• 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.

*** AMD "Zen" thread (inc AM4/APU discussion) ***

Panos said:
Because if they put it at 2560x1440 or 4K the 7700K will seem pointless and total crap.
As it does on the 1080p benchmark video of the same reviewer who did the 720p video......

The Ryzen CPU is pushing the GTX1080 to it's limit, with half the cores snoring, while the 7700K struggles with it's all cores running.

Wait for the 1080ti to hit the market, and then going to be fun seeing the reviews with the Ryzen against the kabylake ;)
Click to expand...

you know the 1080Ti is just a TitanXP, we wont see more than what we already saw, most ppl reviewed with TitanXP already
 
AlamoX said:
you know the 1080Ti is just a TitanXP, we wont see more than what we already saw, most ppl reviewed with TitanXP already
Click to expand...
Yeah but the TXP reviews were last year, was 1 card only and not reviewed by everyone.
After next week the internet will be flooded with the various AIB versions of the 1080Ti for full run across systems etc.
Especially showing off 2560x1440 and 4K :)


teppic said:
My world is upside down :D
Click to expand...

I believe they make the Wraith some time now, it was powering their humble APUs.
And is a damn good one. Few pages back, there is someone who runs the 1700 @ 3.9Ghz with just the stock cooler

Can you do that with the Intel stock cooler? ergh no because there is no one, and the one exists is crap and from the Pentium 3 era hasn't changed
 
Lokken86 said:
You guys probably have no idea what I'm on about here but I wonder how fast these can plot a hard drive for burst coin. My 3570k takes like 24hours to do 1TB :(.
Click to expand...

I assume that's some alternative crypto currency? Personally I'm all about the original, Bitcoin, sitting on my stash until the price hits the moon :D
 
So I guess AMD will launch the 6c/12 model at current i5 prices and the 4c/8t model at i3 prices.

That's going to shake things up a hell of a lot if so.
 
Hi guys,

My first post here, but been following the thread with great interest :)
First of all - the ASUS Hero manual has just been posted.

For the guy wondering about code 61: it's NVRAM initialization.

As my day job one of my main interests is architecture optimization on PS games, and all I can say is that a cluster switch for a thread is a major problem.
Seeing that CCX acts somewhat similar to a cluster on Jaguar (PS4, X1) with a decent thread affinity solution there is certainly some extra performance to be had.
The thread affinity solution could be done on the software side - I am willing to bet that's what AMD is asking developers to do as a very quick fix, just in case Microsoft is slow with a scheduler update.
L3 behavior should also be considered when tasks are executed - however this is more in the realm of game development, as the OS sees threads, not tasks, but modern game engines use thread pools with tasks executed on the pools.
The thread pool implementation should take the Ryzen architecture into account for optimal performance - this may require some time to get it right.
Ideally, the OS scheduler should take the cluster switch penalty into account and even if thread affinity is not specified by the application (pre-Ryzen games), it should consider keeping the threads on the same cluster/CCX if possible - this should be a fix for the OS by MS and/or AMD.
You'd probably want experiment with scheduling on separate cores as well, and seeing how utilizing 2 threads on the same core affects performance

Current console games use 8 cores (no HT) and a significant number of games (console ports) should actually work a lot better natively on 8 cores, which is pretty much Ryzen and i7 6900... the latter not really being an option for most gamers, unlike Ryzen.
What you want on Ryzen is very likely thread pools working similarly to the original console code, rather than dumbed down to the currently dominant 4 core Intel solution :)
I'd definitely expect great results once/if console developers get more involved with the Ryzen optimization - that's why AMD is courting them; they've been working on similar architectures for the past few years, unlike PC only developers.

Certainly interesting times :)
 
KrisWragg said:
I assume that's some alternative crypto currency? Personally I'm all about the original, Bitcoin, sitting on my stash until the price hits the moon :D
Click to expand...

Yes it it. It's mining on hard drives. Not as profitable as gpu mining but uses pretty much no power at all. If you have spare storage its worth doing. Planning on getting like 100TB just for fun and hold burst for a while and see if it moons. If not I'll sell up after roi.

You gotta plot the drives first though and takes a lot of cpu power. Perhaps Ryzen could get through a 5TB drive in a just over a day if I'm lucky.
 
IFW said:
Hi guys,

My first post here, but been following the thread with great interest :)
First of all - the ASUS Hero manual has just been posted.

For the guy wondering about code 61: it's NVRAM initialization.

As my day job one of my main interests is architecture optimization on PS games, and all I can say is that a cluster switch for a thread is a major problem.
Seeing that CCX acts somewhat similar to a cluster on Jaguar (PS4, X1) with a decent thread affinity solution there is certainly some extra performance to be had.
The thread affinity solution could be done on the software side - I am willing to bet that's what AMD is asking developers to do as a very quick fix, just in case Microsoft is slow with a scheduler update.
L3 behavior should also be considered when tasks are executed - however this is more in the realm of game development, as the OS sees threads, not tasks, but modern game engines use thread pools with tasks executed on the pools.
The thread pool implementation should take the Ryzen architecture into account for optimal performance - this may require some time to get it right.
Ideally, the OS scheduler should take the cluster switch penalty into account and even if thread affinity is not specified by the application (pre-Ryzen games), it should consider keeping the threads on the same cluster/CCX if possible - this should be a fix for the OS by MS and/or AMD.
You'd probably want experiment with scheduling on separate cores as well, and seeing how utilizing 2 threads on the same core affects performance

Current console games use 8 cores (no HT) and a significant number of games (console ports) should actually work a lot better natively on 8 cores, which is pretty much Ryzen and i7 6900... the latter not really being an option for most gamers, unlike Ryzen.
What you want on Ryzen is very likely thread pools working similarly to the original console code, rather than dumbed down to the currently dominant 4 core Intel solution :)
I'd definitely expect great results once/if console developers get more involved with the Ryzen optimization - that's why AMD is courting them; they've been working on similar architectures for the past few years, unlike PC only developers.

Certainly interesting times :)
Click to expand...

Welcome :) Great post ..
 
Re: RAM specific to Ryzen, so the below is just wild guess on my part...
It seems that high performance DDR4 RAMs (especially larger capacity) mostly use 2T command rate for Intel specific modules, but Ryzen likes 1T instead. There is also the issue with rank.
So the first announced 64GB 3200 module by G.Skill is quite possibly not snake oil, more like 4x16GB 1T modules instead of 2T ones. Not an issue for 2 slots, possibly an issue when all 4 slots are used, hence the downclocking or failure when all 4 slots are used.
I wanted 64GB at high speed for a Ryzen build, so it's a great relief that it will be actually possible :)
 
Once I have a Ryzen build (when I can actually buy said high speed 64GB RAM, not any time sooner) I'll probably do some experiments with CodeXL and thread pool scheduling.
As someone already highlighted I find it extremely unlikely that most old/already released games would get patched (for way too many reasons...), but new ones should have a very nice transition from console to PC for sure, as long as AMD takes developer relations seriously. I'd consider a change if anyone from AMD reads this ;)
An OS scheduler fix is definitely the way to go for already released software.
I could also imagine a solution similar to what we have in GPU drivers: detect the game to be run, and modify the scheduler behavior accordingly.
Not really nice technically speaking (a hack, really), but should work - unlike waiting for most existing games getting fixed (good luck with that!) :)
Besides, GPU drivers do this for years to "optimize" GPU access to their chipset...
 
You must log in or register to reply here.
Back
Top Bottom