You have an SSD so as long as you use it for the OS and your games and use the HDD for storage you are fine. Its a good SSD as well - although any SSD is good.
(Edit: if you didn't have an SSD then NVMe would have been something to look at.)
So I'm seeing some people recommend AMD. So here's the details on what I'm thinking. I'm looking at the i5 8600k so I'm guessing the 5 2600X is the equivalent from AMD. Budget is around $400-500 though obviously lower is better. I would have to get a new heatsink and fan if I switched to AMD (current one supports 1151 not AM4). I'm using this exclusively for gaming.
So. Is it worth investing a little more in a new hs+f? Or would the stock work? Which option will have the most longevity? I'm willing to take a small performance hit now to save me having to upgrade for an extra year or two. I assume no matter what (from what I can see) a 650W psu will be enough to handle the new cpu and a 970?
Also just what about mobos? Because frankly I don't know what I'm looking for in either case. (Though I realize this question isn't really answerable until I decide whether to go with AMD or Intel.)
If you go with a Ryzen 5 2600X as compared to a Core i5 8600K, then you'll get a little less CPU performance in single-threaded programs and a little more in programs that scale well to many cores. That can also save you money in three ways:
I agree with Ryzen 5 2600X over I5 8600K. If you don't like my suggestion about I5 8400, then Ryzen 5 2600X is likely better price/performance than I5 8600K.
But do you have an SSD? Your computer sounds so old that you might not have, and if you don't have one then I think saving with CPU to buy an SSD would be a good idea.
That's one way to bring the price tag down, but if you're going to do that, then you might as well compare it to the Ryzen 5 2600 (non-X), and get about the same comparison on price/performance as before.
I think you're right. A month ago I5 8400 was still cheaper than Ryzen 5 2600, but it looks like they've swapped positions with I5 8400 price increasing and Ryzen 5 2600 price decreasing.
My info was outdated, and Ryzen 5 2600 would likely be better price/performance than I5 8400.
Sorry for posting outdated info.
Agree. The i5-8400, Ryzen 2600 and Ryzen 2600X are all "about the same". And since the Ryzen 2600 is the cheapest it will offer the best price-performance.
They are all very powerful though. So you can probably base your decision on price. And because prices change I suggest looking at 2 possible builds, one around a Ryzen 2600 and the other around an i5-8400. And see what the motherboard options are for both.
Is the fan a factor? If your current fan can be used to cool an i5-8400 then no - but I'm not sure it can. The fans that Intel include though are - imo - abyssmal. Way to noisy. AMD fans much better. So that would be an extra cost.
Do you really think that Sky Lake Refresh Refresh Refresh will be all that much better than Sky Lake Refresh Refresh? Because while sixth generation Sky Lake was something new, all that Intel has been able to do since then is spin their wheels while waiting for their 10 nm process node to not work.
Rumors say that the next refresh will offer 8 core CPUs, while their current mainstream lineup only goes up to 6 cores. If you think those two extra cores are worth waiting for, then maybe you wait. But otherwise, no. If you wait for 10 nm, you'll likely be waiting at least another year for the low end dual cores, and likely longer for anything bigger than that.
It's also worth mentioning that AMD CPUs are competitive now, and certainly the most competitive that they've been since Conroe arrived in 2006. If you buy a Core i7 8700K for $350 from Intel or its successor in the refresh at likely a higher price tag, you can argue that it's a little better than what AMD offers if you put a high value on single-threaded performance. But for anything below that, AMD offers a competitive product at a competitive price. I'm not saying that you should definitely buy AMD, but only that it's worth considering, and available today as they recently refreshed their lineup.
The single thread performance is what kills AMD vs. Intel for me. I've tried too many times to go for AMD cause of the lower cost, but they always felt worse than Intel. I wouldn't buy an AMD for gaming, personally. I'd buy it for something like a Video Editing RIG, where you can take advantage of the greater core count at lower cost and actually come out with a benefit vs. comparable (but higher costing Intel CPUs).
AMD also tends to have better iGPUs than Intel, which is good for workloads where you can use the iGPU for Display and dGPU for Compute (i.e. Video Editing with DaVinci Resolve).
Aside from that, I don't touch AMD.
I also feel like more software is optimized with a bias for Intel/Nvidia, which results in software running better on their CPUs. Anything that supports CUDA tends to de facto run better on Nvidia than Intel/AMD OpenCL.
I didn't even consider AMD when I bought my new machine a few months ago. I've already been bitten by their laggard single thread performance, multiple times, and I feel their support lifecycle for products is not as good as Intel or Nvidia's.
There's so much wrong with that post. Where do we begin?
For starters, yes, a few years ago, AMD had markedly worse single threaded performance than Intel. Ryzen didn't completely eliminate the gap, but it greatly reduced it. And even now, to the extent that Intel offers better single-threaded performance, it's due in large part (though not entirely) to higher clock speeds. If you're comparing a Core i7 8700K at 4.7 GHz to a Ryzen 7 2700X at 4.3 GHz, that's a considerable advantage for Intel. But among the parts that we're actually comparing, the Intel CPU has only a 100 MHz advantage over AMD, not 400 MHz.
