Ryzen 7 1800X Review! What's The REAL Deal With Gaming?
HTML-код
- Опубликовано: 11 сен 2024
- Rich presents his analysis on Ryzen 7's gaming credentials. It's not as competitive to Intel as we hoped, but there are signs that Ryzen could come into its own a little further on. Let's just say that the 1600X should be fascinating...
Subscribe for more Digital Foundry: bit.ly/DFSubscribe
The frame time graph is often noticeably smoother with Ryzen , that's so much more important than an extra 10fps when you are already at 100. Everyone is missing the important bits.
I think people will be disappointed by these predictions that quad core processors will be the retrospective better pick in five years. People keep bringing up the FX series, but this situation is far closer to the Core2 Quad vs. the E8000-series Core2 Duos. The people who picked the Duos retired the systems years ago while the Q6600 can still function as a low end gaming system. Intel needs to move forward, too, and they know it. It's why they eliminated the difference between the Pentium and i3 by giving the Pentium Hyperthreading.
NO MA'AM
I'm still running my core2quad q9650, 8GB of 5-4-4-12 ddr2 800 RAM, 4GB gtx760. Still a decent gamer after MANY years of only updating the graphics card a few times. I did just have to replace the mobo battery though, never kept a computer that long before lol.
mmmm.. How much £$$£'s for you to show us *your* important bits? ;)
Same here Q9550@ 3.6Ghz 8Gigs of DDR2-1066 pushing data to an R9 290X. Plays great especially in GPU limited scenarios.
ok but 1800x costs a lot in general and a lot more than 7700k
I don't know if Richard understands that he's actually advocating R7 over I7. Which I agree with for a few reasons:
1: Why do people prefer higher fps instead of lower fps? Smoothness. Which one provided a smoother gameplay? Ryzen. So then Richard does recommend Ryzen for high refresh rate gaming right?
2: No guarantee that AMD will fix issues? Well, they have given us their word, and all logic points at that being very true from various sources. But we have something guaranteed: 7700k can't give any more than it is doing today. It is being fully utilized. As CPU demand goes up 7700k will deliver lower fps. Ryzen, on the other hand, will more than likely deliver more as it has plenty more left to be utilized. This isn't a bulldozer. It's 8 real independent cores.
3: People buy new components to play the coming games, not the ones they have already played. Those benchmarks say little to nothing about how future games will run.
4: Look at the processors in current consoles? 2 CCX of 4 cores each. Reminds you of anything? AMD is making a very strategic comeback. Pay attention, and stop being so shortsighted.
I can't believe that as a gamer for just $329 I can buy a 8 core/16 thread CPU today. This thing will last far longer than even the next 4c/8t from Intel for pure gaming, not to mention being able to run more background applications which allot of gamers actually do.
Should have shown CPU usage while running the benchmarks.
Rnl Valen 6900k has same cpu usage as Ryzen. 7700k has higher cpu usage due to less core.
heres your answer the 7700k@5ghz still hits 100% with a gtx 1080 in some titles
The Zen Architecture is out for some weeks now, while Intel's had almost a dacade refinement. Ofcourse Developers will have to adapt. It will be the same when Intel comes out with an all new Architecture.
Guys, relax.
I get what you're saying but as mentioned in the video, that means all back catalog games will never see a patch or fix for any issues brought on because of their new architecture. Instead, AMD should have designed an architecture that wouldn't need to be "programmed" for. That should be up to AMD and Microsoft when it comes to their OS and how it handles managing the CPU and it's cores/threads per application.
Also, I'm a huge AMD fanboy but I'm not biased. If AMD fucks up, I'll call them out on it. I think the Ryzen CPU's are genius but AMD needs to get on the ball and really hunker down on the problems that it has so far.
rwells242 I already mentioned that they need to be able to handle all current programs properly even if it's a new micro architecture. What's the point in releasing something new if it's not compatible with the most recent games and applications? AMD isn't dumb, they didn't design this micro architecture just so developers could program properly for it. If anything they need to fix it to work properly with any program on the market. It shouldn't be up to developers to program for the CPU specifically, they should be able to program properly and the CPU should be able to handle it properly, while I think developers need to make heavy use multi threading, they shouldn't need to design a game around a specific CPU. Crysis 3 is one of the only games that's programmed with intense threading support which is why it works so well on Ryzen. Other games such as GTA V are also heavy threaded applications but most of the instructions sent to the CPU are of the kind that need a physical core to process the commands, Crysis 3 splits this up a lot better but that game wasn't designed with the Ryzen chips in mind. So it's down to AMD and Microsoft to properly support the Ryzen chips on their platform.
Because of this, this is essentially the same problem AMD was having with their FX CPU's and the same excuse. FX CPU's were really good at multi threaded applications but lacked really hard in the single core performance department. That's essentially the same problem they're having now but the Ryzen chips have much better single core performance over their last CPU generation. We never saw developers aside from Crytek make use of heavily multi threading in their applications which would have essentially been programming with the FX CPU's in mind.
The Ryzen chips are already powerful for their price so they're already great CPU's but they could be better with more support from Microsoft and AMD.
People say Ryzen is not good for gaming and then the same people say that 2600k is still great for gaming nowadays.
Logic 404
Martysoft 2600k is a beast. whats wrong with it
Nothing wrong with a 2600k... but the Ryzen R7's are superior in every way.
Roasted 6900k is perfect for gaming. Didnt you see that it beats the i7 7700k in all cpu intensive games?
