• 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

Engell

Member
Nice power consumption graphs there:
power_gaming.png

An 8 core more power efficient than a 4 core. They're gonna have good laptop chips it seems.

Though, I have to wonder why it uses more energy with SMT off.
edit: ahh ok the power measurement was done under some weird conditions with the 7700K overclocked with raised voltages and probably locked to its frequency in games.

not sure that graph is super correct.. that said the power usage for zen is great, but its not a freakin miracle
 
Has this been posted?

TECHPOWERUP - AMD Ryzen 7 1800X Review

Once the motherboards & BIOS mature, I think I'm going to build a 6/12 core work desktop.

That's probably the worst Ryzen review I have seen.

He only compares the 1800X against a 7700K in his tests, two processors in completely different price categories for a start. Then negatives include 'lack of integrated graphics' when he didn't do the same when reviewing the Intel HEDTs.
 

Steel

Banned
not sure that graph is super correct.. that said the power usage for zen is great, but its not a freakin miracle

Different tests have different results, same reviewer has this for superpi(single-threaded):


and Prime 95 which, according to the review, loaded all threads and cores.


Which aren't as favorable as the gaming test, but still quite favorable.

Cinebench is definitely not the power draw one would experience while gaming.
 
I was under the impression the impact of RAM speed --> Infinity Fabric speed was more widely known (discussed here and elsewhere), but for those unaware:



"I asked AMD a follow-up question about Infinity Fabric and how the CCX modules hooked up to the IMCs, and the response is interesting!"
https://www.reddit.com/r/Amd/comments/5zr8lv/i_asked_amd_a_followup_question_about_infinity/


a5gdj003gsly3fkg6.png



AMD Ryzen Infinity Fabric Ticks at Memory Speed
https://www.techpowerup.com/231585/amd-ryzen-infinity-fabric-ticks-at-memory-speed

Memory clock speeds will go a long way in improving the performance of an AMD Ryzen processor, according to new information by the company, which reveals that Infinity Fabric, the high-bandwidth interconnect used to connect the two quad-core complexes (CCXs) on 6-core and 8-core Ryzen processors with other uncore components, such as the PCIe root-complex, and the integrated southbridge; is synced with the memory clock. AMD made this revelation in a response to a question posed by Reddit user CataclysmZA.

Infinity Fabric, a successor to HyperTransport, is AMD's latest interconnect technology that connects the various components on the Ryzen "Summit Ridge" processor, and on the upcoming "Vega" GPU family. According to AMD, it is a 256-bit wide bi-directional crossbar. Think of it as town-square for the chip, where tagged data and instructions change hands between the various components. Within the CCX, the L3 cache performs some inter-core connectivity. The speed of the Infinity Fabric crossbar on a "Summit Ridge" Ryzen processor is determined by the memory clock. When paired with DDR4-2133 memory, for example, the crossbar ticks at 1066 MHz (SDR, actual clock). Using faster memory, according to AMD, hence has a direct impact on the bandwidth of this interconnect.
 
Infinity Fabric, the high-bandwidth interconnect used to connect the two quad-core complexes (CCXs) on 6-core and 8-core Ryzen processors with other uncore components, such as the PCIe root-complex, and the integrated southbridge; is synced with the memory clock. AMD made this revelation in a response to a question posed by Reddit user CataclysmZA.

Infinity Fabric, a successor to HyperTransport, is AMD's latest interconnect technology that connects the various components on the Ryzen "Summit Ridge" processor, and on the upcoming "Vega" GPU family. According to AMD, it is a 256-bit wide bi-directional crossbar. Think of it as town-square for the chip, where tagged data and instructions change hands between the various components. Within the CCX, the L3 cache performs some inter-core connectivity. The speed of the Infinity Fabric crossbar on a "Summit Ridge" Ryzen processor is determined by the memory clock. When paired with DDR4-2133 memory, for example, the crossbar ticks at 1066 MHz (SDR, actual clock). Using faster memory, according to AMD, hence has a direct impact on the bandwidth of this interconnect.

This is good news. Means the CPU Complex and latency between two CCX's can be alleviated in part by faster memory and thus improve performance in tasks that are effected by this, like games. I think this was already known though.

