Vishera Review Up - Anandtech

[bononos]

Actually if you are on a deadline or a schedule with encoding every minutes may count, whereas FPS differences are not really important, and the CPU may never become a bottleneck here, provided you chose video card+display resolution wisely......
......
Let's not forget the bad Ivy Bridge thermals, useless yet mandatory IGP graphics, OC restrictions, Intel's mainboard chipset (and RAM support) policies, and the heat density wall that CPUs hit on 22nm.

There are games as shown in the benchmarks where fps differences are still significant - around the 40-60fps area.

Your remarks about bad thermals is strange if you want to place the Vishera as a serious workhorse cpu. If your Vishera is going to be important to beat time pressures of work, why are you going to overclock it at all? Why bring up the uselessness of the intel igp when the Vishera has none?

 

[bononos]

Sorry for quoting a post so early in the thread, but this bothers me. Yes, I don't much care either when I'm playing Sims 3 and I get 400fps with my i5 instead of 300fps. However, I do care that when playing Guild Wars 2, I get 60+fps instead of 40 because it's such a CPU heavy game, and doesn't take advantage of 8 cores. I do care when I'm late game in a Starcraft 2 mod and I get 30fps instead of 20. I also care when I'm playing Mario Galaxy 2 in Dolphin that it stays pegged at 60fps instead of dropping into the 40's when there's a lot going on.
......

Its not just the average fps that matter. I forgot which site had that article that measured the number of times some games failed to render a certain frame within a number of milliseconds - which would cause a break in realism even if it didn't cause outright stuttering which would be larger number of consecutive frames which failed to render smoothly - and the bulldozer didn't fare that well. I wonder how the Vishera does here.

 

[Allio]

Its not just the average fps that matter. I forgot which site had that article that measured the number of times some games failed to render a certain frame within a number of milliseconds - which would cause a break in realism even if it didn't cause outright stuttering which would be larger number of consecutive frames which failed to render smoothly - and the bulldozer didn't fare that well. I wonder how the Vishera does here.

Hardcoreware tested this and found a substantial improvement over Bulldozer in every game they tested, with the exception of Skyrim (which seems to just flat-out suck on every AMD CPU). Obviously the 3570k isn't threatened, but the 8350 actually manages to edge it out in Crysis 2.

 

[know of fence]

There are games as shown in the benchmarks where fps differences are still significant - around the 40-60fps area.

Your remarks about bad thermals is strange if you want to place the Vishera as a serious workhorse cpu. If your Vishera is going to be important to beat time pressures of work, why are you going to overclock it at all? Why bring up the uselessness of the intel igp when the Vishera has none?

To embed a picture to the benchmark of 40-60 FPS area games put [lMG] [/IMG] tags around the URL or just post a link.

Bad Ivy Bridge thermals mostly refer to strange temperature differences between cores and also this http://forums.anandtech.com/showthread.php?t=2261855

Vishera is better/equal at encoding stuff. What is there to argue? Higher load temps and consumption are no trouble here, as long as idle consumption is about the same. Throwing money at the problem also helps with encoding speeds, if you happen to subscribe to this philosophy. (LGA 2011, dual - octacore E5 Xeons here we go, or how about shelling out four grand for the 2*6 core Apple™ Mac bro™)

One of a few ways to waste money on hardware is to pay for features you're never going to need. While Ivy Bridge boasts a 17% bigger transistor count, how many of those are committed to the IGP? - About this many.

 

[Olikan]

32 to 22 isn't going to knock off half their consumption, they need a better design.

not sure... nehalem a used a lot of power too, it was sandy that made the arquitecture very efficient

Still ~20%. That's not going to be addressed until they widen the front end in Steamroller.

it would be nice to see steamroller with a core disabled, and see the "new" CMT penality

 

[Idontcare]

To embed a picture to the benchmark of 40-60 FPS area games put [lMG] [/IMG] tags around the URL or just post a link.

Another way is when you click the "Go Advanced" button at the bottom of your post prior to submitting the post then you are taken to a screen that has a toolbar across the top of the post.

In that view you can simply click the insert image icon

and give it the URL there, then VB will embed the image without you having to add image tags manually.

