Dictator93
Member
It's science.
-
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.
Anandtech: Intel's new Atom CPU beats AMD's Jaguar in performance
- Thread starter Horse Armour
- Start date
It's science.
JonathanPower
Member
It's 2.5W versus an estimated 6W. Once Jaguar will move to 22nm it should have a similar, if not better, TDP.
Yeah, a little bit more considering the higher clock speed. However, we still have no info regarding the final PS4 CPU clock speed.
That's not an issue if you can provide those yourself. Raw benchmark data is exactly what I want when reading about a new piece of hardware -- let me do the interpretation.
maverick40
Junior Member
I am an engineer in Intel. Worked on this Atom processor and now just starting on the newly revealed quark chip which will be used for wearable technology! Intersting times ahead for the company
DeFiBkIlLeR
Banned
Don't forget the PS4 will only have 6 cores available for gaming.
People really need to start thinking of the PS4 as a 6 core 6GB console for games, not 8 Core 8GB.
Stallion Free
Cock Encumbered
The devs did say that Infinity Blade 3 was like next gen so I guess they were right.
Because when you run a game on a desktop it runs on those magical OS processors right?.
Stallion Free
Cock Encumbered
Yes?
maabus1999
Member
Before heavy optimization, it would not surprise me at all if most games hit the CPU wall first on both consoles; it isn't the strongest beast in the world.
Now after they optimize their code, they (hopefully) will be fine and get more out of the GPU in the process. But those who don't and get lazy will just hit that CPU wall.
Gemüsepizza
Member
While the CPU might be remotely comparable, the GPU is not. So this is not enough to bring such games to a true next gen level.
Stallion Free
Cock Encumbered
But dat retina displayyyGemüsepizza;82470253 said:
AdmiralPain
Banned
So I' guessing that these Atoms are quite cheap, right? Since they can at least match Jaguar performance, what are the chances that a higher-clocked Atom ends up in the Steambox in order to keep costs down?
I only heard them mocking the GPU (with some ridiculous comparisons to power hungry chips). Why would nVidia comment on CPU anyway?
Very unlikely. One thing what devs will dislike though (but will have to do anyway) would be to embrace multi-threaded programming.
JonathanPower
Member
We have no confirmation about this either. In addition, it was also rumored that PS4 reserves less than 1 core for the OS.
http://www.neogaf.com/forum/showpost.php?p=47871129&postcount=1620
Is it x86 windows 8 or ARM?
No, that's not correct. Only in a minor few synthetic benches that exist to load up cores. More cores is not a linear increase.
You think FCAT works on consoles? I'm very seriously thinking about making an FCAT setup if one could.
There was a thread about it. They said PS4 had a low end CPU and a mid range GPU.
This is accurate, though.
JonathanPower
Member
In fact, we are talking about the Cinebench test here. So yes, an 8 core Jaguar should perform twice as good as the new Atom in that same bench.
Not in the 8 core configuration. The 8 core Jaguar will not be used in low end devices.
Let me put it this way.
If you have a 4 core CPU that can be used in tablets and cellphones, and is even outclassed by other CPUs that are used in tablets and cellphones, then doubling the core count just makes it an 8 core low-end CPU.
When those 8 cores could be outclassed by a single core from a 2500K, it is a low-end CPU.
Just talking about Cinebench is silly. There's a reason why they run multiple benches. But yes, it will have a little under twice the performance in cinebench.
If you have a 4 core CPU that can be used in tablets and cellphones, and is even outclassed by other CPUs that are used in tablets and cellphones, then doubling the core count just makes it an 8 core low-end CPU.
When those 8 cores could be outclassed by a single core from a 2500K, it is a low-end CPU.
Just talking about Cinebench is silly. There's a reason why they run multiple benches. But yes, it will have a little under twice the performance in cinebench.
lyrick
Member
Anandtech Posted their iPhone 5c & 5s reviews and the A7 is trading a few blows with the Atom z3770 in the CPU benchmarking sections.
http://www.anandtech.com/show/7335/the-iphone-5s-review/5
So... Is the PS4 & Xbox One CPU already falling behind the iPhones?
http://www.anandtech.com/show/7335/the-iphone-5s-review/5
So... Is the PS4 & Xbox One CPU already falling behind the iPhones?
