AMD Launches Ryzen 3000XT Series CPUs At Higher Clock Speeds To Battle Intel (hothardware.com) 44
MojoKid writes: Last month, AMD made its Ryzen 3000XT series processors official, after weeks of leaks and speculation. Ryzen 3000XT series processors are tweaked versions of the original 3000X series products, but with higher clocks and the ability to maintain turbo frequencies longer. Launching today, AMD's new Ryzen 5 3600XT is a 6 core/12-thread processor, with a 3.8GHz base clock and a 4.5GHz max boost clock. That's a 100MHz increase over the 3600X. The Ryzen 7 3800XT is an 8-core/16-thread processor with a base clock of 3.9GHz and a max boost clock of 4.7GHz, which is 200MHz higher than the original 3800X. Finally, the Ryzen 9 3900XT is a 12-core/24-thread processor with a base clock of 3.8GHz with a max boost clock of 4.7GHz, which is a 100MHz increase over the original Ryzen 9 3900X.
AMD also notes these new processors can maintain boost frequencies for somewhat longer durations as well, which should offer an additional performance uplift, based on refinements made to the chip's 7nm manufacturing process. In testing, the new CPUs offer small performance gains over their "non-XT" namesakes, with 100MHz - 200MHz increases in boost clocks resulting in roughly 2% - 5% increases to both their single and multi-threaded performance in most workloads. Those frequency increases come at the expense of slightly higher peak power consumption as well of course. The best news may be that AMD's original Ryzen 5 3600X, Ryzen 7 3800X, and the Ryzen 9 3900X will remain in the line-up for the time being, but their prices will be slashed a bit, with the new Ryzen 5 3600XT, Ryzen 7 3800XT, and Ryzen 9 3900XT arriving with the same $249, $399, and $499 introductory prices as the originals.
AMD also notes these new processors can maintain boost frequencies for somewhat longer durations as well, which should offer an additional performance uplift, based on refinements made to the chip's 7nm manufacturing process. In testing, the new CPUs offer small performance gains over their "non-XT" namesakes, with 100MHz - 200MHz increases in boost clocks resulting in roughly 2% - 5% increases to both their single and multi-threaded performance in most workloads. Those frequency increases come at the expense of slightly higher peak power consumption as well of course. The best news may be that AMD's original Ryzen 5 3600X, Ryzen 7 3800X, and the Ryzen 9 3900X will remain in the line-up for the time being, but their prices will be slashed a bit, with the new Ryzen 5 3600XT, Ryzen 7 3800XT, and Ryzen 9 3900XT arriving with the same $249, $399, and $499 introductory prices as the originals.
100 mhz per core (Score:5, Interesting)
A 100mhz might not seem like much until you give that to 12 cores and 24 threads.
Re: (Score:1)
But why? AMD already beats the crap out of Intel in the multicore segment. TFA states that AMD is battling Intel here, which these processors do not help with at all. Intel has the crown in lightly threaded performance, around 2 cores, with their superior frequency and boost management. If AMD wanted to step into Intel territory they'd need some way to boost single core clock speeds up into the 5GHz range. The XT family seems misplaced and without a clear purpose.
Re: (Score:1)
Comment removed (Score:4, Informative)
Re: (Score:2)
were coming out the fab easily running higher than stock clock on the same voltage
Define stock, because last I heard AMD was massively struggling to meet their "up-to" performance figures which they advertised, which effectively means they were struggling to meet their stock numbers under boost scenarios. Now maybe that was due to an over zealous marketing department, but quite frankly what marketing says is what is considered "stock".
If you mean higher than base, well no kidding, Core boosting has been a thing for a while now.
Considering they easily could have bumped up the price but instead LOWERED the price on the original X chips and slotted these in at the old X price?
You generally expect cost reductions as time goes on. This is
Re: (Score:1)
AMD did relaunch the 1600 on the zen+ process, but it's been impossible to buy it for months. Most attribute this to being replaced by the 3300x.
It's been pretty obvious that none of the ryzen processors are getting good silicon, with the 3800x being the worst offender.
Literally all of the "best of the best" silicon is going to epyc, with the 2nd best going to threadripper. It would be interesting to see if this cost advantage of having all the segments use the same chiplet is making AMD noncompetitive si
Re: (Score:3)
Re: (Score:2)
Yeah, the news here is that even the garbage tier silicon is now better. Paying $100 extra for slightly less garbage silicon still doesn't make any sense, just like it didn't a year ago.
Re: (Score:1)
Re: (Score:1)
A 100mhz might not seem like much until you give that to 12 cores and 24 threads.
Well you're not, at 12 cores loaded you're thermally limited and tests show no difference to speak of at all. You get a slight boost to single threaded performance but overall this is the dullest upgrade in ages and creating a whole new XT line for it is silly. Should have just called it 3910X, 3810X and 3610X, at least then they'd get points for honesty.
Re: (Score:3, Informative)
It's literally the same design with improvements to the printing process so the frequencies could be improved, so the naming seems about right.
This isn't a "new" processor, but the latest batch of the same design with quality improvements. Instead of just tucking it into the same product model they incremented it so people knew that.