Bad Ivy Bridge thermals mostly refer to strange temperature differences between cores and also this http://forums.anandtech.com/showthread.php?t=2261855

The strange temperature differences are unavoidable consequences of the die layout not being symmetric, it changes the lateral heat diffusion for the cores:

 

[Chiropteran]

My newly redesigned Core CPU, to fix the heat issues.

Intel: I will offer you rights to my design for a mere $1 per CPU built.

CORE1--------GPU--------CORE2
CACHE---CACHE-CACHE---CACHE
CORE3---MemController---CORE4

 

[Yuriman]

My newly redesigned Core CPU, to fix the heat issues.

Intel: I will offer you rights to my design for a mere $1 per CPU built.

CORE1--------GPU--------CORE2
CACHE---CACHE-CACHE---CACHE
CORE3---MemController---CORE4

lol

Believe it or not, there are good electrical reasons for pushing the cores together, but I'll admit it would be nice to shave a few °c.

 

[ShintaiDK]

My newly redesigned Core CPU, to fix the heat issues.

Intel: I will offer you rights to my design for a mere $1 per CPU built.

CORE1--------GPU--------CORE2
CACHE---CACHE-CACHE---CACHE
CORE3---MemController---CORE4

I would settle for 0.01$ a pcs. Would yield 3½mio$ a year or so

 

[know of fence]

Thanks, I actually only had "basic editor" set in Miscellaneous options :thumbsup:

The strange temperature differences are unavoidable consequences of the die layout not being symmetric, it changes the lateral heat diffusion for the cores

With a dual-core i3 I still get a ~15 degree difference. Is there any way to test this (beautiful) explanation, by clocking down the memory controller maybe. Or using torture tests that don't hammer the memory controller. (I have to double check if using "small FFTs" in Prime95 will change anything, though i'm pretty sure I did that already.)

I didn't mean to bash Intel, just pointing out the differences. Back on topic: Piledriver layouts, so nice and symmetrical.

 

[hokies83]

To embed a picture to the benchmark of 40-60 FPS area games put [lMG] [/IMG] tags around the URL or just post a link.

Bad Ivy Bridge thermals mostly refer to strange temperature differences between cores and also this http://forums.anandtech.com/showthread.php?t=2261855

Vishera is better/equal at encoding stuff. What is there to argue? Higher load temps and consumption are no trouble here, as long as idle consumption is about the same. Throwing money at the problem also helps with encoding speeds, if you happen to subscribe to this philosophy. (LGA 2011, dual - octacore E5 Xeons here we go, or how about shelling out four grand for the 2*6 core Apple™ Mac bro&#8482

One of a few ways to waste money on hardware is to pay for features you're never going to need. While Ivy Bridge boasts a 17% bigger transistor count, how many of those are committed to the IGP? - About this many.

You figures only work when comparing a non over clockable 3770 to a 8350..

I could get a i7 920 and overclock it to and say it was equal to a 3770..

What happens when you overclock the 3770k?

The 8350 gets Crushed all the way down the board and it is not even close.

How do i know? Cause ive been running what those 8350 club boys have been running benchmark wise and destroying them by a Landslide ....

AMD put a big Stock clock on that cpu to fool newbies to think it was more then it is.... and in truth it is not much at all.. Only equal in performance with a 3 generation old i7 950 clock for clock.

 

[Idontcare]

Thanks, I actually only had "basic editor" set in Miscellaneous options :thumbsup:

With a dual-core i3 I still get a ~15 degree difference. Is there any way to test this (beautiful) explanation, by clocking down the memory controller maybe. Or using torture tests that don't hammer the memory controller. (I have to double check if using "small FFTs" in Prime95 will change anything, though i'm pretty sure I did that already.)

I didn't mean to bash Intel, just pointing out the differences. Back on topic: Piledriver layouts, so nice and symmetrical.

I'm sure you could work the Vdram value as well as the DDR3 clockspeed to push around the heat generation of the IMC to test it.

For me the proof comes in enabling the iGPU. The picture and data above is for my 2600k with the iGPU disabled. When I enable the iGPU the core temp goes up by quite a few degrees (nearly 10C IIRC).

