-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
OcUK Ryzen 2000 series review thread
- Thread starter CAT-THE-FIFTH
- Start date
Caporegime
The TechReport also have applied all the fixes:
https://techreport.com/review/33531/amd-ryzen-7-2700x-and-ryzen-5-2600x-cpus-reviewed
However,they have not finished the game tests yet it seems.
Actually it depends on the games - for instance DigitalFoundry saw some games dip as much as 10% and they tended to be openworld games,which tended to stream data at a constant rate off their drive.
I run a modded FO4 playthrough on an IB CPU,and the patches make performance more choppier(less when switched off) as the game tends to stream data off the SSD.
https://techreport.com/review/33531/amd-ryzen-7-2700x-and-ryzen-5-2600x-cpus-reviewed
However,they have not finished the game tests yet it seems.
Actually it depends on the games - for instance DigitalFoundry saw some games dip as much as 10% and they tended to be openworld games,which tended to stream data at a constant rate off their drive.
I run a modded FO4 playthrough on an IB CPU,and the patches make performance more choppier(less when switched off) as the game tends to stream data off the SSD.
Deleted member 66701
D
Deleted member 66701
Yeah, just productivity tests. but looking excellent on that front as well.
Associate
- Joined
- 27 Dec 2013
- Posts
- 51
I'm liking the poor hopefulls who really believe amd improved performance by 20% over zen1 Haha
Sensible comment on anadtech failure review:
"Tropicocity:
My main gripe with people saying "other reviewers didn't use the meltdown/spectre" patches and stuff...
1. Those patches have already been tested and they do NOT affect gaming much at all, we're talking lower than 1% !
2. Even if you take out the entire meltdown/spectre thing, look at ryzen 1800x vs 2700x. a 3% IPC increase and some memory latency improvements do NOT account for 20% increased performance in gaming, not at a 200mhz clock change.
3. Even if Ryzen 2 series DID somehow gain 20% on Ryzen 1...why do other websites not show this? They all show at most 10%. Completely remove intel from the situation and you still have glaringly large performance jumps from Ryzen 1 to 2, this is what sticks out the most here."
Sensible comment on anadtech failure review:
"Tropicocity:
My main gripe with people saying "other reviewers didn't use the meltdown/spectre" patches and stuff...
1. Those patches have already been tested and they do NOT affect gaming much at all, we're talking lower than 1% !
2. Even if you take out the entire meltdown/spectre thing, look at ryzen 1800x vs 2700x. a 3% IPC increase and some memory latency improvements do NOT account for 20% increased performance in gaming, not at a 200mhz clock change.
3. Even if Ryzen 2 series DID somehow gain 20% on Ryzen 1...why do other websites not show this? They all show at most 10%. Completely remove intel from the situation and you still have glaringly large performance jumps from Ryzen 1 to 2, this is what sticks out the most here."
Look at the audio benchmarks - a big improvement there!
Some of these reviews are just plain weird.
Comparing 2700X chips to non OC 8600k and 8700k chips
Graphs omitting a ton of chips that have been reviewed but just not on graphs, so e.g. comparing 2700x with no ryzen1 chips, a 6600k and a 8700k but no 8600k, and no 2600x as well.
Graphs with 2700X OC slower than stock 2700X.
Comparing 2700X chips to non OC 8600k and 8700k chips
Graphs omitting a ton of chips that have been reviewed but just not on graphs, so e.g. comparing 2700x with no ryzen1 chips, a 6600k and a 8700k but no 8600k, and no 2600x as well.
Graphs with 2700X OC slower than stock 2700X.
Its good for consumers that AMD have closed the gap, no remorse from me neither.
As has been pointed out before if we had no ryzen 1, the 8 series intel chips would still be quad core and possibly higher priced as well to boot. Competition is a good thing.
The stock vs stock tests are heavily in amd's favour as they've made the cpu hit it's highest single core boost at stock and it boosts all cores at stock with XFR 2 also. Where the intel cpus are literally at base clocks.
