Sandy Bridge

[Sticky]

I have been really pushing the burn tests, and mine is "only" completely rock stable at 4.2 GHz. It ran for a few minutes at 5 Ghz :) (49 multiplier and some baseclock oc, it wouldnt boot at all at 50 multiplier) and it seemed very stable at 4.8 at first but wasnt, same thing with 4.7, 4.6 etc all the way down to 4.2, they seemed stable, but really wasnt completely burn proof until 4.2. I use IntelBurn Test and prime 95 and really pushed it. I dont know how hard others have pushed their cpus, but the high numbers, Im not sure anymore that they are completely stable at those frequencies. Or it could be just mine. But 4.2 Ghz is good for a burn proof 24/7 IMO :)

[RSharpe95]

Well unfortunately, that's really not that much better off than a i7 9XX..

[Sticky]

Well unfortunately, that's really not that much better off than a i7 9XX..

 

Its good for others, in my case I upgraded from socket 775.

[Druid_]

I have been really pushing the burn tests, and mine is "only" completely rock stable at 4.2 GHz. .....:)

 

Did you increase the cpu core voltage to 1.35V and your memory to 1.6V ? Also what cooler have you got sitting on top of it? Memory can sometimes be the bottleneck when you overclock so run memtest first to make sure you memory is happy at the set speed.

 

Well unfortunately, that's really not that much better off than a i7 9XX..

hardly ! very few get their i7 9XXs to over 4.0ghz and certainly not without some serious cooling involved. Also ghz is no indicator of processor power any more when comparing different processor architectures.

[Sticky]

Did you increase the cpu core voltage to 1.35V and your memory to 1.6V ? Also what cooler have you got sitting on top of it? Memory can sometimes be the bottleneck when you overclock so run memtest first to make sure you memory is happy at the set speed.

 

 

hardly ! very few get their i7 9XXs to over 4.0ghz and certainly not without some serious cooling involved. Also ghz is no indicator of processor power any more when comparing different processor architectures.

 

Yes on the cpu core voltage to 1.35, I tried that, it was stable at 4.8 for maybe half an hour of burntesting, then BSOD, and the same at other multiplier settings, every time I thought was stable, tried all kinds of voltage settings, at 4.4 I was sure it was stable, but it finally BSOD after a couple of hours in a game, and then I went back for more burn testing at 4.4 and sure enough, BSOD again.

 

I set my memories (corsair xms3 1333) at 1.55, I didnt try at 1.6, I will try that too and see if that helps. I already did a round of memtest, but I will do another one at 1.6.

 

I went kindof half cheap with the cooler, artic cooling 7 rev2, I had very good results with that cooler before so I thought since its 32nm it should do the job, but it gets to 83 degrees on the highest voltage settings (at 4.6-4.8Ghz) when I burn test all 4 cores at the same time. If I go down to 3 cores for the burn test temperature stays at 78-79. With 2 cores at max load temp is low, 65ish. Think the cooler is the problem?

 

If it isnt the cooler, then maybe something is wrong with my methodology, or maybe I have a chip thats not as good as the reviewers had, or I have too high demands on the burn tests. Or maybe 4.2 just feels low because I thought it was stable at 4.8 and I really should stay at 4.2 :)

Edited January 21, 2011 by Sticky

[Sticky]

My mainboard has ALOT of cpu power saving options, I had to turn off 5 or 6 different setting to make sure that the cores stayed at the set frequencies, and not automatically go down to 3.8-4.0 at load. At default settings when you set the cpu frequncy to 4.8 for example it only stays there for a few seconds, then goes down to 3.8-4.0 to save power, while still being under heavy load. And I made sure to turn all that crap off, so maybe those who claim stable CPUs at 4.7-4.9 havent really looked into their power options, they only run at the set ratio for a few seconds.

 

For me that kind of behaviour on the CPU was something new, it took a while before I could make the CPU stay at the set frequency, but I suppose if you had the previous generation i5 or i7 maybe those funtions were already there, and you are already familiar with this and it will be easier to set things up.

 

edit: I still have it set to run at 1.6GHz at idle, but at load it really stays at the set ratio, 4.2 in this case, and doesnt go down during load. I like that function, that it runs lower at idle, and it took a while to find the settings to have both set up the way it is now, only high and low, nothing inbetween. With socket 775 it did that by itself, but now there is so many options..

Edited January 21, 2011 by Sticky

[Druid_]

83 degrees is hot but no hotter than I have seen some overclockers get to at 5Ghz. A good cooler is a definate must however for anyone who is going to overclock their CPU.

 

Memory is ok up to 1.65v

Not familiar with your BIOS or settings for Sandy bridge but as you say maybe your CPU die can only reach 4.2.

[Sticky]

83 degrees is hot but no hotter than I have seen some overclockers get to at 5Ghz. A good cooler is a definate must however for anyone who is going to overclock their CPU.