The thing is there is no need to prove the cooling merits and advantages of dark (or gray) silicon relative to the hotspots of active circuits on the die. That was proven decades ago, and layout of today's ICs is done with an optimization effort towards reducing hotspots by getting them closer to lower activity factor circuits which are expected to inherently run cooler.

 

[know of fence]

The thing is there is no need to prove the cooling merits and advantages of dark (or gray) silicon relative to the hotspots of active circuits on the die. That was proven decades ago, and layout of today's ICs is done with an optimization effort towards reducing hotspots by getting them closer to lower activity factor circuits which are expected to inherently run cooler.

I'm sure this is true for the smaller areas, but my scepticism is a question of scale, really. Is the same true for the biggest "core" size scale.
Assuming it is given, that the inner cores are the hottest, then the staggered, thermally regulated Turbo certainly makes sense. It probably even serves to equalize temperatures across cores, by applying the biggest Turbo OC to the coolest core, the second biggest to the second coolest core etc.
Thus if turbo is disabled shouldn't we see even bigger differences, due to cooler outer cores? Shouldn't OC'ers avoid disabling turbo at any cost?

Maybe by the same logic having turbo on the symmetric Piledriver isn't all that crucial, which among other things makes a higher overall baseclock possible.

 

[Abwx]

What happens when you overclock the 3770k?

The 8350 gets Crushed all the way down the board and it is not even close.

Let s take a look rather than spreading non sense , let say
a 4.9ghz 3770K and a 4.5g base/4.8G turbo 8350FX..

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/13

Where would you place your uber alles 2600K if clocked at 4.8G ,
in these charts ??. higher than the 3770K ?...:biggrin:

 

[Phynaz]

Let s take a look rather than spreading non sense , let say
a 4.9ghz 3770K and a 4.5g base/4.8G turbo 8350FX..

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/13

Where would you place your uber alles 2600K if clocked at 4.8G ,
in these charts ??. higher than the 3770K ?...:biggrin:

Lol, a 500Mhz overclock cranked power consumption to 305 watts. Good thing Techreport couldn't get the 5Ghz AMD told them to expect, they'd probably have a 450 watt furnace on their hands.

I had hoped to include a quick Skyrim test to see how the FX-8350's gaming performance is improved by higher clock frequencies, but when I went to test it, our overclocked config wasn't entirely stable.

 

[Abwx]

Lol, a 500Mhz overclock cranked power consumption to 305 watts. Good thing Techreport couldn't get the 5Ghz AMD told them to expect, they'd probably have a 450 watt furnace on their hands.

The point was about perfs with ock. CPUs but since the results
doesnt suit your expectation you re showing your ugly trasher face
that run to elude the subject , not without twisting the numbers ,
of course :

From the TR article :

The overclocked and overvolted config tested above peaked at about 262W

The overclocked 3770K in this test use 244W , that s not better
considering it s 22nm , but , hey , this cant be possible ,
it s an intel CPU , you know , the ones you told us you were
so proud selling them at the time of the P4.....:biggrin:

IDC s review , with a voltage tweaked CPU :

http://forums.anandtech.com/showthread.php?t=2261855

 

[CTho9305]

Thanks, I actually only had "basic editor" set in Miscellaneous options :thumbsup:

With a dual-core i3 I still get a ~15 degree difference. Is there any way to test this (beautiful) explanation, by clocking down the memory controller maybe. Or using torture tests that don't hammer the memory controller. (I have to double check if using "small FFTs" in Prime95 will change anything, though i'm pretty sure I did that already.)

I didn't mean to bash Intel, just pointing out the differences. Back on topic: Piledriver layouts, so nice and symmetrical.

Yeah, but look how much unused space there is, compared to the Intel die. It looks like there's room for another whole 2MB bank of L3 cache between the two chunks on the right of that picture, let alone the other large gaps all over the place. It makes you wonder why they didn't use whatever technique on the 6-core 45nm products allowed the L3 cache to fill in the irregular shapes making up the right side of that die.

 

[hokies83]

Let s take a look rather than spreading non sense , let say
a 4.9ghz 3770K and a 4.5g base/4.8G turbo 8350FX..