I guess that's clever on amd's part.
I guess that's clever on amd's part.
A lot of the reviews are still using 2666MHz RAM as well.
Well yeah both companies played at this.
When 8xxx was released to press I noticed it was just 8700ks been reviewed and lack of 8600k's. This was likely because for gaming the 8700k has minimal gains over an 8600k and it would be seen as the poorer value chip.
For AMD they know they cant win the top clock race, so they release chips with more generous stock configurations vs the turbo clocks compared to intel, perhaps slip into the "review guide" compare to stock intel chips and to get favourable as possible results. Possibly also a list of games that are "recommended to bench" as well.
Dude seriously,they do depending on the game - look at the DF video. Some games showed no decrease but some did.
In games like W3 and ROTTR,peformance of CPUs like the Core i5 8400 dropped by nearly 5% to 10% after the last set of patches:
https://i.imgur.com/UwffANG.jpg
https://i.imgur.com/Hh5Rp1L.jpg
DF does their own custom run-throughs. In other games the patches made ZERO difference.
Here is one of the worst games for Ryzen,ie,Fallout 4:
https://cdn.sweclockers.com/artikel/diagram/14687?key=09ecc77da6d620128ae74cbb1b51940b
https://cdn.sweclockers.com/artikel/diagram/12689?key=3d59d063582dc8af5546a26671e89ceb
Yes,you read that right - a Ryzen 7 1800X was equal to a Core i7 3770K in the game.
Minimums are up around 14% to 16% depending on the SKU.
Plus,games are sensitive to not only IPC,but things like latency and clockspeed where Ryzen 2000 has improved.
Remember,some websites are also testing at stock RAM settings,ie,2666MHZ for Intel and 2933MHZ for AMD which AT has done. Both are the officially max supported memory clockspeeds.
Once you overclock the Core i7 8700K,with some fast RAM it probably will push ahead again.
Last edited:
Deleted member 66701
D
Deleted member 66701
Ocing disables Xfr, so it's often slower in single threaded apps. Same thing happens with Threadripper.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,818
Hence my 'running at stock' comment earlier. Almost pointless to overclock!!
in regards to meltdown, it can actually reduce kernel calls by a whopping 50%
spectre the hit is potentially even worse.
The real world affect of meltdown is simply based on how much cpu load is userland vs kernel load, on servers its often a big chunk on kernel and hence bigger performance hit, desktops often mostly userland, but its not a given, some games e.g. could have extra kernel calls and as such could easily get a bigger performance hit than 3%.
On my laptop I had security software that did massive kernel calls on loading task manager it has a i3 cpu. Pre meltdown task manager loaded in under a second, post meltdown over 8 seconds. Removing that security software fixed it, but is proof that affects of meltdown patches are not equally spread across all software.
Spectre is basically a flawed branch prediction logic, for the past decade most of intel's IPC improvements have been based on branch prediction enhancements and speculative paths. For spectre to be "fully" mitigated intel cpu's would easily lose around 60% performance on "everything". To avoid the PR meltdown microsoft only patched spectre on kernel calls. Meaning spectre at least on windows is not very well mitigated at all. Just for kernel calls, but again like meltdown if software heavily uses kernel calls the performane hit will be felt more.
So I am not immediately jumping on anandtech here. Also on my 8600k when I patched meltdown, the rig definitely felt slower, it felt like I cancelled my haswell to CL upgrade, and that was just meltdown, no spectre bios microcode updates on my rig. Right now I dont have meltdown patched for this reason as I am still assessing its value on my system. Spectre patches will probably never ever be live on my desktop, they clearly have little value, only "partial" mitigation for performance loss heavier than meltdown.
In fact in the AMD ryzen 2 thread I even predicted meltdown/spectre would close the gap possibly even completely.
spectre the hit is potentially even worse.
