|
People at tomshardware told me upgrading my old LGA 775 rig wasn't worth it, I should just bite the bullet and get a Skylake system. It couldn't even play Hulu shows without video lag (running an e2180, a 2GHz dual-core CPU. An upgrade to the x9770 is still ~$200), so I was getting ready to toss it when I heard that Xeon processors for LGA 771 motherboards work perfectly well with a simple cosmetic modification: cut off a couple plastic nubs, and put a sticker over two pins. That's it. I popped a $20 Xeon 5450 in and now I have a 3GHz quad core HTPC that can run every streaming service smooth as cable, as well modern games. The 771 Xeon processors are more power efficient versions of their 775 counterparts, and cheaper to boot. They're a great upgrade, so why were they made to be cosmetically incompatible with my 775 motherboard? Fansy fucked around with this message at 15:36 on Jun 17, 2016 |
#
?
Jun 17, 2016 07:09
|
|
|
|
| # ? Apr 26, 2024 23:20 |
|
|
Because money.
|
|
#
?
Jun 17, 2016 10:17
|
|
|
That was my knee jerk conclusion, but how exactly does intel profit by not supporting a gigabyte motherboard?
|
|
#
?
Jun 17, 2016 15:37
|
|
|
Intel charges more for chipsets for motherboards that are compatible with Xeons, so if they let you use Xeons in consumer boards they wouldn't make as much money selling those expensive chipsets.
|
|
#
?
Jun 17, 2016 16:21
|
|
|
Alright, that makes more sense. Off the wall question: if the industry prioritized keeping old hardware relevant, with new CPUs for old sockets, firmware and BIOS updates and so on, how old do you imagine a motherboard could be before you'd no longer be able to upgrade and run modern games / software?
|
|
#
?
Jun 17, 2016 17:04
|
|
|
Fansy posted:Alright, that makes more sense. At the current time, forever. The reason they change is so that you are forced to update when you finally go to get a new upgrade. Otherwise mainboard manufacturers wouldn't have a good stream of income. If you bought a good powerful C2D or C2Q back in the day, you could still be using that assuming no hardware failures. The only reason to upgrade an average system is for specific needs, such as yours. I know a lot of people still trucking with C2Qs and have no intention of updating until they catch on fire because for what they do, the new processors don't really add a lot. It's different if you are running something Doom 4 fully maxed or something - then performance increases are nifty. But that's the enthusiast. If the average person took their broken system to Best Buy and get their C2Q upgraded to the latest i5/i7 - they would be out $200-300 and only intel would benefit. If you have to replace your motherboard, then multiple vendors benefit: The OEM (Dell, etc), the mainboard manufacturer, the ram manufacturer, the hard drive manufacturer, Microsoft, Intel, and so forth. On the other hand, there's some logistical issues with not changing your pin layouts each time you shrink the die and move to a new process, too. It would cost more (And anger business partners as noted above) to design your fabulous new 3nm fab process to match something like an LGA771 pinout.
|
|
#
?
Jun 17, 2016 22:49
|
|
|
There are a few hurdles I can think of to longer motherboard lifespans. The most basic is that planned obsolescence is profitable for Intel through chipsets and for the board manufacturers themselves. Additionally, board manufacturers do not like having to develop and test lots of EFI updates that they are not getting paid for to support substantially different new generations of processors. Many board manufacturers already do not support all of the pin- and chipset-compatible processors that they could, in fact. Maybe they could make newer firmware profitable by selling it like game DLC, but besides the obvious pirating problem this seems too niche to be feasible. Processors of different generations operate in different voltage ranges, so motherboard voltage regulation would have to be more flexible - mostly in terms of providing good stability and granular enough settings as the voltage goes down with new generations. This could be addressed with good design and EFI updates, but again it's more work that would have to be somehow profitable. Most people never consider upgrading their old processor even when good options do exist, so this would mostly affect the niche of enthusiasts who started out buying high end. I feel like CPU updates alone might be a small bonus to old motherboards, but to be honest while it's the main reason to upgrade it's far from the only one. Memory bandwidth seems to double roughly every 5-6 years, so putting a new processor in an old board would be kind of like running single channel memory. Storage is in the middle of switching over to PCIe as an interface due to SATA being clearly insufficient to keep up with SSDs, but if you put a new processor in an old board you might be stuck with 300MBps SATA so you'd have to hope that your EFI updates give you the ability to boot from PCIe if you ever want to improve that. For that matter, you'd have to hope your board is new enough to actually have more than one PCIe slot and not be full of PCI ports which are largely useless these days. That's especially true if you wanted USB 3.0 too. Gigabit Ethernet was still pretty common 7-8 years ago, but on a cheap system you could totally have Fast Ethernet integrated so if you wanted to upgrade from that you'd need another card still. And like Arsten says, if you're OK with all of that then the old processor probably isn't much of an additional limitation anyway. For web/office/multimedia stuff a fast Core 2 Duo or Core 2 Quad is still not bad if you put it with enough memory and an SSD. Since Nehalem and especially Sandy Bridge, progress in sheer single-thread CPU power has slowed down anyway and Intel has shifted focus to power efficiency for the mobile sector and core count(+power efficiency) for servers. Desktops are caught in a land of small increases and added features, and so I'm still happy with my 2500K and going to continue that way at least until I see what Skylake-E and Kaby + Cannonlake have in store. Eletriarnation fucked around with this message at 00:18 on Jun 18, 2016 |
|
#
?