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/13

Where would you place your uber alles 2600K if clocked at 4.8G ,
in these charts ??. higher than the 3770K ?...:biggrin:

You make no sense 8350 is equal in performance to a i7 950 clock for clock that is about it...

 

[Phynaz]

The point was about perfs with ock. CPUs but since the results
doesnt suit your expectation you re showing your ugly trasher face
that run to elude the subject , not without twisting the numbers ,
of course :

From the TR article :



The overclocked 3770K in this test use 244W , that s not better
considering it s 22nm

IDC s review , with a voltage tweaked CPU :

http://forums.anandtech.com/showthread.php?t=2261855

244w isn't better than 305w?
I always thought lower numbers were better for power consumption. Yes I know that you want to give AMD a break because Intel's manufacturing capabilities are far more advanced than what AMD has available to them. Sorry AMD doesn't get one, either they perform or they don't. After all you are the one saying it's not fair to compare a 45w Intel CPU to a 35w AMD CPU, right? So of course it wouldn't be fair to compare a 305w AMD system to a 244w Intel system .

So in summary, IDC's 4.9Ghz 3770K uses 20% less power than a 4.5Ghz 8350, and according to Techreport performs roughly 25% faster in workloads that are favorable to the 8350.

Here I thought you were arguing for the 8350. Sorry for the error on my part.

the ones you told us you were
so proud selling them at the time of the P4.....

You keep saying this over and over, and I keep telling you I never sold a P4 system in my life. Not that it would have anything to do with Vishera even if I had.

 

[Phynaz]

The point was about perfs with ock.

From Techreport:

"the FX-8350 is the highest-performance AMD desktop processor to date for gaming, finally toppling the venerable Phenom II X4 980. Yet the FX-8350's gaming performance almost exactly matches that of the Core i3-3225, a $134 Ivy Bridge-based processor. Meanwhile, the Core i5-3470 delivers markedly superior gaming performance for less money than the FX-8350."

Awesome, AMD is now able to out-perform their 2009 CPU lineup.
Great "perfs" they have. You're going to need that 5Ghz "ock" to match a lowly frequency locked i5 in gaming.

 

[Abwx]

8350 is equal in performance to a i7 950 clock for clock that is about it...

You re still under the influence of viral marketing
unless you re performing it yourself or living in denial...

In cinebench the 950 do 5.31 while the 8350 do 6.94
so the 950 peformances should increase by 30% for 30%
frequency increase , wich wont be the case according to
the overclocked benchs wich show about 25% better perfs
for 40% overcolcking.

At even higher frequencies forget your 950...

I posted TR link for overclocked CPUs and we can clearly see
that the intel 3770K do not scale as well as the 8350 , hence ,
there s a lot of people that think that at 4.5G the 950 or
2600K will keep the lead largely...

 

[Abwx]

244w isn't better than 305w?
I always thought lower numbers were better for power consumption.

So IDC's 4.9Ghz 3770K uses 20% less power than a 4.5Ghz 8350, and according to Techreport performs roughly 25% faster in workloads that are favorable to the 8350.

Here I thought you were arguing for the 8350. Sorry for the error on my part.

Still with your twisted numbers ?...
Yet , i did give you a clue above...

From TR whose link is (forcibly) above and apparently
out of reach for your stretching eyes..

The overclocked and overvolted config tested above peaked at about 262W

It s the 8350 they are talking about , so from 262 to 244 there is 20%..??

I understand better the kind of guy you are , shamelessly
and obviously trolling all the way.

In principle such guys do not deserve any answer but for the sake
of the forum i have sometime some mercy even with ruthless trolls....

 

[hokies83]

You re still under the influence of viral marketing
unless you re performing it yourself or living in denial...

In cinebench the 950 do 5.31 while the 8350 do 6.94
so the 950 peformances should increase by 30% for 30%
frequency increase , wich wont be the case according to
the overclocked benchs wich show about 25% better perfs
for 40% overcolcking.

At even higher frequencies forget your 950...

I posted TR link for overclocked CPUs and we can clearly see
that the intel 3770K do not scale as well as the 8350 , hence ,
there s a lot of people that think that at 4.5G the 950 or
2600K will keep the lead largely...