JonathanPower
Member
Are we talking about gaming performance here or general performance when a CPU is used for non-gaming applications? Because if it's the latter then yeah, the new Atom may be faster at loading a web browser; however, if it's the former, then what really matters is whether the new Atom is faster than Jaguar in gaming applications. And in this regard, the Cinebench test is surely more relevant than the web browser test.
Why?
*edit*
Also, besides this, you're still talking about whether a bottom barrel chip is faster than another bottom barrel chip, which doesn't really discount the fact that we are talking about low-end CPUs here, which was the original point being discussed
It's not bad that it's a low end CPU. Maybe next gen engines are doing something special which will reduce the typical CPU bottleneck that currently exists on all UE3, Blizzard, Source, BF3 MP (or really any MP game) and other single/dual threaded engines.
Though, that does make me a bit worried about Titanfall.
Yeah, I'm not surprised. Thing is though, PS4/X1 games, "are," going to be heavily multi-threaded, by necessity. While the single thread performance of a single PS4 core will be sufficient for some genres, any genre that'd normally strain a CPU is going to have to have it's engine designed to spread the load across as many cores as possible, in order to pull out the performance it needs.
This is actually a good thing, seeing as a number of engines over the years, still haven't made efficient use of multi-core CPU's.
This is actually a good thing, seeing as a number of engines over the years, still haven't made efficient use of multi-core CPU's.
I really hope that's the case. To give you guys some perspective, a 2500K 4C/4T (3 year old $200 part) is within spitting distance of gaming performance you get out of a $1000 6C/12T proc released two weeks ago. This is because games just do not care as it is.
It's a win:win for PC gamers and consoles alike.
It's a win:win for PC gamers and consoles alike.
cartman414
Member
Cool.
I know you're just one individual there, but when are you guys planning on pushing performance ceiling for the consumer market again? I never got an actual response when I e-mailed the customer department.
Not half as likely as one of the R processors. Take a look at these:
http://techreport.com/news/25377/gigabyte-brix-crams-gaming-pc-into-well
Not to mention that Atom is merely up to the IPC level of Westmere and only has 4 EU on the IGP side.
JonathanPower
Member
Why?
*edit*
Also, besides this, you're still talking about whether a bottom barrel chip is faster than another bottom barrel chip, which doesn't really discount the fact that we are talking about low-end CPUs here, which was the original point being discussed
It's not bad that it's a low end CPU. Maybe next gen engines are doing something special which will reduce the typical CPU bottleneck that currently exists on all UE3, Blizzard, Source, BF3 MP (or really any MP game) and other single/dual threaded engines.
Though, that does make me a bit worried about Titanfall.
The problem is that I disagree with your definition of low end. You said in your previous post.
Now, a single 2500K core gives you around 50 GFLOPs at 3.2GHz. An 8 core Jaguar gives you 102.4 GFLOPs at 1.6GHz. So I completely disagree that 50 GFLOPs are better than 102.4 GFLOPs, even if the i5 is a much stronger CPU than Jaguar.
I have read the source article of that graph and those numbers are accomplished by running cinebench in four threads, this will of course limit the performance when the fx-8350 is an octocore processor.
You could try to find an article that displays bad cinebench scaling.
Dude. C'mon. Turbo boost is at 3.7, just about any gaming PC has them at 4.2-4.5, with 4.6-5.0 even being fairly common.
And again, we are talking about a single core on a 3 year old processor.
*edit*
You also dodged the question of why you think cinebench is a more accurate representation of gaming performance. You seem to be playing word/forum games rather than looking to have a conversation, so I think I'm done here.
JonathanPower
Member
Even at 3.7GHz you have 59 GFLOPs, which is still not better than 102 GFLOPs. Talking about overclocking does not make sense in this context because you can't overclock the CPU in a console.
Cinebench is a 3D rendering bench and is a FPU heavy application. Surely it's more relevant for gaming than a web browser test.
There's two games I can think of off the top of my head that are n-threaded. So, it's not really accurate. At least not right now with any empirical evidence we have to go on.
Thats a 6C 12T proc getting bested by 4C/4T and 4C/8T procs. Even a 2C/4T proc is within spitting distance.