Yeah, if you bought the 2600K 6 years ago. All Ryzen gaming benchmarks are barely ahead of 2600K. *Ryzen CPU's in 6 years are gonna be shit.*
No one is recommending you to by a 2600k in 2017, ffs. They just mean that if you have it there is no reason to upgrade, because 500-600 dollars for a few FPS more is not worth it.
I know this has nothing to do with the video, but I wanna thank you Rich and all the Digital Foundry team members for all the passion and devotion that you guys put in every single video. Thank you!
Everyone is bashing amd for how it performs in games but no one thinks about why that is... well for last 6-7 years most games were optimized with intel in mind, here we have brand new cpu. When Batman:AK run like shit on intel and nvidia people said it's "optimization" when games run bad on ryzen people say it's shit cpu when benchmarks on production level shows that this cpu is amazing for what it cost.
Let's be honest last intel's cpu are laughable if you have i7 4xxx you don't need to upgrade it because all "new" intel's cpu are just worthless... it's that funny when 4th gen is not worth upgrading to 7th gen when nvidia with 1 gen difference shows how big jump can be (between 9xx and 10xx).
I use intel and nvidia but I'm not a fanboy or loyalist to billion dollar company. Competition is the most important part for consumers and it's driving force for progress.. for last 5 years we had monopoly and it hits our wallets and tech progress.
Someone with a brain in the youtube comment section.
well AMD stated an optimisation because they made a 4+4 CPU cores and not just a single one like Intel achieved. that's why it's cheaper and no I won't do like other people said. buy rx480 because games will be more optimized for that GPU. well the RX 580 will replace that card with faster performance at the same price . people has to understand that if you want to game right now get the best CPU that actually performs better. if one day AMD runs 80% of the games faster I will get one but since then there's no reason buying one just for the future but think as for today not for tomorrow
Neurotyczny Kot g4560》ryzen
Stop damage controlling! AMD messed up again. I want to run Emulators, and AMD is useless in that department!
Such a low IPC is a sad joke.
Don't forget to mention that the 1800X is literally just an overclocked version of the 1700X.
I don't understand how you can call a quad core more future proof than an 8 core in 2017.
Because core count is not everything. FX 8350 has already proven this.
FX 8350 had a severe lack of per thread performance though. It also wasn't really a "true 8 core".
The reason why 2500K is relevant today is because IPC per thread performance improvement in CPUs has slowed down to a crawl. Intel's last 2 architectures have had 2.5% IPC improvement. Kaby Lake's IPC is only about 20% better than 2500K. So if game developers want to utilize CPU more in future games they have but one choice, and that's to utilize more cores.
Today's games peg 7700K to near 100% in these test scenarios reviewers use (on all threads). 1800X on the other hand is about 50% utilized. Yet somehow it's the less future proof of the two. According to DigitalFoundry.
It's the same story. No matter how many threads there are, they don't matter if they aren't going to be used. Even with dx12 (which still isn't even standardized on launch titles) there are no benefits at all passed 6 threads. Yes, look it up. By the time games even begin to really utilize 16 threads, the 1800xs IPC will lack in comparison to what's out there.
@club4ghz Wasn't the Fx 8150 released in 2011?
hussein hussein Yeah and to this day that still holds true, core count is not the be all end all for games.
Just want to leave something here.
There is a problem with the L3 Cache on Ryzen CPUs :) The Games Basically switch Threads very often, or windows does it i dunno, and that forces the CCXs to communicate way too much with their L3 Cache. The Latency which this issue brings to the table is often 200% greater than on Intels side! :)
Windows/AMD confirmed that there will be an Update which adresses this issue, and that makes the Scheduler to not switch Cores/Threads that often, which will lead to lower Latency in the L3-Cache which will result in better Performance in Games (because they are heavily Cache consuming).
So let the Platform/Microcode/Games mature a bit :) The last 8 Years, Games were primarily made for/on Intel CPUs which have a complete different architecture, so don't say something like "The really bad Gaming performance" or "This CHip is soo much slower in Games" etc. when you can see the actual potential in synthetic tests.
Also, bear in mind that Intels first 14nm CPU was not able, or at least it was very difficult, to achieve 4GHz on 8 Cores/16 Threads! It is actually very impressive that the first 14nm chip AMD produced, is able to hit that Frequenzy that often.
This. The very reason I will not go for Ryzen when I'm upgrading from Haswell i7 later this year or early next year.
To be honest, you have a pretty good CPU already MadFinnTech. You should probably wait for zen2 or Cannonlake at this point.
Disobeyedtoast
Im waiting for x299 platform thats my upgrade
hehe, let him buy intel :P who cares :)
BeaterBurito I thought Ryzen was very good with its L3 cache speed, that's weird. Also Broadwell E (first Intel 14nm 8 core CPU) could reach past 4GHz very easily compared to Ryzen, so don't spread bullshit on that regard, Ryzen sucks for overclocking in general.
I feel like hardware makers try to fit as many Xs into their names without it sounding weird
The best I've heard about must me the XFX Radeon RX 480 XXX Edition. Grand totale of 6 X's.
NO MA'AM Well, given the name there should be ;-)
You also have to take in count that ryzen is an entirely new architecture and has NOT been optimized at all. Ryzen 7 power will be most likely shown in the long run when games take advantage of 8 cores and 16t
Yeah, and when this is the case, you could buy a new generation of CPU's which is faster than the 1800X... So what's the deal?
Why should you buy a 8 core / 16 threads for gaming NOW, when you can't use all of its power because most games are not optimized for 8 cores / 16 threads?