If you're using a discrete GPU for compute, then why do you care how good the integrated GPU is? Are you going to play games on the integrated GPU and mine ethereum on the discrete GPU at the same time? Besides, the AMD parts that we've been discussing here don't even have an integrated GPU. And the original poster already has a discrete GPU: a GeForce GTX 970 that he isn't looking to replace.
Meanwhile, bringing up Nvidia constitutes going way off on a tangent. Unless you're using an integrated GPU--not what you want for gaming except on an extreme budget--the choice of a CPU and GPU are basically independent. You don't have to go Intel/Nvidia or all AMD. For example, the computer I'm typing this from has an Intel CPU and an AMD GPU.
And of all the reasons you might give for a gamer to prefer Nvidia, CUDA has got to be one of the most ridiculous. It's not that CUDA runs better on Nvidia. It's that CUDA can't run at all on anything other than Nvidia GPUs. That's the main reason for CUDA to still exist. To the degree that it has any other point, it's to say, look how easy Hello World is. If Nvidia going far out of their way to break compatibility is a reason to prefer one vendor over another, it's a reason to avoid Nvidia.
Even as an API for GPU programming, CUDA is disastrously awful. It's fine if all you want to do is put a wrapper around CPU code to get it running on a GPU, not particularly caring if it runs fast or correctly. Most GPU programming APIs will push you to write sane GPU code, which can mean more rewriting to make a quick proof of concept; CUDA actively discourages writing sane code. But it's a completely insane thing to use if you're trying to make a polished product and aren't being paid by Nvidia specifically to use CUDA. Even apart from pushing people to write an unmaintainable mess of code by not giving you the usual ways to clean up your code, CUDA is also an awful pain to support. Among other things, it requires compiling kernel code before host code, which completely breaks the forward compatibility features of every GPU programming API that a reasonable person might use.
If you want to talk about a support lifecycle, then yes, Nvidia is likely to offer driver support on a new architecture for a year or two longer than AMD. But we're already four years into the lifecycle support of Maxwell/Pascal (both AMD and Nvidia tend to end support for an entire architecture all at once), making it not at all clear whether an Nvidia GPU or AMD GPU that you could buy today will get driver support for longer. It's less clear where AMD will draw the line on ending driver support, as even Vega is still very heavily derivative of GCN from 2012. If optimizations is the issue as opposed to bug fixes, then rumors have a new Nvidia architecture launching very soon, at which point, the optimization work for Maxwell/Pascal will largely cease. AMD's next new architecture seems to be a little further away than that.
But again, that's all irrelevant, as the original poster already has his video card.
@Quizzical Thanks. Any thoughts on the build I suggested above your response?
What else do you have in your current build? Give your full system specs, including the particular SSD, power supply, case, and so forth.
Ahaha! It is mostly pretty old so don't drag me.
Case: Antec 300 Illusion ATX Fan: Cooler Master Hyper 212 Plus PSU: Corsair Enthusiast TX 650W HDD: Seagate 2TB SSD: Samsung 850 EVO 500GB GPU: MSI GTX 970 Current motherboard: Asrock P67 Extreme4 Current CPU: Intel i5-2500k Current RAM: GSkill Ripjaw DDR3 4x4GB Also an ASUS DVD burner
So here's what I'm currently thinking (and some reasoning).
MSI X470 Gaming Plus (seems like a reasonable board from reviews) Ryzen 5 2600 (I can oc it later to bring it more in line with the X) Stock cooler (not going to oc off the bat. I'll wait until I have the money to get a decent cooler) Corsair Vengeance LPX DDR4 3000 2x8GB (reasonable speed, good cost per GB)
This puts the total cost around $450 after promo on the motherboard, which is a fairly comfortable price for me right now.
Welcome to an era of even slower chips due to having to address a lot of this stuff.
Here is a question -- would you prefer 1) A "secure" system that runs 50% as fast at maximum and per $ cost, or 2) A less secure system that runs 200% as fast at maximum and per $ cost?
Personally, I'd be tempted to buy TWO systems. One a low-mid end system that was secure and I browsed the web, did email, banking etc on, and a second system that was very unsecure but blazingly fast that I would use for games.
People who are worried about security for their mail and banking and such then use their cell phone to check said are looking in the wrong place for their problems.
Welcome to an era of even slower chips due to having to address a lot of this stuff.
Here is a question -- would you prefer 1) A "secure" system that runs 50% as fast at maximum and per $ cost, or 2) A less secure system that runs 200% as fast at maximum and per $ cost?
3) A properly engineered system.
1 and 2 aren't the only two alternatives. 3 won't happen on old gen hardware. That's just screwed. The best thing Intel (or AMD) could do for architecture problems is fix it in new generations which is what they're doing.
