AMD responds to 1080p gaming tests on Ryzen. Supports ECC RAM. Win 10 SMT bug

[Curious Pineapple]

Just now, zMeul said:

and you are assuming people will just top using VS2015 because?!

 

mainstream support ends 10/13/2020; extended support ends 10/14/2025  https://support.microsoft.com/en-us/lifecycle/search/19591

there goes your excuse out of the window 

I'm not assuming anything, nor is anything an excuse. Simply pointing out the fact that code can be optimised for a specific architecture, and if code was built for Bulldozer, it won't run that well on Zen. I won't move from 2015 unless there's a good reason too, but as I said, why update an older compiler just before releasing a new one. Who will move to VS2017 if the updates to it get backported to what's already installed?

[Curious Pineapple]

2 minutes ago, Prysin said:

Lol. I told you Ryzen would be a disappointment. I told you back in November. 

Cost less and performs around the same, damn I was really expecting to mine Bitcoin on this thing too. Such a shame.

[zMeul]

Just now, Curious Pineapple said:

code can be optimised for a specific architecture

there's but one arch, x86

Intel has the original implementation while AMD wings it - and by winging it they cut off few of their toes

 

code doesn't not get build for Bulldozer nor Zen, code gets built for x86

 

who will move to VS2017? people that want the latest shit and have the money to pay for; VS licences aren't cheap

[MakeAMDGreatAgain]

7 minutes ago, Curious Pineapple said:

Cost less and performs around the same, damn I was really expecting to mine Bitcoin on this thing too. Such a shame.

A 1800x performs worse than a i7 7700k in gaming and is 200$ more. It doesn't cost less lol

[Curious Pineapple]

5 minutes ago, zMeul said:

there's but one arch, x86

Intel has the original implementation while AMD wings it - and by winging it they cut off few of their toes

 

code doesn't not get build for Bulldozer nor Zen, code gets built for x86

 

who will move to VS2017? people that want the latest shit and have the money to pay for; VS licences aren't cheap

You really don't get it do you. Do some research on compiler optimisations, and the history of the x86 architecture.

 

Someone who wants the latest shit, as you put it, won't need to buy anything if it gets backported into an old compiler will they.

[zMeul]

@Curious Pineapple

 

I find this really fuckign funny since HPET was co-developed by AMD and MS starting 2005

now they're blaming HPET and BIOS not being on par .. what the fuck are they doing over there?! it's like they hired a guy to design the CPU, fired him, sent the designs to the fab, got the product and now they don't know what to do with it 

[Citadelen]

3 hours ago, LAwLz said:

-snip-

You're exactly right. But, as far as we know from AMD we should be seeing fixes in these in a couple of weeks to months.

[zMeul]

7 minutes ago, Curious Pineapple said:

You really don't get it do you. Do some research on compiler optimisations, and the history of the x86 architecture.

 

Someone who wants the latest shit, as you put it, won't need to buy anything if it gets backported into an old compiler will they.

the optimization of the code does not sit on the compiler creator, it sits on the software product developer

compiler optimization is not CPU dependent; unless you are talking about CPU's manufacturer created compilers that are used to compile code a very specific way; but MS' compiler is manufacturer agnostic, so this is not what you're talking about

[Teddy07]

2 hours ago, Hunter259 said:

 Except they use 4 cores because that is where it begins to become insanely difficult to continue making improvements. Some games will use more cores due to being very parallel in nature (Civ games come to mind).

 

 

Why is it so difficult past 4 cores? Splitting it up into 6+ than 4 pieces shouldnt be that hard?

 

Would love if you could explain it shortly because i am interested

[Mira Yurizaki]

6 minutes ago, Teddy07 said:

Why is it so difficult past 4 cores? Splitting it up into 6+ than 4 pieces shouldnt be that hard?

 

Would love if you could explain it shortly because i am interested

Multithreaded programming is more complex than "split tasks evenly."

 

Also it's a misconception that applications care about how many cores or threads are available to it. Only the OS cares about it for scheduling purposes. The way applications run on an OS is they submit their threads that are ready to run to a queue. The OS then takes whatever is in the queue and throws them onto whatever CPU is ready to take it.

 

So in other words, if an application doesn't appear to perform better past X amount of cores, it doesn't mean the application can only "see" X amount of cores. It just means Amdahl's Law caught up to it.

[Valentyn]

37 minutes ago, MakeAMDGreatAgain said:

A 1800x performs worse than a i7 7700k in gaming and is 200$ more. It doesn't cost less lol

Really depends on who you look at it seems.

1700 @3.9 vs 7700K @5.0 

 

 

[Curious Pineapple]

11 minutes ago, zMeul said:

@Curious Pineapple

 

I find this really fuckign funny since HPET was co-developed by AMD and MS starting 2005

now they're blaming HPET and BIOS not being on par .. what the fuck are they doing over there?! it's like they hired a guy to design the CPU, fired fim, sent the designs to the fab, got the product and they don't know what to do with it 

That must be it. AMD helped develop it 12 years ago, it must still be of use today. Maseratti developed a hydraulic braking system that fails if the engine stops. I guess we should still be using that too. What once improved performance, isn't always of use today. Don't see Netburst around now do we?

 

Fucking hell, it's a brand new product and you expect it to be perfect and the entire platform have zero bugs?

6 minutes ago, zMeul said:

the optimization of the code does not sit on the compiler creator, it sits on the software product developer

compiler optimization is not CPU dependent; unless you are talking about CPU's manufacturer created compilers that are used specifically to compile code a very specific way; but MS' compiler is manufacturer agnostic, so this is not what you're talking about

Not everyone uses Microsofts compiler do they. GCC can produce targetted binaries and the performance difference is noticeable. I would be very surprised if MS didn't put the work into the compilers to detect CPU manufacturer/model and use hardware features that are available to make the end result better. Not really a hard task.

 

1 minute ago, Teddy07 said:

Why is it so difficult past 4 cores? Splliting it up into 6+ than 4 pieces shouldnt be that hard?

 

Would love if you could explain it shortly because i am interested

It's more than just splitting code up into pieces. For rendering and transcoding, you just give each thread a number of frames to chew through, it doesn't matter what order they are done in as long as they are put together right at the end. It gets tricky when multiple threads need to read and write the same memory locations, or use data in a certain order. If you have 8 threads running and 7 of them need to wait for the result of the other one, things start to slow down. You can also have race conditions where 2 threads need to access the same data at the same time. For example thread 1 reads data in memory and starts working with it. Thread 2 changes that data in memory. Thread 1 is now working with incorrect data, and depending on the application that could either be a miscalculated physics computation and you end up in a wall, or an incorrect memory address and your application starts dumping memory.

[VerticalDiscussions]

Oh okay, makes sense xD.

[Colonel_Gerdauf]

On the topic of ECC, there is a bit of ambiguity at play. There is no mention so far of whether ECC support is merely by installation, or by actual operation.

[zMeul]

5 minutes ago, Curious Pineapple said:

Not everyone uses Microsofts compiler do they. GCC can produce targetted binaries and the performance difference is noticeable. I would be very surprised if MS didn't put the work into the compilers to detect CPU manufacturer/model and use hardware features that are available to make the end result better. Not really a hard task.

here we go again, and falls flat on it's face

if the compiler was skewed wouldn't all the code compile with it would show similar results? and yet 1800X does better than 6800K in certain workloads

 

and the titles that were developed with AMD's help? how did those ended up performing better on 7700K?

 

again, more BS excuses

[Curious Pineapple]

Just now, zMeul said:

here we go again, and falls flat on it's face

if the compiler was skewed wouldn't all the code compile with it would show similar results? and yet 1800X does better than 6800K in certain workloads

 

again, more BS excuses

Who the fuck said it was skewed? Can I please have some of whatever you are on. Seriously, I ran out of the good shit ages ago, only got Tramadol and codine left now

[Mira Yurizaki]

5 minutes ago, zMeul said:

here we go again, and falls flat on it's face

if the compiler was skewed wouldn't all the code compile with it would show similar results? and yet 1800X does better than 6800K in certain workloads

 

again, more BS excuses

Not all applications use exclusive features and often times for benchmarking software, they specifically avoid using hardware features because they want to give everyone a fair test.

 

Of course this pissed a bunch of people off when Futuremark said 3DMark's Time Spy was not optimized for anyone (despite the fact Intel, NVIDIA, and AMD said don't do it).

 

EDIT: By the way, using hardware features specific to one vendor caused a huge stink with Futuremark on one of their earlier PCMark suites. Ars Technica found out if you change the CPUID on a VIA CPU to look like an Intel CPU, the benchmark would give an impossibly better score because PCMark used instructions that the VIA CPU supported, but would not use them on any CPU that didn't identify as an Intel one.

