AMD vs. Intel 2020: Who Makes the Best CPUs?
tomshardware.com
tomshardware.com
I've been building systems for almost 30 years and in that time Intel has mostly been the performance champion AND has always screwed you over on price. Sometimes they lead by a lot, sometimes by very little or not at all but their platform is always more expensive.
Both Intel and AMD introduce a brand (e.g. Pentium, Ahtlon) as their new flagship and then overtime erode the brand each year with lower spec products. Eventually they stop introducing higher spec products altogether and the brand sinks to the bottom of their product lineup as a new brand takes the top ranks (e.g. i9, Ryzen).
Depending on when you get into building PCs you might have brand loyalty because maybe your first system was a Celeron you unlocked and ran at PIII speeds. Or maybe you picked up a first gen Althon that was the performance king for a few quarters, or the Athlon II when Intel was floundering with Netburst. Maybe you started out with Core and in your view Intel has always dominated AMD. Regardless your perceptions are skewed.
Identify your workload and your budget, buy the best thing that meets your needs regardless of brand. Intel is historically more stable and less problematic. AMD usually offers significantly more performance per $ on the low end. Then again that all takes too much effort, just buy whatever makes you feel good about yourself on the internet according to these click bait articles.
If you're making purchasing decisions, you should be looking mostly at your specific needs, and the big picture around what you need and how to get good value for your money. If you can afford it, vote your conscious as well, and for a healthy future.
I've owned Cyrux, Pentium, Athlon, Core Duo, Phenom, Core i# and Ryzen. I'm partial to AMD as an underdog, and having a personality that always rooted for them. I'm unimpressed with Intel lately, but up until this year I still wouldn't buy an AMD laptop (though I've recommended a few for friends and family if the need and budget fit).
This article does try to paint "one answer for all questions" and of course that's silly. For most consumers I think they'll be better off this year with choosing Ryzen. But the Ryzen laptop chips that are impressing everyone seem to be in short supply. So I can't cheer them on just yet. I don't purchase for datacenters or professional rendering so I wouldn't dare weigh in there.
To sum up, focus on your needs, and individual options, not brand names!
AMD has largely taken the performance crown thanks to their Zen CPU platform and move to using TSMC for their process as Intel has lagged behind, with major problems deriving from their inability to get 10nm working.
If it helps to give you some perspective on how such a drastic change could occur, Intel’s 10nm process was originally meant to be ready for release in 2014!, and yet they still haven’t managed to scale out production after six years of debugging and failure.
It’s a combination of AMD simultaneously doing extremely well and Intel severely stagnating that has allowed this to happen.
I have no idea what you're talking about with your vague remark AND I think your attempt at pedantry has only helped to solidify the point I was making.
You remarked "You’re a bit out of date here" without any clarification. Given that I spoke historically and in generalities, it's unclear what you were referring to.
As far as I could tell you were dismissing everything I said AND then you went on to state technical facts surrounding recent events that in no way conflicted or invalidated anything I said.
Now you're asserting that it was a failure on my part to read your comment in full and comprehend what you said, allowing you to dismiss anything I say in response.
Basically you're trying to say I'm either misinformed or stupid, when in reality you're just not taking the time to write a coherent argument.
x570 -> Requires active cooling (fans) on the chipset because it throws off too much heat. The proprietray fan will die and your motherboard needs replacing. Congrats! Also, the PCIe lane configurations are weird (x8 x8 x8).
x470 -> Don't even look at your memory the wrong way if you want it to boot. Also, you need an older gen Ryzen/Athlon to update the BIOS to something that works with Ryzen 3000. Sure, there are stickers all over the box touting 'Ryzen 3000 compatibility!' but that's only after you update the BIOS. The BIOS the board ships with won't work at all.
b450 -> Dicier to run new stuff on such an older platform. The implementations are supposedly power weak for higher core count 3900/3950 CPUs.
AMD really dropped the ball on these chipset. They're the things that are leading to terrible user experiences and why folks end up going back to Intel after countless dollars wasted on shipping, RMAs, and swapping gear. No motherboard manufacturer actually designs anything. They just bundle chips together and change logos in the BIOS. Some might have a "flashback" button that works on half the boards, but there's little to differentiate one from the other.