If I see the real gaming performance of the 1800X in like 3 years, I could wait and buy a better 8 Core CPU then which will beat the 1800X
Thats what they said for fx series. This cpu will perform even better when games start to use 6-8 cores! When games start to use 8cores ryzen will be weak... benchmarks vs intels 4core shows it
Daxter we can hear it since FX 6xxx/8xxx processors were out
AtgreatR9 fx series was just shit in it self. But I mean look at bf1, it LITTERALLY maxes out the 7700k. And now since even consoles have 8 cores its the time to utilize more cores , hell we have been stuck with 4 cores for like FOREVER. Also Intel has been VERY anti consumer in the last years since they had a monopoly for like 4-5 years
AtgreatR9 completely different circumstances. the fx series 8 cores were just glorified quad cores. ryzen are true octa cores
Meanwhile in DF PC benchmark tests: - 140fps
Meanwhile in DF console benchmark tests: - 30fps
2017 is the year of AMD, RYZEN VEGA SCORPIO.
Cabron loco
Low quality bait.
Cabron loco $699 GPU vs a $399 console. Nice comparison nub.
Scorpion wont have ryzen, at best a very high clocked puma cpu.
Cabron loco of course it can't but MS never told consumers that they'll do 4k60 they just said 4k...... It will do 1440p60 for sure....
Scorpio? haha
8:30 "Fundamentally all we are doing here is Influencing CPU scores with GPU limits." I've been through more than 30 Ryzen reviews and no one managed to said it that clear and simple, great review!
+DigitalFoundry, Richard, as you know the AMD Zen architecture has 2 CCXes each of them house 4 cores, L2 and L3 cache. And the connection that is called "Infinity Fabric" to communicate between the CCXes, is directly link to the DDR4 memory speed. The communicating |frequency| speed of the "Infinity Fabric" is 50% of the |frequency| speed of the DDR4 dimms. It makes sense that faster DDR4 frequency speeds makes the CPU calculate faster and show better performance.
But the connection of the "Infinity Fabric" is not fast enough to do calculation fast enough between the CCXes, which cause some of what delay and some what low performance. People say that the speed is about 20GB/s to 50GB/s. Which is on the slow part when you are doing CPU calculation.
One way to not loose performance between the CCXes, is that you force the game to run on only 1 CCX, and force all other windows, and windows background process on the second CCX. Now you won't have cross communication between the CCXes , and the game will get full 8 threads compute power. This will translate in to stable FPS and less frame times spikes. Richard i hope will look in to this, and a program called "Procces Lasso" can assist you in getting this tested.
In multi threaded programming,constantly moving data between your threads is unorthodox, and most likely you will end up with a code slower than its single threaded version. There is a chance of sharing cache, but locality counts always, even in single threaded programming. So no, I do not think inter CCX communication is the bottleneck here, it almost seems like most games are single threaded with a bunch of extra worker threads pushing data to a main thread, that is not scalable, instead each thread should push its data to GPU, which is possible with Vulkan/Dx12, before that every OpenGL context had to be in a single thread, not thread safe as we call it.
+utkua
Have you seen zolkorn's review where he tests a Ryzen cpu with 1 CCX disabled clocked at 4Ghz and an i7 7700k also clocked at 4Ghz, the difference between them is only 3-9% which is in line with the benchmarks that ryzen does well on.
www.zolkorn.com/reviews/amd-ryzen-7-1800x-vs-intel-core-i7-7700k-mhz-by-mhz-core-by-core/3/
+i would also recommend checking out this french websites benchmark where they compare a 4core+0core and 2core+2core one to check if the latency between the CCX's is really whats causing performance issues in certain scenarios
www.hardware.fr/articles/956-24/retour-sous-systeme-memoire-suite.html
as you can see the latency between the CCX's does seem play a significant role.
Never heard of zolkorn, his review is in Indonesian? I can't read that, and google translate is not doing a well job. So sorry i didn't. And the second review is in french, and again never heard of them. I only follow reviewers that speak English and have a good reputation.
well you should be able to understand those graphs even if you don't speak english.
For example in the french bench battlefield 1 shows the biggest performance difference of 20% between 4+0 and 2+2 configuration (in 4+0's favor), while 2+2 is actually favored by 3% in 7-zip due to the 16 MB of L3 cache
That might be because remaining cache is shared between the active cores, so more cache per core. If their "critical thread" experience less cache misses, the game might get a little performance boost. However, it is still a very little gain for a significant computing power loss.
Now thats what I call an extensive review, great work DF!
I am glad that you are taking the pragmatic approach and have promised further testing.
The performance picture in gaming is certainly a victim of its own immaturity, particularly with DDR4 speeds etc. While The Idea that Titan XP at 1080p puts stress on a CPU is generally true, It overlooks the fact that it is also stressing the inter chip memory and PCIE communication bandwidth that supports the CPU+GPU load being applied to the hardware. It also overlooks that there may be an issue with the Nvidia Drivers themselves.
Turning off SMT, because there are less threads accessing memory over the Data Fabric, the contention that is being observed.That is a bandwidth restriction. CCX thread switching and scheduling, sit on top of the Data Fabric and the contention is causing the thread switching issues, making the scheduler a symptom, not the cause.
The Issue that I have with the methodology, is that the assumption of only being CPU stress has the entire internet putting blinkers on and obsessing over the windows scheduler. None of the Big channels so far have tested what happens with SMT if you use a 1060 or 1070 for example. Do you get the same results? The answer is no you don't. SMT On performs better than SMT off all the time. Is 1080p performance lagging behind with a 1060? No, it is quite comparable to the Intel platforms.
