Yeah, for gaming, faster ram is better than quad channel. With quad channel, the pipe is wider. But it isn't faster. Dual channel is wide enough. Speed is what you need. And indeed, Skylake and Kaby Lake scale nicely in most games, with faster RAM.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
AMD Ryzen Thread: Affordable Core Act
- Thread starter ·feist·
- Start date
That said, Intel's 6- and 8-core CPUs show much better scaling.
Is it likely for the gap between the Intel 6+ core and Ryzen to narrow with the potential updates to OS/BIOS/games(?), or can that gap only be filled by targeted optimizations that need to be added while in development?
Ryzen is behind in both IPC and clocks to a 7700k, but I am willing to accept the poor performance in current games if Ryzen can, going forward, offer gaming performance closer to the 6900k in computerbase's review.
AstroNut325
Member
I'm not sure I understand. First, Ryzen isn't performing poorly in games. Second, that chart you quoted, based on my understanding, shows a lower and flatter line for the 1800X compared to the 7700K. And to my understanding, that's more ideal. Also, if AMD was able to beat the 6900K in everything, I'm not sure the would be charging $500 for it.
Thanks Paragon and Durante for answering my question
So I guess that with time the AMD's 8 core CPU's will have "smoother" frame rate than intel's 4 core CPU's?Not many outlets are using Watch Dogs 2 in their tests, but the few I've seen other than ComputerBase have the 7700K showing better results than the 1800X.
For example:
Though in this test, the 7700K system has faster memory (3866 vs 2933).
On another note, the 5960X has slower memory and it's also doing better (2400). ¯\_(ツ_/¯
Ah well, I'm really curious on what Digital Foundry's results will be, if they do videos on Ryzen.
For example:
http://techreport.com/review/31366/amd-ryzen-7-1800x-ryzen-7-1700x-and-ryzen-7-1700-cpus-reviewed/8
Though in this test, the 7700K system has faster memory (3866 vs 2933).
On another note, the 5960X has slower memory and it's also doing better (2400). ¯\_(ツ
_/¯Ah well, I'm really curious on what Digital Foundry's results will be, if they do videos on Ryzen.
FX8350 vs R7 1800X by Wccftech
More benches @ http://wccftech.com/ryzen-fx-performance-gains-vishera/
GPU
GTX 1080 Founders Edition
FX Test System
CPU FX 8350 w/wraith
Memory 16GB Corsair Vengeance LP DDR3 1866
Motherboard Asus 970 Aura/Pro Gaming
Storage 480GB OCZ Trion 150 SSD
Power Supply Corsair AX860i
Ryzen Test System
CPU Ryzen 7 1800x
Memory 16GB Corsair Vengeance LPX DDR4 2400
Motherboard MSI X370 Xpower Titanium
Storage 480GB OCZ Trion 150 SSD
Power Supply Corsair AX860i
Conclusion
A long way and time have things come since the introduction of the FX 8350 and it shows. The Ryzen 7 1800x is far from an inexpensive chip bringing itself to the market at $499, but if the eight cores and sixteen threads is what you're after, they can be had for as low at $329 in the form of the Ryzen 7 1700, which is my personal pick of the bunch. But it is clear that a lot has come with Ryzen, a solid performing workhorse that has brought some arguable benefits to the gaming scene. Was the wait worth it? Sure, if you've got the money and the need. But me personally, I'm suggesting that those looking to build a sweet Ryzen gaming rig hang tight while the early bugs and issues are getting ironed out for now. I have a feeling that the Ryzen 5 1600x (6-core/12-thread) and Ryzen 5 1500x (4-core/8-thread) are going to be the real winners in this lineup as we're likely to see similar gaming performance for a whole lot less money. But so far Ryzen is performing on all cylinders right where I had personally anticipated. Now, if they could just get the memory speeds up.
More benches @ http://wccftech.com/ryzen-fx-performance-gains-vishera/
GPU
GTX 1080 Founders Edition
FX Test System
CPU FX 8350 w/wraith
Memory 16GB Corsair Vengeance LP DDR3 1866
Motherboard Asus 970 Aura/Pro Gaming
Storage 480GB OCZ Trion 150 SSD
Power Supply Corsair AX860i
Ryzen Test System
CPU Ryzen 7 1800x
Memory 16GB Corsair Vengeance LPX DDR4 2400
Motherboard MSI X370 Xpower Titanium
Storage 480GB OCZ Trion 150 SSD
Power Supply Corsair AX860i
Conclusion
A long way and time have things come since the introduction of the FX 8350 and it shows. The Ryzen 7 1800x is far from an inexpensive chip bringing itself to the market at $499, but if the eight cores and sixteen threads is what you're after, they can be had for as low at $329 in the form of the Ryzen 7 1700, which is my personal pick of the bunch. But it is clear that a lot has come with Ryzen, a solid performing workhorse that has brought some arguable benefits to the gaming scene. Was the wait worth it? Sure, if you've got the money and the need. But me personally, I'm suggesting that those looking to build a sweet Ryzen gaming rig hang tight while the early bugs and issues are getting ironed out for now. I have a feeling that the Ryzen 5 1600x (6-core/12-thread) and Ryzen 5 1500x (4-core/8-thread) are going to be the real winners in this lineup as we're likely to see similar gaming performance for a whole lot less money. But so far Ryzen is performing on all cylinders right where I had personally anticipated. Now, if they could just get the memory speeds up.
