-
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.
Haswell E review from Anandtech: 5930K, 5960X, and 5820K
- Thread starter LordOfChaos
- Start date
Kensairabbit37
Banned
Although I'm excited, I'm deeply concerned nothing much will support >4 cores for a good while. :/
Insane Metal
Member
Damn, Intel has been on a roll for many years. The "i" series is awesome. Imagine the PS4 or XBO with one of these. Even slowed down. The console would be a monster. And cost $1000, lol
LordOfChaos
Member
Yeah, the issue was likely both die size adding to the cost, and Intels pricing structure. Intel and Nvidia both maintained too much control over their architectures, making them less than ideal partners for consoles. But yes, even four lower clocked Haswell cores would make either console that took them sing. I think 8 jaguar cores may come close to the total performance, but 4 cores is easier to use in their entirety without scaling issues.
Is it me or is that memory controller larger than usual?
Here's hoping that the two leading consoles this generation being 8 cores may make highly threaded games more commonplace. That, and 14nm Broadwell may make octacores more accessible for everyone.
I was planning on a new build in November, once all the dust settled with new cpu / gpu's, but it's looking like I will just hold off. My 2 sticking points in my current machine (2600K Sandy Bridge / 680gtx) is the lack of true pci-e 3.0 features, and the gpu's 2 gig Vram. I'm certain it will still hold it's own going into the next cycle, albeit I won't be able to run all the bells and whistles.
I've still some OC'ing room yet, so I'll eek out every bit I can.
I've still some OC'ing room yet, so I'll eek out every bit I can.
Insane Metal
Member
Yep. But that has made them both advance significantly technologically speaking. Specially Intel. AMD has absolutely nothing to face their i architecture. Which is very unfortunate.
pffffthahahaha
I'm sorry, I don't think that would be the case
I think he's saying that, at the same frequency, the chips would be close in performance.
Kensairabbit37
Banned
So say we all. I'm probably gonna hop onto Broadwell once it arrives. I heard what's coming inititally in Q1 2015 is all the low power chips?
LordOfChaos
Member
pffffthahahaha
I'm sorry, I don't think that would be the case
Adjusted for the clock speed, half the performance (so 8 in total come close to 4) isn't a crazy thought.
AMD A4-5000 (1.5GHz Jaguar x 4) 1323 in 7Zip single threaded
Intel Core i5-3317U (1.7GHz IVB x 2) 2816
And 1.5 vs 2.39 in Cinebench, which if you double Jaguar would actually put it OVER Intel. So saying half the performance is a fair judge.
So I'm pretty close (even slightly over if adjusted for clocks?), nothing to laugh at.
http://www.anandtech.com/show/6974/amd-kabini-review/3
Yep
Yeah. Maybe, I should cancel my 5930K order and go with a 5820K instead. I don't plan to go more than 2-3 GPUs.
LordOfChaos
Member
See my post two above
efyu_lemonardo
May I have a cookie?
as long as operating systems do there's always multitasking..
I wouldn't. 3 GPUs could definitely benefit from the extra PCI-E lanes with the newer cards coming out.
Plus if you are spending the extra premium on a X99 setup, which is significant with the higher priced mobos and DDR4 memory, I see little reason not to spend a few hundred extra to get all the PCI-E lanes.
Zombie James
Banned
Haswell-E is a different market segment so I doubt it.
The Witcher 3 devs mentioned that both consoles allow devs to access 6 cores and that their goal was to "make the [Witcher 3] engine scalable." So it looks like there's already effort being put into taking advantage of extra cores if you have them.
So...At equal clock an Haswell I5/I7 is two times faster ? 4 times if we take an Haswell I5/I7's clocked at 3.4ghz ?
But then there is this quote by Sebbi
http://forum.beyond3d.com/showthread.php?t=64968&page=24
Felix Lighter
Member
I have time to digest this. Now I'm just waiting for Nvidia's cards to be announced and then it's time to build.