Turbo boost, by it's very nature, is overclocking. It's in fact part of the reason the 2500K is such a great gaming processor. It overclocks itself when playing games, most of which are 2 threads or less.
Regardless, you've dragged the conversation out this far to insist that the Jaguar 8 Core CPU isn't low end. I'm not sure what a single post following that disagreement has done to prove that. WTF is going on here?
*edit*
Also, for reference, here's the same processors on cinebench.
*edit 2*
Now compare scores of this:
With even 100% scaling, you're still below the performance of a low-end desktop part released a year ago. That's not even looking at single/dual threaded performance, which is still very important for examining gaming performance.
It's low-end. Accept it.
JonathanPower
Member
No. An 8 core Jaguar at 1.75GHz, which is the Xbox One CPU clock speed, should score around 3,5 at Cinebench. An 8 Core Jaguar at 2GHz should score around 4, which is more than the FX-4350. However, we don't know what is the PS4 CPU clock speed.
Even if the PS4 clock speed was 2ghz, it would still be a low end CPU.
You can't make up numbers as an argument.
*edit*
Are you trolling me? If so, well done.
JonathanPower
Member
I am just estimating the score of the 8 core Jaguar. Jaguar@1GHz scores aroung 1; Jaguar@1.4GHz scores around 1.4; Jaguar@1.5GHz scores around 1.5. So I think we can assume that an 8 core Jaguar@2GHz should score around 4.
JonathanPower
Member
Sure, but then we should consider the FX-4350 a very low end CPU.
You don't judge a CPU's performance by a single synthetic benchmark though.
*edit*
And the FX-4350 is a low end CPU.
JonathanPower
Member
By low end I mean something with the performance of what you find in low end devices like Tablets and Ultrabooks. The FX 4350 is still a pretty powerful processor, and thus I consider it at least mid end.
dr. apocalipsis
Banned
Sweet.
As I said, an octocore Bulldozer derivative CPU only have 4 FPU's.
But then...
Jaguar is a low end CPU because it have abysmal IPC. You can stack as many cores as you want to get good multithread scores in benchmarks, but that doesn't change that fact.
No matter how well optimized your code is, it will run as fast as the slowest thread. And games need fast IPC over multiple threads. There are a lot of sub-routines that need good IPC, so I'm actually worried about how next gen consoles CPUs willl hold next gen games innovations beyond graphics.
FFS, dual core i3s can beat this things easily even in heavily multithreaded scenarios, how can you affirm that they are more than just low end parts? CPU isn't about GFLOP additions. CPUs have to do the job in a lot of different heterogeneous workloads, and Jaguar is far from that.
This.
blu
Wants the largest console games publisher to avoid Nintendo's platforms.
Durante, that paper is truly amazing. I just hope you have no part in it. I mean, I read it up to this paragraph I'll quote here, for coder's gaf amusement:
"Compiler: Our toolchain is based on a validated gcc 4.4 based
cross-compiler configuration. We intentionally chose gcc so
that we can use the same front-end to generate all binaries. All
target independent optimizations are enabled (O3); machine-
specific tuning is disabled so there is a single set of ARM bi-
naries and a single set of x86 binaries. For x86 we target 32-bit
since 64-bit ARM platforms are still under development. For
ARM, we disable THUMB instructions for a more RISC-like
ISA."
Using gcc 4.4 (in a research from 2013..) with no machine-specific optimisations across widely varying architectures (i.e. in-order and out-of-order) to draw conclusions about performance.. Using x86 on amd64 machines, under the excuse 64-bit ARM is still unavailable (nevermind it's using a known 32-bit op encoding, and is already supported in emulated form by well-established toolchains), while refusing to use Thumb encoding on the ARM, to draw conclusions about the code density.. Amazing.
JonathanPower
Member
Wrong. Jaguar IPC is similar to Piledriver. And Piledriver is the core used in high end AMD CPUs.
http://beyond3d.com/showpost.php?p=1667815&postcount=48
That is the fucking x86 decoder, (as in the max instructions per cycle that it can decode into micro ops) that is not the CPUs IPC when processing a load!
Dictator93
Member
It is as if people can convince themselves things that are very much so not as they imagine.
The new console CPUs are very much so lowend... even with 8 cores.