The real world affect of meltdown is simply based on how much cpu load is userland vs kernel load, on servers its often a big chunk on kernel and hence bigger performance hit, desktops often mostly userland, but its not a given, some games e.g. could have extra kernel calls and as such could easily get a bigger performance hit than 3%.
On my laptop I had security software that did massive kernel calls on loading task manager it has a i3 cpu. Pre meltdown task manager loaded in under a second, post meltdown over 8 seconds. Removing that security software fixed it, but is proof that affects of meltdown patches are not equally spread across all software.
Spectre is basically a flawed branch prediction logic, for the past decade most of intel's IPC improvements have been based on branch prediction enhancements and speculative paths. For spectre to be "fully" mitigated intel cpu's would easily lose around 60% performance on "everything". To avoid the PR meltdown microsoft only patched spectre on kernel calls. Meaning spectre at least on windows is not very well mitigated at all. Just for kernel calls, but again like meltdown if software heavily uses kernel calls the performane hit will be felt more.
So I am not immediately jumping on anandtech here. Also on my 8600k when I patched meltdown, the rig definitely felt slower, it felt like I cancelled my haswell to CL upgrade, and that was just meltdown, no spectre bios microcode updates on my rig. Right now I dont have meltdown patched for this reason as I am still assessing its value on my system. Spectre patches will probably never ever be live on my desktop, they clearly have little value, only "partial" mitigation for performance loss heavier than meltdown.
In fact in the AMD ryzen 2 thread I even predicted meltdown/spectre would close the gap possibly even completely.
Well the only RL impact I have seen on any games is open world ones or modded ones,where they really push SSD access,and I noticed it with my modded FO4 playthrough,which has hastened my need to upgrade now. Intel still is better in the game,but my main concern is if I/O is hit,I might have issues. So not sure what to do.
However,for a number of other games I didn't notice any impact TBH.
However,for a number of other games I didn't notice any impact TBH.
Found some gaming benchmarks from computerbase.de with 3466 MHz RAM.
https://translate.google.com/transl...nt-2-speicheroptimierungen&edit-text=&act=url
An interesting set of results to say at least. Those with fast RAM who like to tweak can get themselves a decent gain in performance, especially since Ryzen 2 can support 3400-3600MHz RAM out of the box.
https://translate.google.com/transl...nt-2-speicheroptimierungen&edit-text=&act=url
An interesting set of results to say at least. Those with fast RAM who like to tweak can get themselves a decent gain in performance, especially since Ryzen 2 can support 3400-3600MHz RAM out of the box.
i/o generates system(kernel) calls for sure, but its not the only thing that does.
The game where I felt it the most was FF15, without meltdown I can run for the most part stutter, hitch free. With it, its very evidently worse. That game does have denova DRM possibly meaning lots of i/o but regardless of whats going on under the hood, I definitely felt it.
Also chrome on startup blasts a ton of kernel calls, and it takes almost twice as long to start (about 170 tabs) with meltdown mitigated.
Also AMD is affected by spectre like intel, but because AMD's branch prediction is not as good as intel's their performance hit is less, so basically on mitigated systems the performance gap closes.
The game where I felt it the most was FF15, without meltdown I can run for the most part stutter, hitch free. With it, its very evidently worse. That game does have denova DRM possibly meaning lots of i/o but regardless of whats going on under the hood, I definitely felt it.
Also chrome on startup blasts a ton of kernel calls, and it takes almost twice as long to start (about 170 tabs) with meltdown mitigated.
Also AMD is affected by spectre like intel, but because AMD's branch prediction is not as good as intel's their performance hit is less, so basically on mitigated systems the performance gap closes.
Last edited:
Its my dilemma with FO4 - it runs better on Intel,but meltdown/spectre has manifested itself in more stutter especially with mods. So the question is do I go for Intel but potential have more I/O issues,or go with AMD and have less issues on that front,but worse CPU scaling which has other effects. Not sure if I should wait another year for Ryzen 3 or the CFL successor which hopefully should be fixed hardware.
Last edited:
