CPU) The main difference is that in this configuration (slot #3 routed to CPU) the third slot isn't fighting the NVMe M.2 and SATA ports for bandwidth across the DMI link (from PCH to CPU). If you intend to run a third or even fourth NVMe M.2 with an x4 conversion card, this is the mode you want! By necessity, you will be required to run your GPU at x8 - the CPU only provides 16 lanes to divvy up between all three primary slots. By stealing four lanes for the third slot, the first slot will run at x8. Note that even if you are reduced to x8, the speed difference is negligible, on the order of 2-3% slower vs x16 for even an RTX Titan. If you're looking to simply add an extra PCI-E card to get back some SATA connectors due to running dual NVMe M.2 drives (and subsequently losing some onboard SATA ports), the best option is a small x1 PCI-E add-in card like this one. The PCI-E slots all operate through the chipset, over DMI, and do not utilize lanes directly connected to the CPU. This will then let you feed all 16 PCI-E lanes to the top slot for your graphics card, without needing to fiddle with the third x16 (physical) slot at all. It's still confusing, but hopefully that helped to sort of visualize it a bit. There's a post on Level1Techs discussing this feature, which also includes a press image showing a sample configuration utilizing this mode - a workstation with a Quadro and two Intel PCI-E SSDs. Both drives would be able to achieve maximum speed as they are not fighting for bandwidth across the DMI link. Good stuff.

Your GPU is definitely the most modest piece of hardware in the equation, but I think you'll be pleasantly surprised at how well the GTX 980 does when paired with a stronger CPU. It performs on par with the more recent GTX 1060 6GB or RX 580 in terms of performance, and you'll find plenty of happy users with those cards. 1080p is the perfect resolution for that GPU. This still gives you the ability to achieve 50-90FPS in most modern games. With a few detail setting changes even 1440P isn't out of the question if your display is capable of showing a higher resolution. If you're not a competitive gamer chasing the highest refresh rates, it'll serve you perfectly fine for now! If you're curious about some expected performance with that graphics card, this particular benchmark performance summary of a GTX 1060 6GB at TechPowerUp is a very good reference for you, for a few specific reasons: · The review system uses an overclocked i7-8700K - this helps show the various GPUs in their best light. Your 3700X results will vary from this, but it's close enough to get you within 5-10% of the benchmark results as the cards we're focused on will dictate the performance limits. · The games tested are all fairly modern, popular, and diverse - they include strategy games like Civilization V in addition to some first person shooters and racing sims. · As you flip through the individual game benchmark results, you can gauge for yourself if an upgrade would bring you a tangible benefit, or if you're comfortable holding off until a generation or two down the road. · This link is to a more recent overall performance summary which includes the RTX Super series cards as well as AMD's RX5700/5700XT GPUs as comparison points. On the chart, your GTX 980 will rank almost identically to the GTX 1060 6GB and RX 580. If your display is limited to 1080P at 60Hz, a new card can help eliminate performance dips or slowdown, but will also spend a lot of time underutilized. With a faster GPU, being able to drive a high refresh rate display (eg, 144Hz or 240Hz) or a higher resolution (1440P, 4K, etc) is the most obvious benefit. For now? For playing the occasional game, you're good. Check back in six months to a year when the next batch of GPUs is released and take advantage of a discounted midrange GPU.

Absolutely. Most tasks that fully utilize a processor (such as encoding) will show almost linear increases in performance. So if the eight core 3700X renders a video project for you in ten minutes, a 12 core 3900X would finish in just under seven minutes. While that doesn't sound like much, it adds up fast. If you work from home and do five videos a day, that's over an hour each week saved. Obviously the speed benefit will improve performance for other tasks as well, such as applying fades, filters and other video or photography editing operations, to say nothing of typical day to day usage. To be clear, not all programs will make use of a high core count CPU. Some programs (or program features) utilize only one or two threads, often necessarily due to the way data needs to be processed sequentially. While video encoding software (eg, Handbrake) will make use of as much power as you can throw at it, programs like Photoshop CC make little use of the extra cores. Another factor is that some programs rely on the GPU to process video encoding - eg, nvidia's NVENC. These encoders are limited more by the performance of the GPU than the system's CPU, so the improvement from a higher core processor is lessened. This is why it's so important to try and find real-world benchmarks for the software you use. One often underrated advantage of the current AM4-based processor ecosystem is that there is a clear option for upgrading to a faster processor, should you choose to first build a system around a less expensive six or eight core CPU. The 12 core 3900X ($500) and "any week now" 16 core 3950X ($750) are essentially drop-in replacements on X570 based motherboards. Many existing older AM4 based motherboards are also compatible, but support varies by vendor and model. If you go with an eight core model, you'll be in the sweet spot for content creation - sufficient oomph to power through most any application right now. Going to a 12 core will give you an appreciable speed increase for situations which fully load your processor, such as video rendering. Whether or not the +30% performance bump is worth the extra $100-150 is up to you - if you run jobs overnight, you wouldn't necessarily benefit and would be wasting money; if you're wanting to run jobs while continuing to use your computer, a 3900X or upcoming 3950X (with an approximate +50% performance bump over the eight core CPUs) might be just the ticket. Thanks for the feedback, I appreciate it! In the real world I've spent the majority of my time in retail management within the sporting goods industry. Being (happily) between companies at the moment, I've spent the summer hiking and camping; as winter returns I'll be hitting the sales floor again. Computers and tech have been a lifelong hobby of mine, and I firmly believe that keeping that interest separate from work has allowed me to remain enthusiastic and objective. I enjoy helping people with what I know and working to learn more about what I don't. I've been active on other forums going back to the '90s and on newsgroups before that (I was a geeky kid), but hadn't gotten around to creating a profile on LTT. Going to give it a whirl and pop in occasionally. The 50TB? It's mostly for my Plex server! I do have a fair amount of storage dedicated to games (my Steam directory is about 2TB worth of games I'll probably never finish), but most of it is media. I keep my DVD and Bluray collection on my system so it can be cataloged and streamed to TVs, along with my CDs which can be sent to my stereo or wireless speakers. With Bluray movies coming in at about 30-50GB each (and 4k/UHD approaching 100GB), that space doesn't go as far as I'd prefer. My internet speed is pretty bad, so my collection is almost entirely physical media. Changing discs gets pretty old when I can just pull it up with the television's Plex app without having to get up off the couch.

I'll bite. The fans on most X570 boards are pretty quiet. Some models (for example, MSI's "Zero Frozr" mode") have a mode where the fan won't turn on unless the chipset exceeds a pre-specified temperature threshold. This means that it is both quieter in normal operation as well as putting less wear and tear on the fan. For most users, the fan is not a big deal. Your CPU fan, case fans and GPU fans are all likely to be more audible, even accounting for any difference in sound frequency due to the RPM or size of the chipset fan. If it does fail years down the road, they're pretty easy to replace for a few bucks. The PCH consumes ~11W under load, compared to the ~5-7W on B350/X370 and B450/X470. Overall power consumption increase is quite minimal, on the order of adding a single SSD. NVMe M.2 drives can get hot, due to the small area, under extended read or write operations. Some (typically high performing) models will reduce performance if this happens, but it will not affect reliability. Most X470/X570 motherboards include NVMe M.2 heatsinks as part of their design, and aftermarket heatsinks are available fairly inexpensively. Some people like to watch blinky lights. If you're doing rendering, heavy gaming + streaming, database work, AI/computation or other "heavy" tasks, the more cores the better. In general, if you don't know whether or not you need more cores, you don't need more cores. The Ryzen 3700X is considered a sweet spot since it gives you enough cores to run a modern AAA game at full tilt and still have a few cores left over for background tasks or streaming. For only gaming, a 3600X will perform nearly identical. The 3800X is effectively a higher quality 3700X, and will hold slightly higher boost frequencies under typical conditions than a 3700X. Generally considered to be not worth it unless you intend to overclock or don't mind throwing money at a product for an extra 1-2% speed increase under Precision Boost conditions. Similar to the Intel i7-2700K back in the Sandy Bridge CPU days. Nearly identical performance to the cheaper i7-2600K of the era, but at questionably extra cost.