EDIT - ^^^ snap
 

Renekton

Member
https://arstechnica.com/gadgets/201...s-ryzen-at-games-but-how-much-does-it-matter/

But the current compiler does not offer any specific ability to target particular architectures. Competing compilers, such as gcc, do retain this ability, but Visual C++ aims for a one-size-fits-all output. Given the broadly similar constraints of both Intel and AMD processors over the last decade, this is arguably justified, but it has a consequence: if Ryzen has any uniquely peculiar requirements for its instruction scheduling, they're not going to be met. Developers are trusting their compilers to do something sensible, and as far as we can tell, they're not going to write their game engines in hand-crafted assembler to eke out extra performance that the compiler is leaving on the table.
 
Makes me think that a hypothetical future game console built off this or similar architecture would preform better then the PC version, given it's compilers could be a heck of a lot more targeted.
 

Sinistral

Member
Good video review from DF for games. Ryzen, is another sloppy launch from AMD. Par for the course. They really should have had the full lineup from the get go... and more bugs ironed out. It will be interesting to see what doesn't change when the R5 series comes out...
 
Good video review from DF for games. Ryzen, is another sloppy launch from AMD. Par for the course. They really should have had the full lineup from the get go... and more bugs ironed out. It will be interesting to see what doesn't change when the R5 series comes out...
AMD didnt really have the luxury of waiting any longer with investors on them heavily.

Anyway its a new architecture launch, bugs will br ironed out etc. shit happens.
 

Sinistral

Member
AMD didnt really have the luxury of waiting any longer with investors on them heavily.

Anyway its a new architecture launch, bugs will br ironed out etc. shit happens.

Oh I see why, and understand, but it's no excuse. Just their current circumstance. I'm enjoying my new 1800X computer immensely, and it is performing just as I expected... but the fact that my Ram is still forced to run at 2133, opposed to 3200 is irksome. I may have to wait until May for it to be fixed, long after the R5s come out.

I think the Zen core is going to be amazing going forward for Servers and HEDTs because of the scaleablity and flexibility the architecture affords, with APUs absolutely being stellar for low power applications, such as consoles, tablets and laptops if/when adopted. It'll remain to have the respectable but not kingly PC gaming performance for a long while.

But the marketing impact this small minded Gaming market permeates is a problem.
 

Their "emulated" 1600x tests are the most interesting part. Only a 5% drop across the games they tested for a $250 processor at stock.

It could end up very competitive with the 7600K in games and absolutely demolish it any heavily multi threaded tasks. Throw in the $220 1600 into the mix and OC it to 4ghz and you're looking at some fantastic performance for the money. Heck, an OCed 1600 could well end up outperforming a stock 1800x in gaming based on these results.
 
AMD have got their own high-end desktop platform coming second half of this year. 16 cores/32 threads.

Public knowledge by now but AMD has a new HEDT platform coming out in a couple of months
You'll see more of it at Computex I believe.
It's a 16 core /32 Thread, quad channel behemoth. And it is insanely quick in the tests that Ryzen is already excelling at. So Cinebench, and all other related productivity programs. The gaming issues that were causing the Ryzen AM4 CPUs to behave erratically to say the least have been ironed out. It's akin to a newer revision on a newer platform. This should be competing with the Xeon and of course 6950X Intel offers for $1700~$1800USD, but at about $1,000 USD if not less for some Skews. Coming soon.
CPSs are pretty big physically, about twice the size of surrent 6950X CPUs and a bit more perhaps.
And if you were hoping for pins, nope it's strictly LGA!
IT's NOT 8 channel, but Quad.

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-34#post-38797798

Been more or less confirmed by CanardPC recently as well.
 

Paragon

Member
AMD have got their own high-end desktop platform coming second half of this year. 16 cores/32 threads.
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-34#post-38797798
Been more or less confirmed by CanardPC recently as well.
Makes sense after seeing what they're doing with the R5 line.
Sell the lower-binned Naples (server) parts as high-end consumer parts.
Intel are in real trouble if AMD start selling 16-core CPUs with quad-channel memory and ECC support for $1000.

Their "emulated" 1600x tests are the most interesting part. Only a 5% drop across the games they tested for a $250 processor at stock.

It could end up very competitive with the 7600K in games and absolutely demolish it any heavily multi threaded tasks. Throw in the $220 1600 into the mix and OC it to 4ghz and you're looking at some fantastic performance for the money. Heck, an OCed 1600 could well end up outperforming a stock 1800x in gaming based on these results.
Yes, those were very interesting results.
I guess it's more that the games tested just don't really scale much beyond 12 threads, and if they did, the 1800X would prove to be faster.
But 95% of the performance (in games) at a lower price makes the R5s a lot more interesting.
It seems like the R7 is maybe the only area where Ryzen is failing to deliver if you have interest in them as a gamer.
 