Re: (Score:2)
It might be just enough to win a pissing match!!!1!1!!1
Re:100 mhz per core (Score:4, Interesting)
I think only Gamers Nexus got a review sample of the 3600XT [youtube.com] that easily overclocked to a stable 4.6GHz all core at only 1.3V. They also note that IF clock now comfortably goes up to 1900MHz. That is pretty fantastic.
But when they compared the performance differences in real world application with the other models, the results were a bit more sobering.
According to what I've seen XT series is objectively better and outperforms their non T counterparts in most tasks. And it's good that AMD closes in on Intel by showing that they can improve their silicon that way. Intel for example had various years of silicon improvement to arrive where they are today with the 10 series without having made any major change in their architecture. So again, good for AMD. But their pricing for the XT series simply stinks in comparison to how much of a performance improvement is there compared to their considerably cheaper counter parts.
The reviews for the 3900XT are not released yet, but I do not expect any big surprises there.
The issue of the 12 cores 24 threads CPU as far as my own 3900X goes and the data I have from other people is that they use 2 differently binned CCDs. One of those CCDs is the good one, the 3900X one and the other looks like a 3600X one. Normalized over voltage my 3900X bin CCD runs at about 5C cooler than the other while occasionally boosting up to 4.641GHz on some of its cores and 4.591GHz on others according to what HWiNFO64 tells me, which would make it a pretty good sample [youtube.com] I suppose considering that I bought it shortly after it was released. The 2nd CCD however only boosts to up to 4.392GHz. This is all at stock settings.
So the only thing that would really surprise me is if they put two "good" CCDs in the same package this time.
Re: (Score:2)
They have the same prices as the previous X parts they seemingly replace, so I'm not sure what the argument against them boils down to unless people were expecting to replace a previous Zen 2 purchase with one of these. At worst it's a marketing ploy to
Re: (Score:2)
Even AMD's own statements were that there will be only minor increases.
They're not replacements as far as I'm informed and still in production. Even the TFA states that the non T series will stay in the lineup.
Like I already said, they're still a good sign for AMD being able to improve their silicon over time. That is generally good news for everyone (the market needs competition
Re: (Score:2)
It's boost not base so if you're lucky then one of the cores will occasionally hit the advertised boost clock, with the initial release. Der8auer did a survey and most users did not get the advertised boost clock: https://youtu.be/DgSoZAdk_E8 [youtu.be]
I have an R9 3900X, I won the 'silicon lottery' and my CPU at stock setting does boost up to 50mhz faster than advertised, but I do have 280mm water cooling. My guess is these XT CPUs won't be noticeably faster in real world usage.
Re: (Score:2)
100mhz doesn't seem like much until you benchmark it and realise the insane cost for such an incredibly minor increase in performance.
So far the results seem to be unanimous, there's no reason to get these unless you're an overclocker and trying to set a specific AMD record. Otherwise you're better off saving the money and getting the previous X model, or spending a few cents more and getting a previous X model with more cores.
Re: (Score:2)
It might seem like it today. At one point in time the difference between the Athlon XP 1700+ and Athlon XP 2000+ also seemed quite large. Not so much now.
"to battle Intel" Erm, right now the battle is won (Score:2)
AMD is currently "winning". Anything they do now is more like aiming the kick at the downed opponents kidneys, rather than randomly kicking.
Re: (Score:2)
Wha... ? (Score:3)
got burned on an AMD chip with Vega graphics, only to learn that AMD doesn't really support graphics in Linux, confirmed by customer support.
LOLWUT?
Bullcrap.
AMD *is* supporting Linux: they have developers on their payroll writing the drivers in the mainline Linux kernel and Mesa.
(Together with Intel, they are among the few companies that do. Looking and you, Nvidia and nearly all of the embed SoCs manufacturer).
The only way customer support would be right is if they meant: "we do not offer direct customer support for these specific SKUs in our proprietary 'Pro' driver line" (the AMDGPU-Pro, one. The one used to provide full OpenCL support on cl
Re: (Score:2)
Except for driver support for Linux. I recently got burned on an AMD chip with Vega graphics, only to learn that AMD doesn't really support graphics in Linux
Hahaha no. I just bought a laptop with Vega graphics that came with Win10S. Under Windows I was getting frequent crashes from the video driver on mundane operations like resume from suspend. Under Linux I am having no such problems, and the 3D works fine — if anything, it's faster on Linux.
AMD is now supporting Linux better than Windows.
confirmed by customer support.
You clearly didn't get a support tech who know what they were talking about.
This was the last AMD chip I will buy.
Because you did something wrong and it didn't work correctly? Have fun with Intel's ever-gr
Re: (Score:3)
That's absolute bullshit. Vega cards work fantastly on Linux, you only need a recent enough kernel.
Source: I run a Vega 56 over OpenSuse Tumbleweed, and just spent 3 hours playing Doom Eternal (which works amazingly well under Proton).
Re: (Score:2)
Enjoy your hardware vulnerabilities (Score:2)
The problem is that as always AMD "wins" by brute forcing the problem. As soon as Intel produce something it always is way more efficient meaning your PC doesn't have to double as a heating and incredibly loud cooler noise unit.