You make a statement that bios updates are coming thick and fast and memory speeds are being improved almost daily. Very fair call and it communicates that things are likely to change. It is not making a definative blanket staement that Ryzen is terrible and slow - end of story. However, Because of the unwavering belief belief in the dogma of isolating the CPU. Blanket pronouncements are being made that Ryzen is terrible for gaming. Full Stop. End of story. The end result is the herd puts down tools, and all rush to blame windows scheduler and CCX thread switching as the root cause with only half the facts and too many assumptions at hand to form a completely valid opinion.
I know digital foundry does not use firestrike as a metric in reviews, but I have yet to see any reviewer who does use it, make any mention of the combined score metric or compare the disparities of combined with graphics and physics levels compared to Intel platforms. The combined score is the only result from Firestrike that provides a metric that measures relative GAMING performance! Yet they are still publishing an opinion that states Ryzen is poor as a gaming platform. All it really does is totally invalidate that review as being meaningless.
The media as a whole has pronounced that Ryzen has poor gaming performance. That is a pretty definitive and final call that in reality is effectively based on a statistical sample of one. In actual fact, the Mainstream Tech media in general have stumbled into the perfect storm of immature system and so much graphics horsepower the immature system doesn't know what to do with itself. The test methodology itself has inadvertently found itself stuck in a similar position to an out of spec benchmark result. The result is an anomaly that a professional reviewer should otherwise discard or decide it is worthy of additional research before forming an opinion.
I am certainly not trying to excuse AMD or be an apologist but everyone wins if AMD can succeed. Making blanked statements that are untrue 99% of the time out of ignorance is not the way to keep this industry healthy. The 3d gaming performance on a Ryzen 8 core CPU has issues if you use a Titan XP/1080TI/1080 at 1080p in March 2017. Yes, the performance issues need to be investigated and resolved if possible as the memory/PCIe/CCX shortfalls can impact other areas that disadvantage the environment as well, but problem resolution can only happen if they actually look at the actual problem.
Rant over. I just hope that maybe this can inspire thoughtful reviews rather than relegating some members of the tech media and the industry as a whole to a similar levels of respect that is reserved for hacking script kiddies.
yeah we better compare them when there its 10000 mhz ram memory, that its what its the problem here, its not fer like this intel running only at same ram frequency
I would have thought that it was better to find a way to make the interconnects to the memory and PCIe bus from the data fabric that we have more efficient by actually doing some physical investigation.
Unfortunately, It seems that no-one is interested in doing that. Much better to quote psuedo science and blame all these things like windows scheduling and Thread switching that cannot possibly be the root cause of the issue.
We're here for the Wichard analysis!
(Which is always excellent indepth analysis)
AMD Ryzen is relatively a new architecture compared to at least 8-9 years old Core ix architecture. AMD and MB partners should work hard to solve problems soon. I don't think AMD is this slower. There should be another problem related to motherboards or chipsets.
Unnamed Internet Citizen Absolutely.
I was starting to worry about the lack of Ryzen coverage here. Welcome back!!
there will be a real shitstorm in the comments below
kosasmerty You can smell it from a mile away lol
I don't subscribe to many channels, but you've certainly earned yourself one with this video. I can't tell you how much I appreciate how *thorough* it is. Pretty much every other vid I've seen I can say "yeah but." I appreciate you taking the time to test as many variables as you did in a relatively short time.
THAT DOESN'T TEST LONGEVITY YOUR METHODOLOGY IS FLAWED. If you want to test longevity and not optimization then you put the system under load and see if it can complete tasks at 100% usage. What you have done is tested how well a game is optimized at the low end for a specific CPU. That is not testing longevity.
I think his argument wasn't that removing the GPU as a bottleneck is bad, not at all, rather that since current games don't load Ryzen cores overall nearly as much as the i7's (clearly seen in the benchmarks on the net), in the long term it should have more headroom for games.
I still think that's in part wishful thinking (even though I share it for now), since it's acknowledged by nearly everyone that utilizing more cores optimally and fully is damn hard in games, so they don't scale nearly as well with more cores, at least in the short term. In the long term, the trend is clearly that games are utilizing multiple cores better (at least up to 4 cores, and lately even 8), evidenced partly by how much the relative gaming performance of the AMD FX series improved over the recent years, a processor that has abysmal singlethreaded performance compared to its direct competition from Intel, but always fared quite well in multithreaded tests. Furthermore current consoles use many slower cores, instead of less but faster ones, so they do provide a good incentive for developers to move in that direction, which could, again, very well benefit Ryzen a lot in the future.
What an insanely in-depth video for an overview. You really took every scenario into account. I'm liking this channel more and more with each video!
Ryzen is also future-proof. Much more than 7700k. It's nothing like Crapdozer was back then. Ryzen has real 8 cores 16 threads, and it beats shit out of 6900k in multi-threading.
Look at CPU load. On ryzen it's around 50% on all cores while 7700k loaded 100% all cores. Ryzen is not utilized by modern games. Almost zero difference between 6 and 8 cores also shows that AM4 platform is still in RAW state.
Remember how 2600k was useless for gaming in times of Sandy Bridge? People been disabling HyperThreading to get higher framerate. Now 2600k leaving 2500k behind in gaming.
Also DF missed in their video that Ryzen has better frametimes and tighter minimum framerate. Like when you comparing i5 and i7.
90-120fps and people are complaining. I remember times where stable 30-40 1024x768 was a dream. Won´t mention the graphic difference.