Makes sense after seeing what they're doing with the R5 line.
Sell the lower-binned Naples (server) parts as high-end consumer parts.
Intel are in real trouble if AMD start selling 16-core CPUs with quad-channel memory and ECC support for $1000.

.

The exact same thing was said about the Ryzen 7s, but thanks in part to the likes of you, Durante and Dr.Rus, those were tanked and it looks like they'll have zero effect on the market ha ha.

No but look, AMD's HEDT platform extra details:

16 cores, 32 threads in 4-6 months on a 2 chip multi-chip module. Will be 2.4 - 2.8 GHz on LGA Socket LGA SP3r2. ~150W with 4 channels of DDR4.

The bit in the original quote about the gaming perf being improved sounds like a fanboy hype interjection. But the rest sounds good.
 

SURGEdude

Member
Yes, those were very interesting results.
I guess it's more that the games tested just don't really scale much beyond 12 threads, and if they did, the 1800X would prove to be faster.
But 95% of the performance (in games) at a lower price makes the R5s a lot more interesting.
It seems like the R7 is maybe the only area where Ryzen is failing to deliver if you have interest in them as a gamer.

I think from a marketing perspective AMD made a mistake by launching the R7s before the R5s. I know why they did it, and it's not a knock on the chips. For their intended use they are frankly beyond impressive. But because of the hype and the fact that consumer non-pro customers dominate the discourse of tech beyond their actually share of the market they ended up focusing not on the impressive parts of the R7, but instead decided that if they weren't tops for gaming then they were not good.
 

dr_rus

Member
AMD have got their own high-end desktop platform coming second half of this year. 16 cores/32 threads.

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-34#post-38797798

Been more or less confirmed by CanardPC recently as well.

This will end well...

The exact same thing was said about the Ryzen 7s, but thanks in part to the likes of you, Durante and Dr.Rus, those were tanked and it looks like they'll have zero effect on the market ha ha.

Yes. We three are at fault for all issues with Zen.
 
Look all I'm saying is I can sleep well at night knowing I tried to stimulate a bit of competition in the market and help out all you lot here. God I wanted to help myself too of course.

Whereas some of you...well, you'll have some explaining to do to your wives in a few years time on why the Intel Salt Lake i3 CPU cost you $500.

Let's hope the R5's do better, becos from where I'm sitting, it looks like the Ryzen 7's have did about the same as Bulldozer in a commercial sense.
 
Oh I see why, and understand, but it's no excuse. Just their current circumstance. I'm enjoying my new 1800X computer immensely, and it is performing just as I expected... but the fact that my Ram is still forced to run at 2133, opposed to 3200 is irksome. I may have to wait until May for it to be fixed, long after the R5s come out.

I think the Zen core is going to be amazing going forward for Servers and HEDTs because of the scaleablity and flexibility the architecture affords, with APUs absolutely being stellar for low power applications, such as consoles, tablets and laptops if/when adopted. It'll remain to have the respectable but not kingly PC gaming performance for a long while.

But the marketing impact this small minded Gaming market permeates is a problem.
What motherboard do you have? Got to 3200 on my ASRock Taichi. TridentZ btw
 

Steel

Banned
Look all I'm saying is I can sleep well at night knowing I tried to stimulate a bit of competition in the market and help out all you lot here. God I wanted to help myself too of course.

Whereas some of you...well, you'll have some explaining to do to your wives in a few years time on why the Intel Salt Lake i3 CPU cost you $500.

Let's hope the R5's do better, becos from where I'm sitting, it looks like the Ryzen 7's have did about the same as Bulldozer in a commercial sense.

? The hell are you saying?
 
ComputerBase —— AMD-Ryzen benchmarks: Games under Windows 7, CoreParking and HPET analyzed


Google Translate:

Conclusion: Ryzen works faster on Windows 10 than on Windows 7, even if the game has been set to "Balanced" as an energy saving profile. The change to Windows 7 brings only in Battlefield 1 and Project Cars clear advantages. In Project Cars, however, the same advantage can also be achieved under Windows 10 by selecting the "Maximum performance" profile. AMD's statement that scheduling under Windows 7 and Windows 10 is essentially the same for Ryzen, can be complied with. As claimed by AMD, obviously, CoreParking is a problem - and only under Windows 10.



