I went from a GTX660 to a 960 recently and don't regret it in the least. Not only are things prettier on higher settings, they are a lot smoother too. So considering it's the same generational gap for the two of us, definitely go for it if it's a reasonable price.I'm gonna piggyback on this thread: recently upgraded except for graphics card. Currently have a GT 640. I'm actually not unhappy with it (I'm more of a casual gamer and I don't really care if I have everything set high), but I would like to upgrade. Wondering if I should hold out for Polaris/Pascal, or if I'm better served getting something like a GTX 950 for now and waiting until more budget-minded cards on the new architecture come out.
One of the bigger reasons to go with a newer card is that they can be used with bigger monitors. Older cards top out at 1-2GB, which really isn't enough for really big monitors ... or for cards which use VSR/DSR to generate a much larger virtual image and then scale it down for quality. Newer cards can be had with 4GB or 8GB (or 12GB if you want to go ridiculous). Along with the larger amount of VRAM, you would also get a newer graphics engine that is more efficient (meaning maybe not many additional fps, but a much higher minimum fps) and can do things like decoding H.265/HEVC in hardware (curently found only in NVIDIA GTX 960/970/980 or AMD Fiji-based cards). If you're going to wait, Pascal is going to be demoed in April (actual release still unclear), while Polaris isn't coming until Q3/Q4 2016.I'm gonna piggyback on this thread: recently upgraded except for graphics card. Currently have a GT 640. I'm actually not unhappy with it (I'm more of a casual gamer and I don't really care if I have everything set high), but I would like to upgrade. Wondering if I should hold out for Polaris/Pascal, or if I'm better served getting something like a GTX 950 for now and waiting until more budget-minded cards on the new architecture come out.
An SSD is great for anything that you want to load quickly from a hard drive. So your OS, definitely. Any games with lots of huge assets, yep. But don't store small files and shit on it, or things that get deleted often. The more you delete stuff off an SSD, the sooner it wears out, and it's irreversible. Even if you put something on new in its place. So whatever you install on your SSD, make sure it's something you want to pretty much always be installed.Upgraded, finally!
Now I'm running with a GeForce GTX 1060, with an EVGA 500W power supply.
It's running everything now, and smoothly, too! Still not bleeding edge, but then again, I don't need it to be right now (nor can I afford that).
Ultimately gonna get an SSD, especially as prices keep on coming down for them.... would there be any advantages to getting a larger size? I thought you really only got the most bang for your buck by running your OS from them? I confess ignorance, here...
Right now there's a memory shortage, so SSD prices, strangely, are actually going up (especially for older-style SLC and MLC drives). For speed and reliability, SLC is best, MLC is a happy medium, and TLC is the cheapest (and least durable). Even if you don't know what the acronym stands for, at least now you know how to rank them.Ultimately gonna get an SSD, especially as prices keep on coming down for them.... would there be any advantages to getting a larger size?
This is technically correct (the best kind!), however you'd have to erase and write vast quantities of data to ever get to the point where this actually matters. What google found (and they do erase and write a lot of data) for SSDs is that the age mattered more than the write cycles:But don't store small files and shit on it, or things that get deleted often. The more you delete stuff off an SSD, the sooner it wears out, and it's irreversible.
If you're going to spend the money on SSD, it's worth it to buy a good M.2 drive with a PCIE interface, and a PCI adaptor if your motherboard doesn't have a PCIE M.2 spot. The adaptors are under $20, the drives cost more, but they're also significantly faster (6x for some) than the SATA SSDs....if you're going to be getting a SATA-based SSD, you're going to be limited by the SATA interface's max speed of about 540MB/s anyway.
Related old link from March 12, 2015: The SSD Endurance Experiment: They're all deadFor a consumer, the reality is that you'll never approach the wear limit.
FYI: for hard drive people, 1TB = 1000GB, and 1PB = 1000TB.Clear evidence of flash wear appeared after 200TB of writes
...