Naw -- you can do either just fine. Only thing is a very secure system would be relatively quite slow and dedicated toward being very secure. In the less secure system, you simply don't care about whether one core can see what the other core is doing AT ALL. Security is a computer sense always has costs both in design and in silicon.
Comments
(Edit: if you didn't have an SSD then NVMe would have been something to look at.)
They are all very powerful though. So you can probably base your decision on price. And because prices change I suggest looking at 2 possible builds, one around a Ryzen 2600 and the other around an i5-8400. And see what the motherboard options are for both.
Is the fan a factor? If your current fan can be used to cool an i5-8400 then no - but I'm not sure it can. The fans that Intel include though are - imo - abyssmal. Way to noisy. AMD fans much better. So that would be an extra cost.
For starters, yes, a few years ago, AMD had markedly worse single threaded performance than Intel. Ryzen didn't completely eliminate the gap, but it greatly reduced it. And even now, to the extent that Intel offers better single-threaded performance, it's due in large part (though not entirely) to higher clock speeds. If you're comparing a Core i7 8700K at 4.7 GHz to a Ryzen 7 2700X at 4.3 GHz, that's a considerable advantage for Intel. But among the parts that we're actually comparing, the Intel CPU has only a 100 MHz advantage over AMD, not 400 MHz.
If you're using a discrete GPU for compute, then why do you care how good the integrated GPU is? Are you going to play games on the integrated GPU and mine ethereum on the discrete GPU at the same time? Besides, the AMD parts that we've been discussing here don't even have an integrated GPU. And the original poster already has a discrete GPU: a GeForce GTX 970 that he isn't looking to replace.
Meanwhile, bringing up Nvidia constitutes going way off on a tangent. Unless you're using an integrated GPU--not what you want for gaming except on an extreme budget--the choice of a CPU and GPU are basically independent. You don't have to go Intel/Nvidia or all AMD. For example, the computer I'm typing this from has an Intel CPU and an AMD GPU.
And of all the reasons you might give for a gamer to prefer Nvidia, CUDA has got to be one of the most ridiculous. It's not that CUDA runs better on Nvidia. It's that CUDA can't run at all on anything other than Nvidia GPUs. That's the main reason for CUDA to still exist. To the degree that it has any other point, it's to say, look how easy Hello World is. If Nvidia going far out of their way to break compatibility is a reason to prefer one vendor over another, it's a reason to avoid Nvidia.
Even as an API for GPU programming, CUDA is disastrously awful. It's fine if all you want to do is put a wrapper around CPU code to get it running on a GPU, not particularly caring if it runs fast or correctly. Most GPU programming APIs will push you to write sane GPU code, which can mean more rewriting to make a quick proof of concept; CUDA actively discourages writing sane code. But it's a completely insane thing to use if you're trying to make a polished product and aren't being paid by Nvidia specifically to use CUDA. Even apart from pushing people to write an unmaintainable mess of code by not giving you the usual ways to clean up your code, CUDA is also an awful pain to support. Among other things, it requires compiling kernel code before host code, which completely breaks the forward compatibility features of every GPU programming API that a reasonable person might use.
If you want to talk about a support lifecycle, then yes, Nvidia is likely to offer driver support on a new architecture for a year or two longer than AMD. But we're already four years into the lifecycle support of Maxwell/Pascal (both AMD and Nvidia tend to end support for an entire architecture all at once), making it not at all clear whether an Nvidia GPU or AMD GPU that you could buy today will get driver support for longer. It's less clear where AMD will draw the line on ending driver support, as even Vega is still very heavily derivative of GCN from 2012. If optimizations is the issue as opposed to bug fixes, then rumors have a new Nvidia architecture launching very soon, at which point, the optimization work for Maxwell/Pascal will largely cease. AMD's next new architecture seems to be a little further away than that.
But again, that's all irrelevant, as the original poster already has his video card.
MSI X470 Gaming Plus (seems like a reasonable board from reviews)
Ryzen 5 2600 (I can oc it later to bring it more in line with the X)
Stock cooler (not going to oc off the bat. I'll wait until I have the money to get a decent cooler)
Corsair Vengeance LPX DDR4 3000 2x8GB (reasonable speed, good cost per GB)
This puts the total cost around $450 after promo on the motherboard, which is a fairly comfortable price for me right now.
Thoughts anyone?
https://www.theregister.co.uk/2018/08/14/intel_l1_terminal_fault_bugs/
Time for some more microcode updates.
Here is a question -- would you prefer 1) A "secure" system that runs 50% as fast at maximum and per $ cost, or 2) A less secure system that runs 200% as fast at maximum and per $ cost?
People who are worried about security for their mail and banking and such then use their cell phone to check said are looking in the wrong place for their problems.
거북이는 목을 내밀 때 안 움직입니다
https://www.techradar.com/news/this-could-be-our-first-look-at-the-intel-core-i9-9900k