Core Parking is not a problem under Windows 7

The top performance benchmarks on Windows 10 show that sleeping the CPU cores through Windows can cost performance in some games. Interestingly, this is not the case under Windows 7, although the older operating system uses CoreParking. Why? The function works differently under Windows 7 than under Windows 10.

Windows 10 lowers 14 threads from the 16 existing threads on the Ryzen 7 1800X at low load. Threads one and two are still active and threads three to 16 are switched off. Probably Windows 10 thus a CPU cores with SMT active. Windows 7 also saves individual threads at low load, but a maximum of half and for a different system. According to the Windows 10 logic, threads one to eight must be affected. But instead, every second thread is put to sleep, ie threads 2, 4, 6, 8, 10, 12, 14 as well as 16.

coreparking7-108zjdt.png


If Windows 7 assigns the threads the same as Windows 10 (with respect to the actual kernel and SMT), then it is conceivable that Windows 7 disables the logical "SMT threads", but leaves the actual kernels active. The processor does not have to be "wakened" in time. This and the fact that Windows 7 leaves six threads more active than Windows 10 might explain the problem of core parking under Windows 7, while the feature under Windows 10 does not work smoothly.


Full 16-thread diagram — Ryzen 7 1800X - Core Parking under Windows 10
https://www.computerbase.de/bildstrecke/77193/2/

Full 16-thread diagram — Ryzen 7 1800X - Core Parking under Windows 7
https://www.computerbase.de/bildstrecke/77193/3/



High performance under Windows 10 also has drawbacks

For the game performance under Windows 10, it is therefore advisable to switch the power management to maximum performance. Since neither the CoreParking nor the downclocking of the CPU is active, the power consumption at idle on the Windows desktop increases however. The result is a plus of nine watts and therefore a power consumption of 62 watts instead of 53 watts.

The softer CoreParking under Windows 7 also shows itself at the energy cost measuring device : Instead of the 53 watt under Windows 10 this system approved under Windows 7 slightly higher 58 watts from the socket. However, these are still four watts less than in the peak performance mode of Windows 10.

power-drawwnji0.png


Conclusion: Windows 10 "Balanced" costs performance

This article confirms the statements made by AMD regarding the scheduling and core parking of Ryzen under Windows : On average over all 13 tested games Windows 10 and Windows 7 do not take much, in the tendency the new operating system is already faster without intervention.

Performance gains due to the change to Windows 10 are usually also possible under Windows, if the energy saving profile "balanced" is switched to "maximum power". Two percent faster the games then run on average, ProjectCars represents an absolute exception with almost 20 percent gain. On Windows 7, however, the change does not bring any performance gain.

The reason is the sleeping of single cores and threads, which works differently under Windows 10 and apparently not correctly with Ryzen. The patch announced by AMD for early April appears to be meaningful. If you already want to use the positive effect of this adaptation, Windows 10 must operate in the profile "maximum performance". From April, "Balanced" should then bring the same power, but in the idle, but also save a few watts of electricity. Intel does not use core parking under Windows 10 with both Kaby Lake and Broadwell-E.



Windows 7 is not an advantage (RE: core parking)

Those who still use Windows 7, as a player, unlike in forums often written, has no general speed advantages with a Ryzen processor. The advantage of the operating system is that the integrated CoreParking in the profile "Balanced" does not cost any game performance. Windows 10 in the "High Performance" profile is ever faster with the exception of Battlefield 1.
 

Datschge

Member
It appears of the upcoming 4c8t models 1500X has 16MB L3 cache, while 1400 has 8MB. This leads to the common guess that the former is a 2+2 and the latter a 4+0 configuration. If accurate the latter would be preferable for gaming even if stock and turbo frequencies are lower (3.2 vs 3.5, 3.4 vs 3.7, no 0.2 XFR). The latter is only $20 cheaper, but all the frequencies and turbo don't matter anyway once one overclocks a Ryzen...
 

Paragon

Member
It appears of the upcoming 4c8t models 1500X has 16MB L3 cache, while 1400 has 8MB. This leads to the common guess that the former is a 2+2 and the latter a 4+0 configuration. If accurate the latter would be preferable for gaming even if stock and turbo frequencies are lower (3.2 vs 3.5, 3.4 vs 3.7, no 0.2 XFR). The latter is only $20 cheaper, but all the frequencies and turbo don't matter anyway once one overclocks a Ryzen...
Anandtech already have confirmation from AMD that they're going to be 2+2 configurations.
AMD are just making one part and binning it for all models.

http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th said:
We have confirmation from AMD that there are no silly games going to be played with Ryzen 5. The six-core parts will be a strict 3+3 combination, while the four-core parts will use 2+2. This will be true across all CPUs, ensuring a consistent performance throughout.