I guess what I am really trying to figure out is, If Ryzen can compare with the 6900k in some heavily multi-threaded production workloads will it ever be able to compare to the 6900k in a heavily multi-threaded game too?
Do games need to have Ryzen in mind from the start to fully utilize its power or, assuming they already perform best on an Intel 6c/12t+ CPU, can updates compensate?
'poor performance' was a bad choice of words. I meant I was fine with less performance than a 7700k in games that don't go beyond 4 core and rely more on single threaded performance.
Who knows. This is something that we'll really just have to wait and see.
How bottlenecked is even something like a 2600k if you're doing 60fps or less in modern games? How relevant are these CPU gaming tests, really, if you plan on upgrading within the next 4 years?
I feel like for the vast, vast majority of people, none of this stuff is going to matter, and they should look at frame times and non gaming usage.
I feel like for the vast, vast majority of people, none of this stuff is going to matter, and they should look at frame times and non gaming usage.
With time? Computerbase has 1800X being smoother than 7700K right now already...Thanks Paragon and Durante for answering my question
...at which point we can close the circle of commonly used gaming benchmark being completely useless since they are misleading. And not only because many of them are actually disguised GPU benchmarks instead testing the CPU.I agree as to the significance of the WD2 results, but your remark about the 7700k doesn't hold true in frametimes at Computerbase at least:
Sqrt_minus_one
Member
Thats a single benchmark.
Not many outlets are using Watch Dogs 2 in their tests, but the few I've seen other than ComputerBase have the 7700K showing better results than the 1800X.
For example:
http://techreport.com/review/31366/amd-ryzen-7-1800x-ryzen-7-1700x-and-ryzen-7-1700-cpus-reviewed/8
Though in this test, the 7700K system has faster memory (3866 vs 2933).
On another note, the 5960X has slower memory and it's also doing better (2400). ¯\_(ツ
Ah well, I'm really curious on what Digital Foundry's results will be, if they do videos on Ryzen.
Results are all over the place for this chip Computerbases Watchdogs 2 performance appears at odds with both Gamers Nexus and techreport which both appear to put 7700k ahead I assume they are simply testing in different sections or some kind of bios issue.
Techspot is another one (though all they have is average and minimum framerate results).
http://www.techspot.com/review/1345-amd-ryzen-7-1800x-1700x/page4.html
Hard to find more results for comparison since there's not many reviewers that used watch dogs 2 in their testing.
Unknown Soldier
Member
I paid USD$350 for my 5820K a year and a half back. Where I got the great deal was on the motherboard, only $130 after rebate. Also at the time DDR4-3200 was pretty cheap, got 4x4 GB G.skill sticks for $90 or something like that. Overall I feel like I got a good deal but 5820K is 6 cores and Ryzen 7 is 8 cores. You get more cores per dollar from Ryzen.
I agree as to the significance of the WD2 results, but your remark about the 7700k doesn't hold true in frametimes at Computerbase at least:
That said, Intel's 6- and 8-core CPUs show much better scaling.
HEDT is badass as fuck because it's built off the server chipsets. People don't believe me when I say quad-channel makes a difference. If it didn't matter, why would Intel support it? They aren't dummies and supporting quad-channel drives up the cost of both CPU package and motherboard so they wouldn't be doing it for no reason.
Computerbase did "frametimes in percentile" for all games they tested. In BF1 DX11 1800X even equals 6850K.
If you mean no other sites did 99th percentile benchmarks then we are walking in a circle again: The whole thread deteriorated once gaming benchmarks arrived since neither the readers nor many of the testers apparently know what there is being tested. They essentially tested GPUs, the state of BIOS versions, drivers, OS support (or the lack thereof) of an all new uArch, platform and chipset and often made imaginary conclusive conclusions of the suitability of the CPU for gaming.
I don't think anyone is suggesting that there is no reason to use quad-channel memory ever.
But I've yet to see any gaming tests where it's made a difference - especially when pit against a higher-clocked dual-channel memory setup - since X99 doesn't support anything like the speed that Z270 does.
Gaming seems to rely more on data rate and latency than bandwidth.