The problem is that Intel does achieve this by using "clever" tricks and cheats that eventually turn out to have been pretty stupid (cough... Spectre/Metldown... cough...), and require workarounds that cause slow-downs and cancel out any gained speed.
Seems everytime I run "grep . /sys/devices/system/cpu/vulnerabilities/*" the list has gotten even longer on the Intel hardware I have (mostly work-provided), while at the same time it seems to be filled with "Not affected" on my various personal AMD and ARM har
Re: (Score:2)
If the CPU overheats and produces a BSOD it's most likely a PEBCAK in these days. There are tons of safeguards in place that you will have to work around manually to make such an issue a common sight. And while there may be a couple of defunct products where this happens without the user causing it, this is far far from the norm.
On top of that there's plenty of empiric data
There are no FPS disparities in any relevant games where
Re: (Score:2)
I'm not sure if you are serious or just trolling, but just in case...
1) Do you realize that latest AMD chips actually run cooler than Intel's?
2) If your CPU overheats to the point of crashing, then the problem is the cooling system.
3) I have let my Ryzen 2700X run at 100% during two weeks (some math heavy code), and it worked perfectly, not a single problem.
4) Yes, I have played the latest at 60 fps, I just spent 3 hours playing Doom Eternal, and it never went below 60 fps.
Re: (Score:2)
Sounds like you need some better cooling. Admittedly it can be difficult with laptops nowadays, where thin and stylish trumps things like avoiding thermal throttling. But on a desktop you have plenty of options for cooling. Sure, I like a quiet desktop too, but that comes after making sure the cooling solution is sufficient that it can run at 100% CPU indefinitely.
The Cycle (Score:2, Insightful)
AMD periodically massively leapfrogs Intel in performance and price, then over the course of some years Intel catches back up, then passes AMD, then a few years later AMD yet again does the leap.
Re: (Score:2)
Re: (Score:2)
AMD isn't even in the manufacturing process race, Intel fabs are running against TSMC fabs, which manufactures AMD's chips.
TSMC's manufacturing process gets funded by all of their customers. It's AMD, Apple, Broadcom, NVIDIA, Qualcomm, etc combined up against Intel's fabs.
Intel is a much bigger company than AMD, but they are only about three times the size of TSMC. And Intel does a lot more than just manufacturing, so Intel's fabs might be significantly smaller than TSMC.
Intel needs to not just keep up, but
The Chinese might help Intel out (Score:3)
With a few of those thousands of missiles they have pointed at Taiwan.
Taiwan has done really well as a very small country to dominate this space. But they are vulnerable.
Re: (Score:2)
China won't attack Taiwan. They don't want to screw up that golden goose.
Anyway they like AMD, they licence their CPU designs and produce domestic versions with the American RNGs ripped out and probably some other unnecessary stuff too. They use an inferior process so they don't run as fast but for government use they are fine, you don't need a Threadripper to run a word processor.
Re: (Score:2)
There was a time when China was very dependent upon Taiwan (and Hong Kong) for investment and expertise that drove their tremendous growth. However, that time has past, and China no longer needs Taiwan's economic input. They like it, but it would not make much difference to China if it disappeared. Tsai knows this, and Taiwan has been withdrawing from China as fast as they can.
I think it unlikely that China would attack Taiwan because it would be a declaration of war against the rest of the world. Even
Re: (Score:2)
Cheers to competition and progress!
It's what we want.
Hard pass thanks. (Score:3)
A 2-5% performance increase is not worth the 11% cost increase (3600X), 20% cost increase (3800X), or 18% cost increase (3900X). It's far better to spend $20 more getting a 3900X than a 3800XT except for a few very VERY specific workloads.
The Zen2 series is incredible, but this move makes little sense for AMD and even less sense for consumers. Eagerly awaiting Zen3 later this year which will make the XT series here make even less sense.
Re: (Score:2)
I don't think you understand how it works. Per core performance is very different than #cores and does not have a linear cost. You always pay multiple times the performance increase when you get to the top performing core. And it makes sense for many cases - i.e. in most environments I've worked in there are always some tasks you cannot further parallelize, for which per core performance is everything. I remember at some point we paid twice the price for a 5% per core increase on a Xeon CPU (Intel really go
Re: (Score:2)
I don't think you understand how it works.
Percent improvement vs percent cost increase doesn't care how many cores you have. A chip that is 5% faster on all cores is 5% faster regardless of how many cores it has, and the results are painfully obvious, unless you have a VERY specific workload the money is better spent on more cores.
i.e. in most environments I've worked in
So you've worked in a VERY specific environment. Good, buy the 3800XT or whatever. The fact remains it's not a chip that makes sense for most workloads.
I remember at some point we paid twice the price for a 5% per core increase on a Xeon CPU
Xeon CPUs by definition have always been for very specific workloads.
Re: (Score:2)
To expand on my reply: If you have a workload where your profit depends on single core performance you wouldn't be buying AMD given you can get faster single core performance from Intel for a fraction of the cost. E.g. An i5-10600K outperforms the 3900XT on single core workloads and costs literally half.