A CPU with a single 4-core CCX could potentially be more expensive to produce, since that's effectively a highly-binned part where all cores in the CCX must be working, vs a lower-binned part where only two cores in each CCX need to be working.
The R5-1400 is just going to be even lower binned than the R5-1500X, where it doesn't even need all the cache to be functional.
 
It appears of the upcoming 4c8t models 1500X has 16MB L3 cache, while 1400 has 8MB. This leads to the common guess that the former is a 2+2 and the latter a 4+0 configuration. If accurate the latter would be preferable for gaming even if stock and turbo frequencies are lower (3.2 vs 3.5, 3.4 vs 3.7, no 0.2 XFR). The latter is only $20 cheaper, but all the frequencies and turbo don't matter anyway once one overclocks a Ryzen...

If that really is true then an overclocked 1400 is going to be the real star for gaming focused rigs. Overclock it to 4ghz and I would expect that the 7700k may be the only stock Kaby Lake chip that will beat it in games.

It's going up against a locked i5 7400 with just a 3ghz base clock speed and no hyper threading.
 
If that really is true then an overclocked 1400 is going to be the real star for gaming focused rigs. Overclock it to 4ghz and I would expect that the 7700k may be the only stock Kaby Lake chip that will beat it in games.

It's going up against a locked i5 7400 with just a 3ghz base clock speed and no hyper threading.
It just means the 1400 is lower binned. They will be the 1500s that were unstable with the full cache active due to manufacturing defects. The same way that all R5s are lower binned R7s due to defective/unstable cores.
 

Durante

Member
The exact same thing was said about the Ryzen 7s, but thanks in part to the likes of you, Durante and Dr.Rus, those were tanked and it looks like they'll have zero effect on the market ha ha.
You are repeatedly being completely ridiculous in this thread.

I find it preferable to have a complete and balanced look at a product in all its performance scenarios, and discuss its shortcomings, even if I want it to do well.

You, conversely, clearly believe that for a product you want to succeed there should be a gag order on anything that could remotely be perceived as negative. You are welcome to that belief, but be prepared to be ignored by more level-headed people.

? The hell are you saying?
He's saying that in his world view, the only ethical thing to do is to hush up any issues a product may have if it is going up against a dominant competitor. It's all rather ludicrous.
 

Datschge

Member
A CPU with a single 4-core CCX could potentially be more expensive to produce, since that's effectively a highly-binned part where all cores in the CCX must be working, vs a lower-binned part where only two cores in each CCX need to be working.
The R5-1400 is just going to be even lower binned than the R5-1500X, where it doesn't even need all the cache to be functional.
All true. Funny that the forum at that place currently runs with the idea that 1400 could be a 4+0 part when an article there already got clarification otherwise.
 

Datschge

Member
Someone did some tests that (pending confirmation) may point to the source of Ryzen's weakness in gaming.

A 10% performance hit in single-threaded performance under Windows 8.1 on a Magny-Cours 2-module CPU is found when three factors are allowed to coincide: thread migration across modules is allowed, the OS timer is at a high resolution, and core parking is allowed.

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-41#post-38802307

The most interesting part is that on that system the issue disappear when the OS timer is not at a high resolution. At which point we come back to the earlier point in this thread that the scheduler moving threads between the two CCX at high speed may be what kills single-threaded performance.

Now someone reproduce those tests on an actual Ryzen system under Windows 10...
 
You are repeatedly being completely ridiculous in this thread.

I find it preferable to have a complete and balanced look at a product in all its performance scenarios, and discuss its shortcomings, even if I want it to do well.

You, conversely, clearly believe that for a product you want to succeed there should be a gag order on anything that could remotely be perceived as negative. You are welcome to that belief, but be prepared to be ignored by more level-headed people.

He's saying that in his world view, the only ethical thing to do is to hush up any issues a product may have if it is going up against a dominant competitor. It's all rather ludicrous.

Oh please Durante, don't ignore me. I find it continuously amusing how seriously you take everything, especially a CPU you have no intention of buying :)
 

Steel

Banned
Oh please Durante, don't ignore me. I find it continuously amusing how seriously you take everything, especially a CPU you have no intention of buying :)

I'm pretty sure Durante said he is buying a 1700, though, and why wouldn't he be interested in what AMD is doing, good and bad? Look, most people here recognize that AMD has come a long, long way with Ryzen and it does a lot of tasks extremely well for its price, but recognizing that Ryzen 7 isn't the best bang for your buck in purely gaming isn't shilling for intel.

