AMD Responds to Intel's 9th Gen Benchmarks
tomshardware.com
tomshardware.com
In my developer workflows that include quite a lot of browser tabs, a bunch of Docker containers running, occasional Go or Docker build I don't get any of the CPUs at 100% often. Only when I do cross compilation to a multiple architectures I do use 6-9 CPUs and even then my machine is totally usable and I can't even feel that in the background lots of cores are at full utilisation.
What really counts is the amount of CPUs available, I don't really care about their maximum frequency. Comments about gaming are valid but I think instead of buying Intel just go with a better GPU and you will win there as well.
I am by no means a fan of the price/value ratio for new i9 CPU and any shady marketing cheap tricks generally, but I don't get this 'Ryzen works for me so it should for everybody' all over the tech net these days.
Use whatever you feel is the best tool for your needs and your budget. Respect other's choices, even if it means that for some people, this new i9 will be the best/preferred option. Or any other option for that matter that isn't yours.
I wouldn't recommend it to most of my friends or my family for the same reason you mentioned, none of them will utilize those cores.
It just works really well for me and it wildly surpassed my own expectations as I was quite used to the fact that my machine wasn't really usable while I run Kubernetes locally. My workflow used to be:
1. write some code 2. start Kubernetes, wait for it to boot and test the code that I wrote 3. tear it down 4. write some more code 5. repeat
I would be loosing quite a lot of time. Also, while waiting for it to start, I would usually get distracted and start doing something else. Ryzen changed this workflow and boosted my productivity. Now I don't even notice whether k8s is running or not, sometimes I forget it for days :D I can't stress enough how important it is to have a lag free development environment.
With an evga RTX2080 in the home PC it murders the games I play (elite dangerous mostly) at 4K.
This doesn't feel at all like what the OP is saying. I know it's common on the internet to ignore "right tool for the right job" but OP managed to nail the point. Each one is right for a different job. His happens to be more suited for AMD.
Oh, and I'm about to upgrade the networking (from 4 x 1 Gbps) but I'm on the fence between 10 Gbps and Infiniband.
I never hit the limits of these machines, so obviously this configuration should work just fine for everyone. :-)
https://www.datacenterknowledge.com/archives/2012/03/23/too-...
This hasn't been true for quite a while in many games. Modern engines often do utilize 6-8 cores (they have to to get it to work well on the anemic CPUs in ps4/xb1). Assassin's creed, for instance, struggles to push 60 fps on Ryzen 1700X[1]. Of course, one can always drop settings to slightly above console quality and enjoy 60 fps. But, the idea that GPUs are almost always the bottleneck in games is just a myth.
[1] https://www.youtube.com/watch?v=kG9Bsv-d8HM&feature=youtu.be...
AMD benefits from having their chips in the consoles...which now have an install base of around 120M+ units...
http://www.legitreviews.com/intel-core-i7-7700k-versus-amd-r...
PS: On high they are both fine.
- Nvidia has been killing it in graphics card design relative to the competition. Consider that a GTX 1070, released 30 months ago, can go all the way to 1440p60 comfortably.
- Nearly every game has to be playable on consoles, and consoles are all running AMD's last generation chipset.
- 4K is still the "storm on the horizon" that no one's graphics cards are truly ready for in an affordable way.
All of that combined really means that raw graphics quality in games hasn't improved a lot in the past four years.
So, most games will continue to show increased performance the more clock you throw at a processor. Combine that with more and more games optimizing for more cores given the prevalence of 6-8 cores chips on the console market and in gaming PCs, and this should make sense to anyone.
It should be said: This is why cloud gaming is quickly becoming a real thing. Beyond connection speed, graphics cards were always a huge bottleneck in making that cost affordable to cloud providers. Today, the graphics card isn't a bottleneck, and all the cloud providers have an abundance of them in their data centers for AI-focused work. And if neither the cards nor the connection speed could handle 4K anyway...
I wouldn't call it a "storm on the horizon". I am fortunate to have a GTX 1070. If I turn off anti-aliasing, 4k at 60hz seems to work perfectly well in just about every game I've played. (Older steam games like Deus Ex:HR and Dishonored, HotS, WoW, Mechwarrior...). Most things are built with antialiasing at 1080p in mind, and running it at 4k (without FSAA) seems to be similar workload.
[1] Battlefield 1 - 97fps [2] Far Cry 5 - 70fps [3] Grand Theft Auto 5 - 62fps [4] Total War: Warhammer 2 - 55fps
And the 99th percentile of many of those titles drops below 60fps. The FE can push a little bit harder.
Its true that at 4k you can pretty safely turn off the more powerful forms of AA and gain a lot of performance back; the resolution is so high that it doesn't tend to matter. How else would the X1X be capable of so consistently reaching 4k30 for titles like FH4 with a substantially less powerful graphics card?
But that's not really how we observe what these cards are capable of. You can play CS:GO at 4k120 on a much less powerful card, but that's not really that interesting.
Also, I consider the "affordable" mark to be ~$300, which is GTX 1060 territory. So when you consider that most gamers [5] are barely running 1080p144/1440p60 right now, that jump to 4k60 is actually pretty far.
[1] https://www.anandtech.com/show/13346/the-nvidia-geforce-rtx-...
[2] https://www.anandtech.com/show/13346/the-nvidia-geforce-rtx-...
[3] https://www.anandtech.com/show/13346/the-nvidia-geforce-rtx-...
[4] https://www.anandtech.com/show/13346/the-nvidia-geforce-rtx-...