If you want something that just works and is fast, stick with Intel because their chipsets are better. I'm writing this on a 3900x, too...
EDIT: Since many are doubting what I went through, let me list the permutations of boards, DIMMs, and CPUs.
CPUs: 3900x and 220GE DIMMs: Corsair Ballistix, G.Skill FlareX, Samsung ECC (unbuffered) Boards: x470 Taichi and Hero VII
I finally got my third Taichi to boot with a 220GE to the point that I could flash a BIOS to something "stable" for the 3900x.
I've literally built hundreds of servers and PCs in my lifetime (including AMD Romes). I know what I'm doing. This ecosystem just isn't mature yet!
But then I'm not an overclocker and the only time I go into the BIOS is when something breaks and/or I need to change the boot order.
Maybe the quality control is what's at issue?
If you read the negative reviews, you'll see reports of "I RMA'd twice and only the third board worked." Not one or two and more than a few.
If you read the comments you'll find a lot of people stating 9 months ago that the problem is fixed with a BIOS update.
> x570 -> Requires active cooling (fans) on the chipset because it throws off too much heat. The proprietray fan will die and your motherboard needs replacing.
AMD does require fans on the chipsets for X570 boards, but different board makers have treated this differently. My understanding is that some boards don't spin it up unless you're putting some serious PCIE 4.0 bandwidth in. Either way, "The fan will die and needs replacing" may be true someday (as it would be with any fan), but is not a guarantee.
>x470 -> Don't even look at your memory the wrong way if you want it to boot. Also, you need an older gen Ryzen/Athlon to update the BIOS to something that works with Ryzen 3000. Sure, there are stickers all over the box touting 'Ryzen 3000 compatibility!' but that's only after you update the BIOS. The BIOS the board ships with won't work at all.
First gen Ryzen (1000 series chips) had some memory compatibility issues, but 2000 and 3000 chips haven't had many issues. It's true that you needed to do a BIOS upgrade on x470/B450 boards when 3000 series chips were released, but those stickers indicate that the board maker installed a newer BIOS and they should work out of the box. Unless you manage to find an X470 board that's been sitting on the shelf since before last July any x470 board you buy today is extremely likely to be compatible with all currently released Ryzen chips out of the box.
> b450 -> Dicier to run new stuff on such an older platform. The implementations are supposedly power weak for higher core count 3900/3950 CPUs.
3000 series chips use a smaller process and are more power efficient per-core than 2000 and 1000 series chips, so you should have no issue running an 8 core chip in a B450 board. 12-core and 16-core chips might have some issues, but who buys a $79 motherboard and sticks a $500/$900 CPU in it?
I went through a pair of x470 Taichi's from Amazon last month. Both had non-compatible BIOS that wouldn't work with 3000-series. One so old that Dr Debug wouldn't even do anything. Just blank.
It's not just me. Read the negative product reviews of almost any x470 boards, and you'll see memory and non-POST issues very recently (followed by the response macro from mfg'er saying everything is fine just use 1 DIMM, etc). It's very much a crapshoot with AMD chipsets. You get a good one and everything's great. You get a few bad ones and you're stuck in swap/return/RMA hell.
Also, how can it be FUD when it's exactly what I went through?
Edit: The "Ryzen 3000 compatible" stickers do NOT mean it will work out of the box. It means that it CAN work...after you update the BIOS with an old CPU.
AMD will offer you a CPU "boot kit" to get it going, but you're risking on missing out on many 30 day return windows with shipping, covid19, processing, etc. Avoid it all and just get an Intel until the next refresh.
I was wary of the issue you raise, but lucked out, and figured the issues you talk of had washed through. Maybe the warehouse staff who boxed it looked at the order and gave me a recent unit.
What action could AMD take to improve the situation? I wanted to go to AMD because I expect to get strong upgrade options from it - certainly to a 3950x, and possibly further. But - a complication of that is that older motherboards need BIOS updates.