With Intel moving from 4-channel memory to 6-channel memory on LGA-3647, and AMD using 8-channel memory with Naples (server Zen CPUs), there are clearly applications for it.
Though I'd like some of the additional features supported by the upcoming LGA-2066 (X299) or LGA-3647 platforms, I suspect that the top gaming CPU released this year is going to be the 6-core Coffee Lake chip, as it should have high IPC and high clockspeeds - but with a limited number of PCIe lanes, only dual-channel non-ECC memory, and all the other compromises you get for buying into the consumer platform vs HEDT/server.
While you get 6 channel memory with the server parts, the fastest ECC memory I've been able to find is DDR4-2666 CL19, and you can't overclock Xeons either, so I expect that those parts won't be great performers in games despite being very powerful for certain types of application.
Thanks Paragon and Durante for answering my question
Again it depends on the game. If the 4 core is hitting 90+ utilization it will have bigger frame drops than a CPU with more cores, and similar IPC performance, if a game scales with them. Those frame drops will be felt a lot more than a higher core count having a higher average frame latency but is more consistent in adhering to that average.
PCGH tested Starcraft 2.
The frametime graph would show this.
If it had big drops/stuttering problems, you would see a sharp spike at the end of the graph.That's why it's arguably the best way to present benchmarks.
Thanks, that doesn't look good :/
yeah but for some reason their 7700k performs worse then on any other site... absolutely no one else show the 7700K with worse frame-times then ryzen. Also for some strange reason it looks like they are testing with a 980ti.
Don't underestimate Starcraft 2. I remember it being very heavy on the CPU when there's lots of units in play. We also don't know what they did. They could've just spawned 2000 units.
Evolution of Metal
Member
I know for sure that they haven't benchmarked the actual game, testing the game in some sandbox mode is unproductive and it doesn't represent the performance of the game. Even with garbage SC2 engine you shouldn't get sub-30 FPS on a modern day CPU on low settings.
michaelius
Banned
Starcraft 2 is one of worst offenders in doesn't scale to multiple threads category.
For typical everyday usage low voltage dual core with HT is enough.
For typical everyday usage low voltage dual core with HT is enough.
JohnnyFootball
GerAlt-Right. Ciriously.
Went to Microcenter and almost bought a 6700K for $280 and a $50 open box Z170 motherboard and some $100 DDR4. $430 minus tax.
Decided to hold off. Skylake is still excellent but I am now determined to wait and see on the 1600X.
Decided to hold off. Skylake is still excellent but I am now determined to wait and see on the 1600X.
Nah Blizzard games are notoriously poorly-threaded before Overwatch.
TC McQueen
Member
Probably should've picked up the RAM.
Coulomb_Barrier
Member
http://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-7-1800x-cpu-review/9/
Reading Kitguru's review. Very promising as for GTA5 a Skylake CPU is nearly 100% loaded, but Ryzen is about 50% loaded with only 30% of its SMT utilised.
It's clear there is huge room for the gaming performance to improve, so that side of the CPU is hardly a dud from a hardware level, with 100% utilisation of the CPU during gaming (likely will never happen) it would blow right past a 7700K.
The question is how much of this spare performance can be unlocked to bring the performance up to where it should be. There's huge potential just sitting there.
Yikes! Steve from Gamingnexus is getting death threats for his Ryzen review - https://mobile.twitter.com/GamersNexus/status/838221363991166981
Coulomb_Barrier
Member
Yikes! Steve from Gamingnexus is getting death threats for his Ryzen review - https://mobile.twitter.com/GamersNexus/status/838221363991166981
That's not a death threat is it that's just a d#ck.
Can we keep this nonsense out of here.
michaelius
Banned
Yikes! Steve from Gamingnexus is getting death threats for his Ryzen review - https://mobile.twitter.com/GamersNexus/status/838221363991166981
Business as usual with AMD fans. Every portal that dares to criticize their company is Nvidia/Intel shill no matter how legit issue is raised.
brain_stew
Member
Given that a 7700K will run 3866mhz memory more often than Ryzen will accept 2933mhz memory, it seems a fair comparison to me. Both platforms are being shown in the best light.
There's definitely some potential here for a good gaming CPU. Another 3 months to work out the UEFI, Windows, SMT and memory issues and a $200-$250 6C12T unlocked Ryzen 5 that can overclock to 4ghz on high end air could still be great competition for the i5 line in gaming rigs.
We don't really know at this point if it's primarily a software or a hardware issue. In any case, I would never expect or make purchasing decisions based on the idea of most PC games actively optimizing for any given CPU architecture.
The 5960X has quad-channel memory. It has a significantly higher memory bandwidth even at lower clocks.Though in this test, the 7700K system has faster memory (3866 vs 2933).