Your way off dude... i had a i7 920 at 5ghz and i was touching high 7 scores in cinebench back in the Day..

3770k Scales like a Monster when overclocked.. Just look at mine...

x264 HD BENCHMARK 5.0 RESULTS

Please do NOT compare it with older versions of the benchmark!
Please copy/paste everything below the line to to report your data
to http://forums.techarp.com/reviews-articles/26957-x264-hd-benchmark-5-0-a.html

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Results for x264.exe r2200
x264 Benchmark: 64-bit
==========================

Pass 1
------
encoded 11812 frames, 90.04 fps, 7754.21 kb/s
encoded 11812 frames, 90.13 fps, 7754.07 kb/s
encoded 11812 frames, 90.31 fps, 7754.08 kb/s
encoded 11812 frames, 91.36 fps, 7754.02 kb/s

System Details
--------------
Name Intel Core i7 3770K
Codename Ivy Bridge
Specification Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
Specification PC3-18300
Specification PC3-16000
Specification PC3-18300
Specification PC3-16000
Core Stepping E1
Technology 22 nm
manufacturer Gigabyte Technology Co., Ltd.
vendor Gigabyte Technology Co., Ltd.
manufacturer Gigabyte Technology Co., Ltd.
Stock frequency 3500 MHz
Core Speed 5001.8 MHz

Northbridge Intel Ivy Bridge rev. 09
Southbridge Intel Z77 rev. 04

I was compressing Larger files..

Please quit trying to Hype a complete failure of another Amd Cpu.. Your still 3 Generations behind Intel and it is quite Sad..

 

[Phynaz]

Still with your twisted numbers ?...
Yet , i did give you a clue above...

From TR whose link is (forcibly) above and apparently
out of reach for your stretching eyes..



It s the 8350 they are talking about , so from 262 to 244 there is 20%..??

I understand better the kind of guy you are , shamelessly
and obviously trolling all the way.

In principle such guys do not deserve any answer but for the sake
of the forum i have sometime some mercy even with ruthless trolls....

TechReports overclocking results:

4.8GHz, 1.475V - reboot
4.7GHz, 1.4875V - lock
4.6GHz, 1.525V - errors on multiple threads
4.6GHz, 1.5375V - errors with temps ~55C
4.6GHZ, 1.5375V, Turbo fan - stable with temps ~53.5C, eventually locked
4.6GHZ, 1.5375V, manual fan, 100% duty cycle at 50C - lock
4.6GHZ, 1.55V, manual fan, 100% duty cycle at 50C - crashes, temps ~54.6C
4.4GHz, 1.55V - ok
4.5GHz, 1.55V - ok, ~57C, 305W
4.5GHz, 1.475V - errors
4.5GHz, 1.525V - errors
4.5GHz, 1.5375V - OK, ~56C

"There's a 48W gap between the TDP ratings of the Core i5 parts and the FX-8350, but in our tests, the actual difference at the wall socket between two similarly configured systems under load was over 100W. That gap is large enough to force the potential buyer to think deeply about the class of power supply, case, and CPU cooler he needs for his build. One could definitely get away with less expensive components for a Core i5 system."

Okay, so let's use your figure anyway. So the undervolted 8350 uses 18 watts more than the 3770K and is 25% slower in highly threaded applications. Is that more to your liking?
(We don't even have to talk about lightly threaded applications)

 

[Abwx]

Okay, so let's use your figure anyway. So the undervolted 8350 uses 18 watts more than the 3770K and is 25% slower in highly threaded applications. Is that more to your liking?

TR results are for a 4.9g 3770K and 4.5G for the 8350 in multithreaded
tasks and likely 4.8G in lightly threaded apps.

Also , if 8350 is 25% slower it means the 3770K has 33% better perfs
than the 8350 , i suppose that you know that ,so let s find out if this
figure is accurate for highly threaded worloads :

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/13

I did put the first pass of video encoding although it s a lightly threaded
case , since it matters for the total encoding time....

See by yourself , keep in mind the 400Mhz higher frequency of the 3770K
with highly threaded loads.