SneakyStephan
Banned
shitty netbook cpu (the i5 3317U) does not have the same ipc as the desktop haswell I believe.
Also the full amd desktop fx cores (not the netbook jaguar ones) are less than half as fast as a haswell i5 desktop core at the same clockspeed, they are still slower (still less than half as fast) even at 5ghz compared to stock i5 4670k
You are massively overestimating the amd cpus
edit: I was right:
Passmark cpu benchmark score:
i5 3317u netbook cpu, 4 cores @1.7 ghz : 3100
Intel Core i7-4770K desktop cpu, 4 cores @ 3.50GHz 10,280
core for core, clock for clock, the desktop i5 is 1.6 times faster than the notebook i5
So even if the 8 amd jaguar cores were clocked equally to a desktop haswell quad core, the haswell quad core would still be 3.2x faster (not 2x)
But with jaguar at 1.7 ghz a quad core haswell cpu at stock clocks is about 6.5x faster, once you overclock it it's another 30 percent so almost 8.5x faster
There is a reason why the desktop amd fx 8 core cpu needs to be clocked at 5ghz to match a quad core ivy bridge i5 at 3.5ghz.
Anyhow, now that misinformation was corrected maybe we can discuss the article and the 6-8 core haswell-e instead of having these dumb jaguar comparisons
again. It's pointless they are more than a full generation apart.
edit: silly me, I can just look up the jaguar's passmark score to compare them without doing this crude math assuming performance scales 100 percent with clocks for either.
AMD A4-5000 APU 1,905
so a stock desktop haswell quad core is 5.4x faster
We would have to pay problably over a thousand dollars for the console and it would have huge TDP (therefore it would probably be bigger than Xbox One).
Sure, they could've done that. But then we would witness the biggest catastrophe in history of SCE, easily overshadowing the PS3.
I don't think you've thought this through.
What are the performance gains in games and Dolphin?
There is a test in overclocking section.
on the one hand i'm glad to see Intel move the CPU power yard stick down the field. On the other hand in terms of Gaming performance anything beyond an i5 is a waste, and things are only going to get more GPU bound with Direct3D 12/Mantle/Next Gen OpenGL. not to mention using GPU for compute tasks. If you doing (a LOT) video encoding maybe but otherwise its just an epeen thing
SneakyStephan
Banned
Anandtech are idiots btw, benchmarking games at gpu limited settings, and on an sli set up to boot, why can't they just set it to low@800*600 and get meaningful data for the cpu results.
Zombie James
Banned
What are the performance gains in games and Dolphin?
Because it's as far from relevant as you can get. Nobody plays at those resolutions and it's not indicative of anything anyone would actually do.
SneakyStephan
Banned
Benchmarks are hard
the point of gaming benchmarks is to see how much faster the cpu actually is in games, which you can only measure if you don't have a gpu bottleneck.
It doesn't matter if you aren't going to play that specific game at 800*600 ,the point is to see how much better it does in games so when future games come out using the same engine or you upgrade your gpu later on you know if the cpu is more capable than what you have or not.
If you test it at gpu limited settings it's a gpu benchmark not a cpu benchmark.
You turn off /down all gpu reliant settings so you can test the cpu reliant ones and see how the cpu performs.
Potential buyer has 2 titans and a 144hz monitor
He doesn't give a shit about a benchmark showing him that those 1 or 2 770s run tomb raider at 50-90fps at 1080p with 4x msaa and tressfx on high, he needs to know if he can get 144 fps with this cpu when he can't with his current cpu.
This worthless anandtech benchmark tells him nothing (it only tells him how well a 770 performs in tomb raider), other than that anandtech are a glorified product overview site rather than a respectable review site.
Might as well look on newegg or amazon product overview page for a 'review'
the point of gaming benchmarks is to see how much faster the cpu actually is in games, which you can only measure if you don't have a gpu bottleneck.