A proud 2600k owner here. Glad I forked out the few extra € to have HT. OC'ed at 4.2 only and still not seeing bottlenecks in games, so no need to upgrade anytime soon.
SauS Gaming Shizzle Channel yeah ! I bought an i5 4690k @4.9ghz but I get bottlenecked so I got a ryzen
Yeah...there's absolutely no way your 4690k @ 4.9GHz would bottleneck. It's almost as fast as Ryzen is out of the box.
it bottleneck on games like bf1 I play at 144hz
Hey you might wanna look at how far that cpu is comparing to current generation, minimum 50 percent slower on all fronts.
50% slower dude, impossible
I agree about using lower resolutions to test - but this assumes one very important thing when considering CPU longevity; it assumes that all games in the future will treat CPUs exactly the same as they do today. Today, IPC per core and frequency are the BIG predominant forces. Tomorrow, this may shift towards core count being more important (particularly with new low level APIs), in which case the quadcore i7's won't have as long legs as it does today.
Intel on suicide watch.
stop replying you got roasted
Gaming market vs servers and workstations.
Nah, Intel on suicide watch. It's the latter category which makes more money and AMD's Ryzen dominates there.
STOP
WATCHING
Dex4Sure yeh they also will force u to buy a new socket...
What about good old compiler optimization, and the utter lack thereof? LLVM has had a small patch that switched Ryzen to the Haswell profile, from the Bobcat one, and just that results into a 5-10% performance increase with no extra tweaking whatsoever.
fps per $ ? AMD wins. Good enough for me
I agree. People are here bitching how Ryzen fails in high end gaming, yet none of these whiny cunts have the money to afford such a system. They're literally complaining how Ryzen fails to outperform the 7700K *in a situation they will never experience*.
Ryzen owns Intel in the production benchmarks, in both single and multithreaded applications. People in the IT and data management sectors are fucking cheering for Ryzen right now. For people who work with tons of data and have to efficiently manage all of that, Ryzen is a godsend. Affordable and on par with a 1,000USD cpu.
STR33TSofJUST1C3 I mean, if you work for a company that deals with data management, you'd probably have a massive budget for your servers anyway. I.e we have dual Xeon 2695s running our setup lol
Yep, the performance is so close, that paying double with intel would be just stupidity. Finally get to work with a AMD based build, it sure has been a long time..
imgur.com/a/G2XK1
or does it?
>i3 7350k on top ... No, just no
you guys know what you are doing, you do a great job isolating the CPU and testing against it and not fuzzing results. great job.
RYZEN CPU USAGE DURING GAMING IS LESS THAN 50%
My ancient fx 8350 uses less than 40 percent in gta5 and witcher and less than 20 percent in most other games. This has always been the case with 8 core cpus.
Because games don't use that many cores... do you even logic
my i5 2500k has been running at 4.0 for 4 years now. zero plans to get rid of it.
ryzen is for engineers like me, not for kids and their computer games
"Users shouldn't have to go into the bios and tweak settings" Has it really been that long since a new CPU architecture came out? This stuff always happens when a new one is released. If you're that concerned about it then wait for a few months while the bugs get sorted out. Also I hope your titan X is either a hybrid or water cooled, because the GPU core will throttle down even without OCing it. When it's Overclocked and not cooled properly it can severely impact 1% and .1% lows.
*watches this 2 years down the line, laughs*
I wish you would have run all the tests in 4+0. I've seen some crazy results where it out-benches the 4+4 Ryzen, which would imply that thread scheduling on cores (not necessarily Windows' job) would play a huge role in performance degradation when threads are across CCX units when they'd run much better on the same CCX unit.
Wichard \o/
intel overclocking is insane, my i7 5820k have a base cloak of 3.3ghz, I run it on a solid 4.5ghz since I'v have it, 1.2ghz boost (I bought it like 2 years ago to raplace my fx8350, I was not able to overclock the fx pass 4.4ghz and the base clock was 4ghz) but still, I'm happy that rysen came to compete, I just hope it works out, my next cpu could be amd again.
if you game on 1440p or higher, ryzen is a no brainer.
John Dough LOL
upvoted for being funny :)
extreme: I see better and better gaming for ryzen in the future :)
sadly 1080p is still the most popular resolution and game devs aren't going to start doing more work because of Ryzen....
On a side note, dual cores and quad cores account for close to what 90% of what gamers are currently using.
source: Steam Hardware Survey.
DaKrawnik 1080p is popular but why would you pair a 500 dollar cpu with a 1080p monitor? have you ever ask yourself that?
#1 theory is that in games threads are jumping from one CCX to another way too often. Why didn't you try 4+0 configuration then?
you keep saying that the 1800x "gets the job done" and "isnt necessarily a bad cpu", but you forget... its $200 more expensive than the 7700k while performing worse in most cases
Unnamed Internet Citizen Everything wrong.
I feel like we should at least wait for the 1600x before we call Ryzen a failure. the 1800x is a workstation cpu (dont know why they even advertised it for gaming) and is fairly equal to the 6900k in that regard. If it is true that the 1600x will have the performance DF is theorizing, the $250 1600x will only have 15% lower performance than the $350 7700k, which seems fairly competitive.
You seem to forget that the R7 Ryzen is competing against 6900k, wich costs... what? 1000$? And the most expensive R7 is what? 499$? Its a no-brainer. Its AMD that wins that one, The Same or better performance for half the price!
That's what the 1700 is for.