The fact that the 840 Pro exceeded 2.4PB is nothing short of amazing
A caveat about this: Some drives are AHCI (older), some drives are NVMe (newer). NVMe drives are VASTLY superior to AHCI, but not every system supports NVMe, and some of them support NVMe BUT not as a boot drive. So don't just buy "blade" storage because it physically fits your board unless you make sure it works with your board, and even if it does, it may require you to reinstall Windows from scratch (in "EFI mode") in order to fully benefit from NVMe. I say this having to painstakingly hunt down an SM951* for my system once I discovered it only supports AHCI. Also make sure your PCIe<->M.2 adapter supports the same bus width and revision as your drive (PCIe x2 v. PCIe x4 and PCIe 2.0 v. PCIe 3.0, and furthermore some of them only support SATA even though they physically look the same) otherwise you're just throwing away speed.If you're going to spend the money on SSD, it's worth it to buy a good M.2 drive with a PCIE interface, and a PCI adaptor if your motherboard doesn't have a PCIE M.2 spot. The adaptors are under $20, the drives cost more, but they're also significantly faster (6x for some) than the SATA SSDs.
I don't remember exactly, but I think that experiment only tested MLC drives, dunno if it also included any TLC drives.Related old link from March 12, 2015: The SSD Endurance Experiment: They're all dead.
From skimming, definitely more than one type of drive (type of memory I mean) in that stress test.I don't remember exactly, but I think that experiment only tested MLC drives, dunno if it also included any TLC drives.
They only had one TLC drive in that test, and it started showing unrecoverable errors after only 100TB of writes, whereas the MLC drives didn't start throwing these errors until after 600TB.From skimming, definitely more than one type of drive (type of memory I mean) in that stress test.
From the article:They only had one TLC drive in that test, and it started showing unrecoverable errors after only 100TB of writes, whereas the MLC drives didn't start throwing these errors until after 600TB.
Showing that reallocations are necessary for some of the drive (But that it's still usable) isn't the same thing as "unrecoverable errors" IMO, but that's getting into semantics here. The article points out how most users only write a terabyte or two a YEAR to a drive, and that's different than your system drive versus your "downloading all the games!" drive. So even worst-case-regular-user of writing over 5TB a year to your drive (which is unlikely), the "bad" type of memory would still last over 100 years before complete failure. Even with 100TB only of headroom, that's still 20 years, with my "much larger than typical" usage of 5TB per year.But it suffered another spate of uncorrectable errors on the way to 900TB, and it died without warning before reaching a petabyte. *snip* Few expected a TLC SSD to last that long—and fewer still would have bet on it outlasting two MLC-based drives.
What about things like the swap file, which is constantly re-writing itself?From the article:
Showing that reallocations are necessary for some of the drive (But that it's still usable) isn't the same thing as "unrecoverable errors" IMO, but that's getting into semantics here. The article points out how most users only write a terabyte or two a YEAR to a drive, and that's different than your system drive versus your "downloading all the games!" drive. So even worst-case-regular-user of writing over 5TB a year to your drive (which is unlikely), the "bad" type of memory would still last over 100 years before complete failure. Even with 100TB only of headroom, that's still 20 years, with my "much larger than typical" usage of 5TB per year.
I still maintain, a typical user doesn't even have to THINK about this type of failure. I'd bet on another random electronic component of the board (solder connection, or whatever) to fail prior to the flash memory itself.
A typical user DOES have to think about this sort of error, but the appropriate preparation would be keeping some sort of backup, which any user should be doing anyway. And given the choice between a drive rated for a lifespan of 140TB v. a drive of 400TB, I'm going to pick the 400 one every time, even if it does cost $30 more.I still maintain, a typical user doesn't even have to THINK about this type of failure. I'd bet on another random electronic component of the board (solder connection, or whatever) to fail prior to the flash memory itself.
"Modern" OSes are supposed to be getting smarter about this, and start to do things in more SSD-friendly ways, but that's a great question. I'd like to see some testing.What about things like the swap file, which is constantly re-writing itself?
There's also the conventional wisdom that SSDs suffer complete failure less often than platter drives but lose data blocks more, which is part of why I said don't put things like pictures and documents on them.A typical user DOES have to think about this sort of error, but the appropriate preparation would be keeping some sort of backup, which any user should be doing anyway. And given the choice between a drive rated for a lifespan of 140TB v. a drive of 400TB, I'm going to pick the 400 one every time, even if it does cost $30 more.
Not sure this is a thing, at least no more often than platter drives. Would like to know where that theory comes from.There's also the conventional wisdom that SSDs suffer complete failure less often than platter drives but lose data blocks more, which is part of why I said don't put things like pictures and documents on them.