[5] https://store.steampowered.com/hwsurvey/Steam-Hardware-Softw... (Far more than 40% of gamers on Steam are running a card less powerful than the GTX 1070).
I think you may be looking at what most benchmarks show, which is often high/ultra details, with AA. Yes, you can compare cards better by putting them through their paces, and showing that card A handles a heavy load better than B, but most games look amazing _right now_ on lesser settings at 4k, on the 1070, even if it's certainly eclipsed by better cards. (I also agree with you on the affordability front.) Just about any game that came out before last year will look amazing, and I've been able to run pretty much every game that I have at high-ish (other than antialiasing) settings quite happily.
You're absolutely right that $300 is a lot better target than $400, though.
On a Ryzen 5 2600X with a Vega 64 at 1440p, most games I've seen run at "Epic" or "very high" detail at 75fps.
I had the 1700, 1800X and now a 2700X. For most people that want a desktop and really utilize it in a way where it's just shared computer for general home usage, I'm recommending Intel's NUC lineup. When willing to build, for productivity and gaming I'm recommending Ryzen, and while the 9900K is going to be at the top of some charts, the combination of outdated Skylake architecture (plenty of examples like this boost example), high price that borders on Threadripper pricing, and increasing the die thickness to nullify a lot of solder's advantages[1]. Publications are starting to add all this up and naming the 2700X the best gaming CPU[2]. For a long time I would've said Intel but I'd have to agree with their findings in their entirety, I also really like the i5-8400.
[0]https://i.imgur.com/IPhCh1G.jpg
Of course such marketing benchmarking HAS to be done at 1080p, because anything more challenging and you're entirely gpu bound with cpu differences becoming margin of error fractions of a percent.
Also, with variable refresh rate, reaching exactly the maximum refresh rate of the monitor isn't very important.
The comment chain also referred to reaching "more than 60 frames on higher than 1080p resolution", there are luckily more options than 4K.
Citation please.
* You don't really want a bigger monitor than 24" for competitive shooters, because then you can't see everything at once. There is only one 1440p monitor in that size that I know of and no 4k monitors that I know of
* You want a monitor with 144Hz (or better!) and low input latency. 1440p and 4k monitors that do this are rare or non existent
* Easier to stream
* Overall appearance of a game at 1080p with more anti aliasing vs 4k with less isn't that difference. text being an exception of course.
the other issue is input latency which is hard to find specs for. I have an old dell 1080p monitor that is a total outlier on that end and I never want to give it up!
but yeah, with more 24" options, higher refresh rates, and the latest nvidia cards 4k, 144hz gaming is going to be viable soon.
If you want to achieve 144 fps (constantly) you often have to lower graphics to levels where you're CPU bound.
Reviewers test at 1080p because that effectively removes the GPU from the equation, at higher resolutions it's mostly all GPU limited.
AMD will be unveiling their Ryzen 2 processors/GPUs in January and it is rumored that there will be a 15% IPC increase. If that is the case, AMD will be on-par with Intel. It is also rumored they will be bumping the Ryzen lineup to have 16 cores.
I've been running a Ryzen system w/ a 1800x @ 4GHz on all cores for well over a year now and have no complaints whatsoever. As a developer/light gamer it was the perfect balance of price to performance.
I also read that there are still some controversy about the CPU frequency for multicore and also that reviewers got speical CPUs, you are not guaranteed to get 5GHz on Intel so it is possible that in 1 month the new CPUs will have not as good turbo.
My point is not to rush a purchase based on this initial benchmarks, wait a few more weeks.
Does this mean that is guarantedd that 1 core will reach 5GHz ? Are there exceptions? Will it work it regular coolers ? Will it turbo without BIOS changes so in default config?
Thanks if you can clear this, because I am not sure if thix Max is considering power users that overclock or not
And it OCs to 5.1 on all cores.
It's quite likely that a normal air-moving fan + heatpipe radiator will not let you do this for more than a few minutes.
1. Your power supply can supply the 200W needed.
2. Your motherboard's VRMs (voltage regulator modules) can supply the 200W needed.
3. Your CPU coolers can extract 200W of heat.
Its all running outside of Intel's 95W TDP specification, but it seems like Intel is comfortable with claiming 5GHz on 2-cores of the i9-9900k if you have the appropriate PSU / Mobo / Coolers.
IMO 9900K is awesome if you need best in class single-thread performance for one task and good multithread performance for another task. It would be good for me, for example. I like to play WoW and this game is pretty much single-core limited to provide even 60 FPS. And I'm using my PC for development, so improvements for compilation time are nice as well. Whether I would want to pay that price is another matter, I definitely won't buy new Intel CPU until they fix those Meltdown issues causing significant slowdown in some tasks. And if AMD will reach 5+ GHz by that time, I'll happily switch without second thought.
In programming community Intel also receives more skepticism because their compiler intentionally cripples AMD users and the disclaimer is put on the web as an image so is not picked up by search engines(last time I checked a few months back was still image).
It has with SSDs.
https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9...
Fully loading an i9-9900K will consume 170W. Note that this is POV-Ray, a ray tracer, which can be reasonably defined as real-world application.
When stress tests are performed, like the following:
https://www.tomshardware.com/reviews/stress-test-cpu-pc-guid...
the peak is 205W, which is extremely high (when compared to the touted 95W TDP).
AMD is actually transparent on the ThreadRipper power requirements - the reported TDP is 180W, and stress tests confirm that value for several models. In this respect, Intel is arguably cheating. I wonder if AMD could get an advantage (marketing and benchmark-wise), by declaring a low TDP of 95W, while having an effective 200+ one.