Wonderwolf it performs far far better than the 7700k in every way, other than some games.
it's all a matter of optimization. RAYZEN is totally brand new architecture, just give it some time for updates and stuff.
this thing barely hits 40% usage when the i5 7600 overclocked to the limits is about the explode at 100%.
Cant wait to play Minecraft with 1000fps and 4k! Who needs quality games like Bloodborne, Horizon, Nioh or Persona 5 if i can play CSGO and minecraft all day with 1000fps?!
Low quality troll bait.
Nunovski You fell for the bait m8
1800x @ 3.92Ghz (on air for now, till I get my AIO bracket) - RAM running only at 2133Mhz (should run at 3Ghz) - GTX980tiSC
Here I have over 70fps in maxed out GTAV (1080p) and my monitors are 60fps. So plenty of power here. I'm more than happy with my new system (coming from a 3770k at 4.6Ghz) as it was an improvement all across the board. And there is still room for improvement.
AMD made it this time. 8c/16t does makes a difference in a lot of apps here.
"End users should not have to use the BIOS to get better performance" thats the exact reason why we have bios options aint it? Overclocking? Tweaking settings for max performance? that statements makes 0% sense. The same exact thing is true for intel, sometimes HT is faster, sometimes its slower, start complaining about that?
Jean T and that probably means that they don't have a 144hz monitor, so as long as ryzen can output 60+fps in games any "normal en user" shouldn't really care
Bockstiegel Gaming yeah but do you want to be doing this for every game? Based on what AMD have told DF - you need to pretty much change the settings for every game you want to play
Bockstiegel Gaming you shouldn't have to tweak the bios for every single game disabling or enabling SMT, that's fucked up.
I think he meant to say 'on a per game basis'
Having to reboot your system, go in to the BIOS and change specific settings just for one game to run well, and then having to do it all over again for a different game - that's not something people should have to do.
1) He's talking about the average consumers, not the enthusiast who knows what they're doing. A motherboard BIOS is far from user friendly when it comes to the average consumer.
2) Please name a current gen game where Intel's HT should be disabled to get better performance?
3) The fact that RAM speeds are directly tied to the performance of the Ryzen chip is proof enough that you have to go into the BIOS on Ryzen chips to improve performance.
So to test your theory @3:00 and @7:30, will you make a vow to revisit this setup on March 18, 2018 - March 18, 2019 - March 18, 2020. That is the only way to properly evaluate the validity of the hypothesis; or more importantly whether or not it is a proper hyypothesis to be making for the consumer to evaluate. Love the channel.
Horyzen zero dawn
This video is greatly improved by turning the sound down and playing some euro pop to Richard flailing his arms around. No limits Richard, no limits.
God, how satisfying this is after all those insecure, salty PC gamers spammed "WAIT TILL RYZEN COMES OUT!!!!1" literally everywhere on this channel these last months.
god you're a moron
Walter White I'm a good analyst... yet I have no clue what the point of this comment is.
intel is still better for gaming tbh
barfyman362 but comically overpriced
I think he likes drinking your tears. You sure your good analyst? Not saying I agree with him, but it was pretty obvious.
Finally someone who benched with at least 1 core on each ccx disabled. It shows that either optimization or utilization of ryzens threads is indeed a limiting factor or that multithreading overhead is just the issue in the end.
What idiot would be an R7 CPU primarily for gaming... It's a workstation CPU. Ryzen is disappointing in terms of gaming. I don't see R5 doing any better.
I dont get this workstation. Ever since Broadwell E came out, Suddenly everyone is a 3D artist of Handbreak render person. Like really where did all this people come from that they need a CPU that is made for workstation.
As far as I remember first tests of Ryzen on public release were handbrake and some renderers.
It was not "marketed as gaming CPU", but rather "jack of all trades, master of multicore workloads"
The thing is that the R7s are not that good compared to the best gaming CPU Intel has which is the i7 7700k and this shouldn't be a surprise to anyone who knows about tech. Now the R5s are going to compete with the i5s. Performance in RyZen 5 is not going to be better than RyZen 7, true, but priced at 200 bucks it is going to be a better option than the i5s
I love when people say this. AMD marketed the R7 as a gaming CPU too.
UnknownUser Can it not play games satisfactory?
Hello Richard! Thanks for your great work here. I've seen an unnecessary and irrational level of hostility directed at you in particular, so I wanted to speak up to remind you that the silent majority appreciates your dedication to your craft. No one else in the world does the kind of testing Digital Foundry does. You should be proud of your team. Keep up the great work!
poor AMD, I remember the hype behind the FX8150 being the first 8 core gaming processor. It was a disaster
FX8350, far behind Intel
FX9590 5ghz chip doubles your monthly power bill and really not much better
fx series has a design flaw: the cores share resources specifically Floating Point Unit which limited it's performance
Dont google 9590 launch price
you do realize that the fx 8xxx series was made with MANTLE in mind right?? RIGHT?
+IT Developer and your source is?
The 8-core FX CPUs were a disaster at launch mainly because: Windows didn't know how to properly work with them, games because were still mainly single/dual threaded and there was not a huge improvement over the older Phenom II CPUs. Where AMD went wrong was they didn't actually improve on the CPU architecture over time and they expected things to be different in a near future, yet that future is only happening now. It seems to me that AMD is much more into innovation than Intel and it they have good ideas, they just have really bad timing (and planning).
On the other hand tho, I can't really undestand the hate for the 8 core Ryzen CPUs. If they weren't made to be gaming CPUs, why don't test and judge them using the things they were born to do? I don't remember people hating Intel 6 and 8 core CPUs this much, and they were poor gaming CPUs as well.