I am surprised at how slap-dash the motherboard is. It supports ECC ram - great. But BIOS configuration is a confused mess.
The workstation/server platforms ARE better. Having a BMC to flash/recover things is much more reliable.
So, you bough the same board twice and expected different results, and then are painting all motherboards with that experience.
It's unsurprising that motherboards need microcode updates to work with newer CPUs. Intel CPUs also worked that way, except that 1) they only rarely support a socket for more than two generations anymore, and sometimes not even that, so you physically need a new motherboard to upgrade your CPU anyways, and 2) Intel CPUs haven't had any major architectural changes since 2016.
I tried an Athlon 220GE in each one, too. It's not just a Ryzen 3000 problem being too new and requiring an update. This platform just isn't solid and that's my point.
What's a physical retailer? San Francisco is the supposed capital of the tech world. It only has a single computer store (Central). It's not even open right now. I guess you could count a big box store like Best Buy, but they're not going to help out much right now other than exchanging like-kind items.
As I've already said, it is totally normal for new CPUs to require motherboard firmware updates to function.
There’s probably a little more dodginess to the overall AMD ecosystem, but I expect that to be shaken out shortly (if it already hasn’t been)
Happens every day. People are still buying boards like that off Amazon. All sale/clearance stock is likely still on the old bios.
I don't know if the fact that I was looking at said recommended list is what negates the issue you're talking about, but I had minimal issues in my build (mine was from also including a PCIe to 8 SATA port card. Moved the main drive off of it, installed the drivers and started working fine). Initially tried to do a Windows VM w/ GeForce 1080 but scrapped it after driver issues with the device in VM (apparently, Nvidia doesn't want non-Quadra GPUs in VMs).
(FWIW: I’ve built 2 Ryzen 2000 and 2 Ryzen 3000 boxes, all on x470 and B450. No RAM issues.)
>If you're used to slapping in name-brand memory to a board and booting it up[...]
Yeah, I am and that's exactly what I did with a new Ryzen CPU. How is this shit the top comment?
I'm not even talking about XMP stuff. I'm talking about booting with basic 2400/2666 stuff. "Yeah that's good RAM but not on our QVL. Get bent."
How do you go from, "one particular mobo from one manufacturer is bad" to "all mobo makers are bad" and you attribute some sneaky malicious motive behind this.
That's too much anecdote and emotional spin for me.
Don't get these problems on AMD server/workstation platforms. You don't get them anywhere on Intel. But hey, the main article here also gave the nod to Intel in chipset/driver so I must just be all emotional spin.
It's not that people are doubting your experience, it's that the Intel boards aren't any better. Many of the Intel boards also need a BIOS update to support later CPUs than they were originally released with, when they support them at all. DDR4 memory is finicky as heck. People end up having to RMA Intel boards all the time.
It's largely caused by the motherboard makers losing the ability to differentiate themselves. Half of what used to be on the motherboard is now integrated into modern CPUs and the other half is in the chipset which is now also made by the CPU manufacturer. Two motherboards with the same chipset are largely fungible so they have to compete on price and the quality suffers. It's a general problem and buying any consumer motherboard these days is kind of a crapshoot.
Here's a motherboard with 5 eggs and 50 reviews:
https://www.newegg.com/p/N82E16813121180
It's a Pentium 4 motherboard from ~15 years ago. See if you can find a current generation Intel motherboard on there with 5 eggs from half that many reviews. I don't see any.
Intel boards do need upgrades to work with newer CPU's. This is true. But their updates actually work when they say they add support. Go back to July 2019 and read all the forum posts and YouTube videos of everyone complaining about Ryzen 3000, "Well Gigabyte said this release works...but it doesn't." And on and on.
It's a sorry state of affairs all around, but Intel is better than AMD. I write this on an AMD rig and after spending way too much money getting it going.
All that is integrated now so every board with a given chipset is basically the same. Then they have to compete more aggressively on price, and the way they get the price down is by using cheaper heatsinks and spending less time developing firmware and things like that.