Jun 18, 2016 00:15
|
|
|
Technology improvements require new motherboards. For one thing, the northbridge doesn't exist any more, it's moved into the processor. Also LGA 775 is two generations of RAM behind, as we are up to DDR4 now.
|
|
#
?
Jun 19, 2016 22:14
|
|
|
Eletriarnation posted:And like Arsten says, if you're OK with all of that then the old processor probably isn't much of an additional limitation anyway. For web/office/multimedia stuff a fast Core 2 Duo or Core 2 Quad is still not bad if you put it with enough memory and an SSD. Since Nehalem and especially Sandy Bridge, progress in sheer single-thread CPU power has slowed down anyway and Intel has shifted focus to power efficiency for the mobile sector and core count(+power efficiency) for servers. Desktops are caught in a land of small increases and added features, and so I'm still happy with my 2500K and going to continue that way at least until I see what Skylake-E and Kaby + Cannonlake have in store. Intel's modern low end stuff (Celeron and Atom etc) do about as good as a job as the mid-range Core/Core 2 series with far less power consumption and are quite inexpensive. It's surprising how functional Windows 10 is when paired with a low-end CPU.
|
|
#
?
Jun 20, 2016 08:03
|
|
|
|
| # ? Apr 26, 2024 23:20 |
|
|
Don't forget that even though the same physical socket is used pinouts may change between generations on the processor plugging in to the same(notionally, keyed differently) socket. Unfortunately long gone are the days when a Socket 7 motherboard meant Socket 7 compatibility across anything you could buy(as long as your MB and CPU at least had one matching FSB frequency for standard clocks, if not is was possible to get things running most of the time with the jumper configurations even if not in the manual). Even if the pinout wasn't changed, the memory controller likely wouldn't support older memory standards. And as mentioned earlier, VRMs are matched to a specific generation of processor, having an "efficient" voltage level and ranges above and beyond which it may serve but not efficiently or for extended periods of time. Physical sockets are changing less often now not only due to the slowing of Moore's law but also the reusing of physical sockets and carriers to save costs in engineering, every physical socket requires engineering in both mechanical and electronic schools. Mechanical to confirm it won't break the core under normal stressing and thermal characteristics. This needs to be redone for each core on the socket as thermals or physical stressing may change despite the same mounting with the core being so much smaller or larger compared to the heatspreader mounted on a modern CPU. Then we have the electronic side, where differential pairs must be kept together while parallel buses may be separated, and power rails must be separated from all these concerns are not only closely matched to the CPU core but also the support chips and all other peripherals. So we're now in a situation due to to CAD for large format PCBs that it's cheaper to reuse a physical socket and change it's pinout to suit the core and support chip than it is to physically re engineer the package. Even though a LGA1151 socket could physically with modification take a LGA1150 CPU, the pinout is likely vastly different to prevent such interchange. And all of this is disregarding the market segmentation argument, which is simply there to make money for the companies involved.
|
|
#
?
Jun 20, 2016 10:23
|
|