It doesn't matter if you aren't going to play that specific game at 800*600 ,the point is to see how much better it does in games so when future games come out using the same engine or you upgrade your gpu later on you know if the cpu is more capable than what you have or not.
If you test it at gpu limited settings it's a gpu benchmark not a cpu benchmark.
You turn off /down all gpu reliant settings so you can test the cpu reliant ones and see how the cpu performs.
Potential buyer has 2 titans and a 144hz monitor
He doesn't give a shit about a benchmark showing him that those 1 or 2 770s run tomb raider at 50-90fps at 1080p with 4x msaa and tressfx on high, he needs to know if he can get 144 fps with this cpu when he can't with his current cpu.
This worthless anandtech benchmark tells him nothing (it only tells him how well a 770 performs in tomb raider), other than that anandtech are a glorified product overview site rather than a respectable review site.
Might as well look on newegg or amazon product overview page for a 'review'
Yep.
People have long abandoned the silly low resolution tests for CPUs because they're literally meaningless.
Also, better benches from Techreport:
Not a single surprise there. IPC still king.
Bottlenecks don't work like that. They can exist on all parts at the same time. The information is still very useful. They are generally more engine dependent than "settings" dependent.
Rentahamster
Rodent Whores
SneakyStephan
Banned
it's like the real mkeynon got kicknapped by aliens
read my bit on the 120-144 hz monitors
testing games with settings noone with such a monitor will use (tressfx high etc) is stupid, because they care about getting the framerate they desire (which is why they'd be looking at a high end cpu to begin with)
gpu limited benchmarks are completely worthless.
frametimes and minimum fps are better metrics to get a feel for cpu performance in a gpu limited scenario, but what you need is minimum fps and frametimes in a cpu limited scenario
What do you mean by this?
Because, the benches I posted above, especially for Watch Dogs and Batman, are both CPU limited but for different reasons. Batman is on UE3 and sings with two speedy as fuck cores.
Can Intel's new chip improve gaming performance?
Verdict
"A super-powerful, octo-core CPU, but has little to really offer in gaming performance."
Full Article: http://www.pcgamer.com/review/intel-core-i7-5960x-review/ (link is external)
Verdict
"A super-powerful, octo-core CPU, but has little to really offer in gaming performance."
Full Article: http://www.pcgamer.com/review/intel-core-i7-5960x-review/ (link is external)
Yeah, games are still really all about IPC. As long as you have a SB or newer at 4.5Ghz, there's going to be an imperceptible difference in performance between any Intel proc in any game.
Wondering when we will finally see a game that properly takes advantage of 6 or above multithreated cores.
There are some already, Civ V is n-threaded.
The rule of thumb right now is basically that the more threads a game uses, the less reliant it is on CPU performance in terms of affecting average frame rate. (assuming a modern quad core is in use). Even still, games that are well known to be multi-threaded, like Crysis 3, are still all about that IPC. The Intel Pentium Anniversary at 4.8GHz keeps up with the 2500K and 8350.
SneakyStephan
Banned
When you turn off 4xmsaa and hbao+ and tressfx and whatever other fps destroying settings that are 100 percent gpu dependant and that noone with a 120hz monitor is going to use if it prevents them from getting the 120fps they desire.
So you turn anything gpu related (resolution, aa, ao,physx, tressfx etc) off/to low so it doesn't interfere with the results of the benchmark and you can see what fps the cpu can achieve in that game or engine.
People can decide on what gpu and settings they use for themselves, they don't need a benchmark to guess for them.
here's what's relevant to performance: (aka to how well and smoothly a game will play)
-minimum framerates caused by CPU
-average fps enabled by CPU
-minimum framerates caused by gpu
-frametime variance caused by gpu
-average fps enabled by gpu
This is the information you need to decide if there are any bottlenecks in your system today, and wether there will be any bottlenecks in future games or when you replace either the cpu or gpu. And wether you are buying a cpu that is overkill for your gpu or vice versa.
the first 2 are represented in a good cpu benchmark, the last 3 in a good gpu benchmark (pcper for example for gpu benchmarks)
If someone wants to know how their gpu performs at max settings in a game they can look at a gpu benchmark for it. it's irrelevant to the cpu benchmark so it doesn't matter if you test the game at 800*600.