That test of 6 cores vs 8 cores was amazing! Thank you! Ryzen 5 is gonna rock :)
Stupid AMD I wanted I7 6900K performance for 200$ or less , AMD always fails !!!
Janis Berzins not really it is half the price and it is around 5-10% slower
Janis Berzins it's not really a gaming CPU, it's much better suited for things like video editing
Janis Berzins Everyone fell for the bait
I wanted 6900K perfomance for free, wtf I hate AMD now!
Renusek Free is the best price !!
@DigitalFoundry
Thank you for looking into the Ryzen topic. What immediately makes most sense right now, is 4+0 configuration you have yet to test. If I may suggest with incoming R7 1700 vs 1800x comparison, could you include this test as well? Disabling one CCX would offload InfinityFabric on CCX vs CCX communication and reserve whole bandwidth for DRAM/SOC. Its being said that this yields substantial FPS increase.
There is an option in the bios to use only 4 cores in 2+2 config and 4+0 config. It would be a great thing to test both of these configs and show us the performance penalty in using 2 ccx units versus only one.
Perfectly happy with my i7 875k here... and that's a year older than the 2500k.
Currently eye'ing the Ryzen platform... though I might be waiting for the Ryzen 5. Because the CAD is too damn low and I can't afford to pay $600 for a CPU.
This must have taken a shit tonne of time to do. Just want to say i appreciate all the hard work you guys do. Keep up the good work.
Thanks for the review Richard, I have been eagerly waiting for it!
Ryzen still has room for improvement, it's a whole new platform with issues to fix,
back in 2011 the 8150 got spanked by the 2500k even though it had half the cores, it destroyed the FX in anything but cinebench. Everyone told us "yeah just wait, wait until more apps and games become multithreaded, then you'll see the true power of the 8150".
6 Years later the Ryzen lineup gets spanked by the 7700k in everything but highly multithreaded apps. And people are again saying the same thing: "yeah just wait until we get ryzen optimized apps and game start using more cores".
Its been 6 years already, how much more do i have to wait?
Thanks for the great review! DF always has some of the most in depth frame time and pacing reviews.
Ryzen is great for workstations but I love my 7700k for pure gaming, which is what my rig is for. A 5.0GHz 1.36v 100% stable OC took me 4 minutes to setup and works flawlessly. Also right now Microcenter has it for 299.99 USD and the i5 7600k for 199.99 USD. Thats a steal.
First of all, great video. I love your work and appreciate your consistent and well developed methodology in your bench marking. The numbers you provide are invaluable to the community. I do have an issue with some of your commentary and implications though.
Your comment at 2:54 that implies relative performance today will mirror relative performance tomorrow is a bit presumptuous. That theory tends to work in apples to apples comparisons...like Kabylake 7700k vs Skylake 6700k. It isn't so cut and dry when you are looking at apples and oranges. Ryzen is definitely an orange to intel's apple. It is a completely new and different animal and there are numerous reasons to believe it has the potential to leverage more of itself in the future than it is currently utilizing in games; while its clear Kabylake's quad core design is already reaching its limits in some workloads today.
It is much more reasonable to hypothesize that Ryzen will make up much of its early deficit through the teething optimizations all new architectures go through. This is supported by the evidence that it clearly can compete and even surpass the 7700k in some gameplay situations today, even without specific software optimizations from developers being leveraged yet. (see Hardware Unboxed's results and many others who show similar anomalies) It is clearly an equal in almost every way to the 6900k both in multi thread and single thread workloads, and there is no reason it should not perform that well in most games...yet it does not. It can in some but not others? Makes no sense, and my intuition tells me it will be fixed in time. This is not FX bulldozer. It clearly has the power FX never did, and some optimization or bug is causing erratic behavior.
The future of gaming is clearly moving from 2 and 4 threaded workloads to 8 thread workloads as 8 threads becomes more and more mainstream, with even consoles supporting and leveraging 8 now. Anything with less than 8 true cores will start to fall behind in performance compared to most CPU's that have them. Speculating how quickly this happens is not an easy thing to do, but make no mistake, it will happen...just like it did from single thread to 2 and from 2 to 4. . As it does Ryzen will age like a fine wine and every quad core chip on the market today will get longer and longer in the tooth until they fall out entirely.
Kabylake is the undisputed king of high refresh rate gaming TODAY...and probably tomorrow as well. But I really feel the future belongs to 8 core chips. Gamers care as much about value as they do about records. A solid 8 core chip today will likely see solid performance long after more powerful quadcores have fallen out of favor. If you don't believe me, simply look at history. Find a single situation where buying a duel core hyperthreaded i3 chip when 2 threads for games was more common, panned out better than getting a quad core i5 of the same timeframe. Didn't happen. Games demanded more cores down the road, and hyper threads only get you so far. The same thing will happen with 8 cores soon, and my guess is in less than 5 years.
Great video again. Really appreciate your work at DF. The future looks interesting for all of us in the tech world. Cheers.
Man...that DFI Lanparty bring back memories...
"''''....i could overclock that *_mother_* to 4.4ghz whith no effort what so ever!" thats a classic line from Wichard!!
I was wanting a choice between CPUs. But results like these tells me if you want a good CPU for video making and lower cost go with AMD. Want a premium gaming CPU go with Intel. If it needs to be optimized the older titles probably won't get that optimization.
That 6 core performance has me intrigued . . Can't wait for April!