The Intel boards have no immunity from that. The ratings there are just as bad. The older Intel boards were better but so were the older AMD boards.
(The same applies to laptops, so far i've never came across one which hadn't a label or engraving that told me all i need to know for choosing the right replacement.)
I had to disable mitigations to complete a filesystem checksum, it was on the order of 1000x slower.
Given the forthcoming i9 will hit 5.3 GHz, I’m choosing Intel for my next build. Until AMD comes out with libraries and compilers targeting scientific computing, Intel may still have the lead.
Ryzen does support ECC, and > 64 GB, so that is a plus.
https://www.amd.com/en/processors/amd-epyc-aws https://www.amd.com/en/processors/epyc-google-cloud
and
https://aws.amazon.com/ec2/amd/ https://cloud.google.com/blog/products/compute/amd-epyc-proc...
1. The single thread performance gap is very small (5-10% depending on application) or even non-existent in some cases.
2. Your computer runs more than one process at any given moment. The more cores, the more processes you can run without an impact on performance
3. That single core performance comes at the cost of efficiency and general power usage. Intel has had to clock their 14nm process very high to achieve that small lead in single thread performance.
Imagine you have a single core compilation process that takes two minutes that you must run 50 times a day. This can happen. Ten percent brings 12 seconds per compile or ten minutes a day. This might not sound great but it means you can get an extra five compiles done. You can also get increased likelihood that you can remember what you were working on. Therefore the marginal gain in single thread performance is well worth having in this use case. The expense of genuine Intel becomes worth it even if you don't have so many cores.
the compilation definitely has parallelism.
Both AMD and Intel are ridiculously fast! You can't really go wrong with either.
The code is all async by default and that is easy to load onto green threads.
This used to be implemented with callback hell, now it's Promises which is actually really cosy (and what Rust is adopting).
Random article I found on it: https://medium.com/better-programming/is-node-js-really-sing...
But it is also like 10 times faster than other compilers.
Who runs builds with a single core compilation process? Are we back in 1990?
You are hard-pressed to find a CPU with a single core nowadays, even in the cheapest of consumer segments. Why are you fabricating such an unrealistic case?
By the way, one of the biggest improvements in software performance in the past decade comes from link time optimisations during build (LTO), that's all done during linking and it takes FOREVER.
From anecdotal experience of the code I am on right now, relinking a few hundred megabytes of .o's takes less than 2 seconds with lld on my 6900k.
Depends how many targets you're building. If you have enough memory, you can link several libraries in parallel.
Linking is also IO-bound, and already a solved problem in the sense that there are already plenty of solutions that render linking to be a non-issue. It's far from a decent CPU benchmark.
Wanting an Intel over a Zen 1 for single threaded performance? Completely understandable.
Can you give some examples of CPU intensive apps that are not multi threaded in 2020?
Honestly my biggest problem right now is that I don't even know if there's anything I can buy/upgrade to speed it up. I don't know how to determine what the bottleneck is, so to speak.
Of course if price is no object then the above doesn't apply.
But, this is a narrow-minded perspective to have. The future cannot possibly be that we magically invent an x86 core that can clock to 8GHz w/ practical power & cooling. One has to push forward in some incremental way. The answer is adding more parallel execution units. Sure, lots of software and techniques used today are not feasible to directly port over, but that doesn't mean it has to stay that way.
As a software developer, I feel it is my duty to help push the state of the art forward wherever possible. I have already begun working on projects that are designed around the concept of extremely high core counts becoming the norm. Having 32 logical threads sitting in your local workstation makes it possible to practically experiment with the lower bounds of what would be required in a larger production deployment.
There are types of software that we can consider building today that would have sounded absolutely insane just a few years ago. Cloud gaming is now a thing falling within practical reach. How many more x86 cores per instance would we need before we could start thinking about rendering 100% of client views server-side (i.e. eliminate most HTML/CSS/JS) or deprecating the usefulness of discrete GPUs altogether? Pulling everything into a single memory domain could do wonders for breaking down the ages-old GPU/CPU functionality silos. One way to do it all with one instruction set and a single cache-coherent memory architecture...