These stupid gpu limited benchmarks for cpus (as well as product overview sites like andandtech rarely bothering to test actual cpu limited games like guild wars 2 , arma or sr4 or planetside 2 or ns2 etc) is why people ended up making the horrible mistake of buyig an amd fx8350 for gaming...
Unless you are alien spy mkeynon you surely will agree with this last point. (fx8350 missbuys due to shitty misleading reviews that leave information out)
obfuscating the results is the exact opposite of what a benchmark is intended to do or what a review is supposed to accomplish.
also on a totally different topic: anyone notice how minimum fps in bioshock infinite is STILL 20 fps even on a haswell-E and sli gtx770? that game was such a disgrace
Just look at any frametime percentile graph on pcper to see how that game should have never passed QA.
I don't get why any review site dignifies the existance of that monstrosity
SpecialAgentZ
Member
I think the i5 3317U is a Dual Core CPU, not quad.
GameAddict411
Member
the CPU in the OP is more than half the size of the XO APU(which houses the GPU and the CPU) + the eSRAM, coupled with the much higher clock speed and you have a monster. it would probably fry the XO power adaptor and melt the plastic lol. these types of hardware is just impossible to integrate in consoles, regardless of the costs. even a much smaller quad core processor from intel will cause problems from very high power consumption to heat and size of the hardware. this isn't 2005 anymore where top end hardware didn't get so hot, the X360 suffered high failure rates during its first couple of years and thats with 2004 technology. people need to understand that costs weren't the only concerns when they developed the current gen of consoles.
This is what I'm saying though, CPU and GPU bottlenecks can (and do) happen at the same time on both parts for different reasons.
Playing with all those high settings is certainly going to show you an average frame rate that is being heavily limited by the GPU. However, by looking at frame time analysis, you will also get a perfectly good picture of what is going on in terms of limiting CPU performance.
The "Frames Above" chart for Batman perfectly illustrates this with the 5960X.
It doesn't take a genius to extrapolate that data to get an idea of how it will affect performance on various setups. Not only that, but it gives you a much clearer picture of how decent hardware in the rest of the system will affect things with this one part switch.
Figuring the data from a standard benchmark done with a single 770 with max settings is far more realistic for determining performance on a 780 with lowered settings than it is with the settings turned way way down at a low resolution.
GameAddict411
Member
the consoles would have used laptop CPUs anyway, even if they went with intel as the supplier. the core i7 in my laptop even after two years eats through anything i give it. from games to many other intensive things. gaming is much more dependent on the GPU.
SneakyStephan
Banned
It's still incomplete information (shows you part of the fps drops caused by cpu,but does not represent the performance during all of the time where the performance is purely gpu limited , at whatever arbitrary settings and gpu the reviewers have decided on)
-which brings me to another point, imagine the 2500k reviews benching all games with a hd4870 at max settings, who in 2013 buying one new or second hand i5 2500k and buying a hd 7970 ghz edition would give two shits about how it performs in crysis with a 4870, their gpu is 4x faster the game running at 35 fps tells them NOTHING about the cpu, they could compare it to an athlon 2 benchmark and it would look just as good an option for their 7970ghz than the i5)
and as for the bolded: really? the majority of people in the new gpu threads etc on gaf are not able to correctly read or interpret benchmarks or put them into context.
Why make it even harder for them...
(unless you're a product overview site and not a review site)
@ the underlined: again who cares about how a gtx770 performs? you can look up gpu benchmarks for a gtx770 for that info, it's totally worthless in a cpu benchmark.