On another note, the 5960X has slower memory and it's also doing better (2400). ¯\_(ツ
I can't believe GamerNexus is blowing up over this AMD stuff lol. I've been watching his videos since pretty much the first time he uploaded to youtube. He is not a bias shill..he does amazing work much better than "bigger" youtube channels.
Kids thanks to the Internet can talk all the shit they want best to just ignore it.
Kids thanks to the Internet can talk all the shit they want best to just ignore it.
It's completely ok if you only want to know the differences between cpus in that scenario.
brain_stew
Member
Only started watching his videos over the last couple of months but I've found it to be among the best PC YouTube channels. His methodology for gaming tests is spot on and I'm pleased he's sticking by his testing.
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-8#post-38775732
Windows 10 is currently not working well with SMT compared to Windows 7. We'll see if the next OS patch addressed this.
Windows 10 is currently not working well with SMT compared to Windows 7. We'll see if the next OS patch addressed this.
LelouchZero
Member
I'm not too fond how long he tests for but I do like the methodology of taking into account the 1% and 0.1% lows.
CPU Test Methodology
When I do benchmarking I test for no less than 1 minute and 30 seconds, 30 seconds is not enough time to test games with many variables at play that can impact the performance in titles such as GTA V, Watch Dogs 2 and Battlefield 1.
lordfuzzybutt
Member
How about not lump every "AMD fans" together? That shit is not okay, nor is this kind of generalisation.
I'm sure if you hit him up he'll address why he tests for 30 seconds (if that is not a typo.). I'd also like to see the tests last for 2 mins or so to be honest.
Sqrt_minus_one
Member
I agree with you but a lot of sites do this though (looking at Computerbases frametime and frame rate graph indicates its 30 seconds as well, techreport is about 60 secs) - IMO it should ideally be a couple of minutes combining mostly intense and not so intense areas.
General Lee
Member
It really depends on what you want to accomplish with the tests. You can do testing in non-demanding areas and get high numbers and that might be 90% of how one might spend gaming, but once you get to the 10% that result won't be useful. That's why it's better to just take the most demanding area of a game and test that, and this might be different for GPU and CPU.
Take for example ROTTR Geothermal Valley, which is highly CPU bound in the village, and highly GPU bound right below the village in the forest. You could do a 2 min run through both and get an average, but that result would not really give you a very good description of the performance since it could be considerably off in both cases depending on the hardware bottleneck. It's important to disclose what sort of benchmark scene is used because of this reason. A 30s bench through the village gives a better indication of what sort of performance benefits one might get from a 7700K over a 1700 in that situation.
A frametime graph with video like DF does is really the best way.
JohnnyFootball
GerAlt-Right. Ciriously.
I've thought about building a system by buying a little bit at a time but I always decide against it because you could end up buying a defective part and not know it was defective until well past the return window.
Completely forgot about that. <-<;
Based on techreport, the 1800X is a bit better than the 4790K in gaming overall (which has slower memory but higher clock I guess).
It's just the 1800X either failing to surpass or barely surpass (depending on the reviewer) the 7700K on a well multithreaded title like watch dogs 2 makes the 7700K look like a more future proof option for gaming, despite the higher core count on Ryzen.
Taking into consideration that the 1600X will have the same clocks as the 1800X, the only way it'll perform better is if the updates/patches AMD talks about actually make a difference.
Coulomb_Barrier
Member
Yes but look at the details. The Skylake and Kaby Lake CPUs are almost at max load when running these types of games (see my above post) whereas the Ryzen is only at 50% load on all its cores in GTA5. So whereas Kaby Lake and SL won't get much faster, there is a large amount of headroom left with the Ryzens.
Wasn't aware of that, but, despite that, there's no guarantee that AMD will deliver on their promises that it'll get better with patches. For all we know, those could end up being just a minor improvement.
If they end up improving alot, that'll be good for everyone, especially since Intel does need competition, but until then, Ryzen can't really be recommended for those focused on gaming based on promises, in my opinion.
Ouch, so it's an OS level regression. As if people need any more reasons not to move to W10...
General Lee
Member
It's better to let the platform mature and work out the kinks. If you've got the luxury of time, waiting for Skylake-X/Kaby-X or even Coffee Lake or Zen 2 is always an option. Hedging your bets on supposed unused potential is always risky, especially when we're talking of devs utilizing a GPU/CPU that's not from the market leader. Fiji had lots of potential too.
Unknown Soldier
Member
People who already made the move to Haswell-E and Broadwell-E are going to be set for quite a long time it seems. Especially since the kinds of people who are on HEDT aren't exactly gaming at 1080p. It will probably be years before GPUs will catch up to the point where my 5820K will be bottlenecking anything at 4K resolution. Being on quad-channel memory means that I won't fall behind either if games continue to become more multithreaded and chew up more memory bandwidth.