What a great video. Sad to see this Ryzen mess. That comparison between 1600x and 1800x with both having 6 cores really shows 1800x's poor optimization at gaming at the moment. Hopefully, it will get fixed in the future.
By far, this is the most complete and informative benchmark on the internet right now.
I've watched, read, saw a lot, and i mean, A LOT of bench about this Ryzen (i just bought my new PC, AMD Ryzen 7 1800x inside ;))
And this one was by far, the most interesting one.
Buying a Ryzen 7 now is a bit like gamble on the future. You know there is a beast in this CPU, but it's not in is final form.
I've choose to bet on it, first because i can go back to AMD, like good old days, but mostly because i'm pretty sure this beast is just not ready.
Wait for few update, wait for devs to work on it, and you'll see, in 1-3 years, they will release the Kraken !
I was waiting for this review so much..I believe only this channel.Brilliant review!!!
Four months later, the performance have improved via patches on both Rise of the Tomb Raider and Ashes of Singularity, making this comparison pretty much obsolete.
A "strategicaly shaved chimapnse" Richard is on Fire. Make more of theese jokes richard Please!
DX12 is known to be closer to the metal. Obviously no DX12 game has been optimized for Ryzen. And yes, CPUs do require optimizations. But Intel compilers have been used for so long that people forget that CPUs are not necessarily plug & play and you don't immediately get max performance.
A good, objective review as always. Looking forward to Ryzen 5 (and Ryzen 3) review + benchmarks. It'd also be good if for the cheaper Ryzen processors you compared performance to older Intel CPUs so those looking to upgrade can see what kind of performance benefits they can expect.
Great job Digital Foundry! Excellent work
If ccx to ccx is an issue, why the hell didn't you test 4+0 core disabling to confirm?
It'd be interesting to see Battlefield 1 cpu benchmarks. I know it's next to impossible to create an identical scene for multiple runs, but 64 player conquest absolutely murders my 4690k@4,2GHz.
"The i7-7700K is ahead in both counts" - yes, in FPS-Numbers, but looking at those Frametimes, the R7 is doing a MUCH better job. The R7 line is a razors edge, even compared to the 6900K. And that thing costs a LOT more. Especially in Crysis...look at those Graphs from the i7, they are bouncing all over the place. ALso at 4k, the difference is noticable.
What I got from this video (if you watched it to the end) is that the R5 1600 and 1600X will be the real heroes of the Ryzen lineup.
Did you try the 4+0 configuration to completely remove the data fabric from the equation. It may be a latency rather than a bandwidth problem
While the Ryzen chips do offer insane productivity value, I think its important to point out not all productivity tasks are highly multithreaded.
Many tasks still perform better with faster cores instead of more.
you guys should do gaming and twitch streaming tests using the 1800x 7700k and 6900k! I'm curious to see if it scales with more cores like AMD has been touting recently. loved the video, thanks!
i've been waiting for this. gj Richard. i knew your in depth analysis would be spot on with what information is available to date about Ryzen. outstanding work! keep it up! more pc master race videos please. the console peasantry is too much to handle. are you guys going to start benching different video games? it's been the same ones for quite a while now. perhaps try Mafia III. it''s supposed to love Ryzen. take care. see you next video.
It would be interesting to see how single channel performance compares to dual channel performance.
oh! wow! that kinda sucks, had really high hopes for this CPU, thought it would be a real alternative to i7. Hopefully they can optimize it in future generations.
Excellent In-depth Fair CPU review! I will still stick with my 4790K until the end of the decade but hooray for some competition!
IMHO Ryzen 7 1700 with an AM4 is a good choice.
The AM4 mobo will accept Ryzen + and Ryzen ++ and will be supported till 2020. The 7700, not only is a far weaker chip (and I am willing to bet will get left in the dust in a year or two) but is on a much less supported platform. Intel changes these things too much and too often.
Can you test and compare the 2+2 or 4+0 core configurations to help identify issues with the CCX interconnect?
Thanks for putting in the diligent work :) Looking forward to the gaming perf mystery eventually being explained.
Do you think this gaming performance is something that can be fixed? The single threaded power is there, and obviously multithreaded is, but it's like games don't understand what's going on :s
Still got my 4790k water cooled at 4.3 I'm very happy definitely wait it out a few years
What was the usage of the 7700k versus the 1800X during gaming benchmarks? You mention longevity. Considering the higher core count of the Ryzen chip and it likely being utilised less, would that not mean it has more room for growth?
I'll wait for the platform to mature a bit, hoping there will be exciting times ahead
Great work as always DF. Love the fact you even took the time to bench 6 cores.
Great review. Reflects what other reviewers have found. 1800x not worth the price for pure gaming, but overall not bad. The Ryzen 1700 is a way better value, cause it can overclock to 3.8 Ghz easily and gives pretty much the same game performance.
I'd be interested in more memory analysis (different modules, XMP vs manual OC) from Digital Foundry with the different Ryzen chips. Actually Hardcore Overclocking (better know as buildzoid) has mentioned that the 32x memory multiplier is broken and seeing as that they are using 3200 Mhz memory, they are either using the bad multiplier or a different multiplier with bclk overclocking (which dowgrades PCIe to Gen 2 or Gen 1).
If you're 100% only gaming, NEVER EVER do any productivity works on your PC (include microsoft office), and "extra 5 fps" really matters a lot to you ?, buy i7 7700K, there is no debate here.
Ryzen 7 design for productivity first, gaming second, that's why in a productivity benchmark, R7 1800X beats intel 6900K while 1800X cost HALF than 6900K.