I can't even tell if you're trolling.
 
Despite recently changing the HSF to one of my dual-tower coolers and having the build be that much quieter and cooler than before... I've bought an EK-XLC Predator AM4 kit, which means I'll likely be picking up a Predator 360 or 280 soon.

If I change the cooling again, there's a strong possibility I may change the motherboard as well (wanted to wait for "gen1.2/gen1.5" X370 models to be shown at Computex) to another X370. Note any complaints I've made about not running my RAM at full speed may not have been clear. They're Intel optimised and 16GB per DIMM. They simply aren't on my QVL (always risky to go against that) so this isn't a major mark against AMD or Gigabyte so much, though they could expand compatibility. They would almost certainly run at max rated or higher on FM2+/AM3+ but those are considerably more mature than AM4.

If I keep the board, it will be nice to see AMD's RAM speed update in May bring me up to speed. I can see most/all A-series boards being left out from manufacturers' update implementations, and selective roll out for B-series. Can't imagine why Gigabyte wouldn't provide AMD's update for my board but we'll see if I still have it by then.

At or near full speed on this RAM, the Fabric clock will run ~25-30% higher or so.
 

JohnnyFootball

GerAlt-Right. Ciriously.
I'm pretty sure Durante said he is buying a 1700, though, and why wouldn't he be interested in what AMD is doing, good and bad? Look, most people here recognize that AMD has come a long, long way with Ryzen and it does a lot of tasks extremely well for its price, but recognizing that Ryzen 7 isn't the best bang for your buck in purely gaming isn't shilling for intel.

I can't even tell if you're trolling.


Yeah, if you want gaming Kaby Lake is the fastest and cheaper, but the 1600X has a decent chance at being the best bang for buck, even though it won't be the fastest.
 

Renekton

Member
Was so anxious about getting a rig ready for Mass Effect lol.

But now, can slowly wait for Vega, Ryzen mobos and Skylake-X.

For the best, I think.
 

Paragon

Member
More benchmarks from Digital Foundry comparing the 1800X/7700K/6900K: https://www.youtube.com/watch?v=HZPr-gNWdvI

Some interesting results there, with the 7700K having a clear lead in some games and being tied or behind in others. It really seems like 4 cores is being pushed to its limit now.
Crysis 3 is interesting in that it seems to scale well with core count in some situations, but is still reliant on single-core performance too.

The SMT On/Off and Windows 7/10 comparisons are a bit concerning.
There seemed to be 10-15 FPS to be gained from it, but what performed best varied in each game.
 
More benchmarks from Digital Foundry comparing the 1800X/7700K/6900K: https://www.youtube.com/watch?v=HZPr-gNWdvI

Some interesting results there, with the 7700K having a clear lead in some games and being tied or behind in others. It really seems like 4 cores is being pushed to its limit now.
Crysis 3 is interesting in that it seems to scale well with core count in some situations, but is still reliant on single-core performance too.

The SMT On/Off and Windows 7/10 comparisons are a bit concerning.
There seemed to be 10-15 FPS to be gained from it, but what performed best varied in each game.

The 1800X in this video performs so closely to the 6900K that it makes the reactions to the gaming performance look absurd. And this is whilst using an overclocked Titan X at 1080p as well!

The May memory update will be very interesting. Will performance continue to increase with memory frequency? Will a 3600Mhz memory kit have appreciable gains over 3200Mhz?
 

Datschge

Member
As you all know there is nothing wrong with the Windows scheduler, it works as intended. They just found a tiny loose screw somewhere.
There likely are a bunch of other tiny screws left there and at other places.
That 35% gain is only in the Unreal Tournament 3 benchmark, an old DX10 game.
The way Ryzen gaming benchmarks have been random up to now updates like this one are going to randomly invalidate previous benchmarks and nobody knows which games are affected. ^^
 

Paragon

Member
The most recent Windows 10 update appears to have tweaked the scheduler a little: http://i.imgur.com/uaLeqhW.jpg
https://twitter.com/BitsAndChipsEng/status/843864982320267265
Certainly looks like it could have been split over CCXes in the original test.
I'm not sure how that does anything but highlight the issue of the dual-CCX design though; anything which spawns more than four threads is going to end up split across CCXes, and performance could vary depending on which threads are split across the CCXes.
 
Top Bottom