[zMeul]

1 minute ago, Curious Pineapple said:

Who the fuck said it was skewed? Can I please have some of whatever you are on. Seriously, I ran out of the good shit ages ago, only got Tramadol and codine left now

you, 8minutes ago:

Quote

Not everyone uses Microsofts compiler do they. GCC can produce targetted binaries and the performance difference is noticeable. I would be very surprised if MS didn't put the work into the compilers to detect CPU manufacturer/model and use hardware features that are available to make the end result better. Not really a hard task.

 

[MakeAMDGreatAgain]

11 minutes ago, Valentyn said:

Really depends on who you look at it seems.

1700 @3.9 vs 7700K @5.0 

 

Eh ,you are partially right, not all of them though. Gamers Nexus had way different results though

 

 

However. why is the R7 1700 only at 3.9ghz? Does it really only overclock that high? Wouldn't these benchmark be insane if the R7 1700 were at 5GHz? 

 

How high can the R7 go?

[Valentyn]

5 minutes ago, MakeAMDGreatAgain said:

Eh ,you are partially right, not all of them though. Gamers Nexus had way different results though

 

However. why is the R7 1700 only at 3.9ghz? Does it really only overclock that high? Wouldn't these benchmark be insane if the R7 1700 were at 5GHz? 

 

How high can the R7 go?

 

That's the limit he hit with the Noctua NH U12S cooler. I would hope it could go higher with better cooling; as he mention it hit 74 degrees under AIDA64.

We might need to wait for better motherboard BIOS updates and the like before being able to see it pushed higher, although isn't the current LN2 max 5.2Ghz?
Could just mean early Ryzen it topped out at 4.1 or 4.2Ghz top on end user cooling.
 

Personally since I use my computer as a workstation, and for gaming I tend to only overclock conservatively. 4Ghz on my 5820K that maxes out temps at 60 degrees celsius under AIDA 64.

 

[Curious Pineapple]

10 minutes ago, zMeul said:

you, 8minutes ago:

 

Please, point out exactly where I claim that the benchmarks are skewed.

[HelplmChoking]

I think the most important thing to keep in mind is that the current 8 cores aren't meant to compete with Intel's 4 core "gaming" CPU's. When the 6/4 core Ryzen chips come out we'll likely see better performance and higher clock speed.

[zMeul]

4 minutes ago, Curious Pineapple said:

Please, point out exactly where I claim that the benchmarks are skewed.

no?! that's not you're implying when saying:

Quote

I would be very surprised if MS didn't put the work into the compilers to detect CPU manufacturer/model and use hardware features that are available to make the end result better. Not really a hard task.

what then?

[Trixanity]

1 minute ago, Curious Pineapple said:

Please, point out exactly where I claim that the benchmarks are skewed.

He thinks because you say that a compiler can optimize for each and every architecture it's skewed because it's optimized for a specific one instead of being completely oblivious to the architecture you're compiling for. So I guess he's ultimately saying that you're claiming compilers favor Intel.

[Curious Pineapple]

9 minutes ago, zMeul said:

no?! that's not you're implying when saying:

 

Do you also believe the earth is inside out because you can fold a flat map into a sphere either way?

 

You are looking for things that are not there.

8 minutes ago, Trixanity said:

He thinks because you say that a compiler can optimize for each and every architecture it's skewed because it's optimized for a specific one instead of being completely oblivious to the architecture you're compiling for. So I guess he's ultimately saying that you're claiming compilers favor Intel.

I think so too. Assuming every program used was compiled with the same compiler, and the compiler includes several different variants of machine code into the one executable, it's unlikely to have compiled anything that can take full advantage of Ryzen, and at best it will use an AMD optimised variant that was built for Bulldozer. It's not possible to compile high performance applications that will run on any x86 CPU without optimisations anymore. Even Intel's architecture changes significantly between generations. Code optimized for a Core 2 processor will be far from optimal on a Core i, and the 2 generations will handle the same "generic" code very differently. You wouldn't do in software what can be done in 2 CPU cycles. A 2 cycle instruction for example on an i7 may replace a routine that takes 500,000 cycles to complete without the hardware support. A good toolchain should work this out, include both, and choose which code to run based on the CPU.