 

The cooler isnt that bad, its performance is not far from the more expensive ones, and at a very low price. I might try pushing higher again if I buy a new cooler, Im going to think about that, but probably not, I will probably stay at 4.2. Considering the low price I went for with the components I think 4.2 is good for this upgrade.

 

Not familiar with your BIOS or settings for Sandy bridge but as you say maybe your CPU die can only reach 4.2.

 

It can reach 5, if you want to use those words ;) But only 100% rock stable at 4.2

[Kuky]

Keep in mind also those very high clocks people claim are mostly for showing off, they run it for benchmarks etc, and not for 24/7 setup :) 4.2GHz is very nice overclock

[theGozr]

All depend of your batch number .. .. some batch have some low overclocking capability

[Prophet]

The single thread performance of the i3 2100 is better than the i7 980X.

 

 

And from what I have been reading, for some reason the SB actually is overclocking better with most of the power saving features turned on. Completely backwards from everything previously.

[GGTharos]

Looks like it's 'the next thing' to buy :)

[Cali]

Looks like it's 'the next thing' to buy :)

 

I agree an that, hope I can build a new computer within the next 2-3 months.

[JaseGill]

I have been saying the same for a while.

Its amazingly fast, smooth in both A10C and BS and I cant see any snags with it at all.

Its a major step up for Intel and currently unbeatable in my humble opinion.

That is at stock speeds too. Over-clocked an i5 2500K like mine supposed to wipe the floor with an i7 980X. Unbelievable until its proven and I can see why some doubted my reports early on but I'm sure its all over the internet by now how damn fast these Sandy Bridge are.

[theGozr]

Yeap Got my 2600K in January 9 ;)

 

Just did two tests one on my server a Q9550 cruising @ 3.6GHZ cinebench 1 CPU Cinebench-10 4484

My Other system is an Q9650 @ 4050 GHZ 1 CPU Cinebench-10 4996

Edited January 24, 2011 by theGozr

[Sticky]

And from what I have been reading, for some reason the SB actually is overclocking better with most of the power saving features turned on. Completely backwards from everything previously.

 

Not sure I believe that, it could be just a misunderstanding on their part, with some of those power settings turned on you might think that you are stable at a sustained 4.9 during load, but you spend almost no time at the set frequency during load. So if you are going to use that, you might aswell lower the freq to under 4 yourself instead, because its the same thing.

[sobek]

Well, isn't it a little lopsided to judge the performance from one test?

[Sticky]

Well, isn't it a little lopsided to judge the performance from one test?

 

I dont understand, who is judging what from one test?

[JaseGill]

Sobek,

Maybe you should look at Custom PC, PC Pro, PC Gamer, the Bit-Tech website, Toms Hardware, Overclockers Forum and any number of other magazines and websites ad infinitum before commenting on how lopsided this test is. You'll find that its about as one sided as a sack of Dodecahedron's ;)

The accepted consensus is that the Sandy Bridge architecture is head and shoulders above anything else available at this time, and at a very good price point I'd add. The only ones upset at this time are AMD.

I'm no Intel Fan-boy either in case you're wondering. I think its about time AMD got back in the game. No progress is ever made when a single company has a monopoly, we need competition to push AMD and Intel to keep outdoing each other, otherwise we stagnate.

Edited January 25, 2011 by JaseGill

[sobek]

Sobek,

Maybe you should look at Custom PC, PC Pro, PC Gamer, the Bit-Tech website, Toms Hardware, Overclockers Forum and any number of other magazines and websites ad infinitum before commenting on how lopsided this test is. You'll find that its about as one sided as a sack of Dodecahedron's ;)

 

Yes, i know it performs well compared to the 1366 i7, and i have read tests, but i talked about conclusions that people jumped at after the graph that was posted above. One single test says about next to nothing.

[JaseGill]

Sobek,

That graph and many like it are on a hundred websites, in a hundred magazines and splashed all over the internet and unless everyone (me included) is in Intel's pay (I wish I was I could do with the extra cash) then it has to be taken as a consensus I believe.

I agree, one graph on one website (even this hallowed ground) is no proof, but its not one graph on one website once one performs a minimum of investigation via Google.

The only conclusion I see anyone jumping to is that its pretty conclusive, hence a good conclusion, wouldn't you agree?

In the end the only thing that matters is real-life experience and I can only offer my own, which is that in a non-overclocked i5 2500K (rest of my spec in my sig) I'm getting in the 70's FPS general flying, 40 over the battlefield, 30 + when there's loads going on and not even the infamous noticeable lag when the cannon rounds hit.

For me that's all the proof I need that the above graph is sufficient to jump to some pretty conclusive conclusions.

sorry if it sounds like a rant or a dig at you, as its not supposed to. Had you said "I'd like to see more proof other than one graph before I believe all I read" then I would have agreed. But as you have already said you've seen all those yourself.

 

+1 rep to you as an aside, as I think this is what these forums are for, good debate and without it, and informed members like yourself, it would be a much worse place for it. I really do follow this forum with great interest and think it is the best of its type.

 

If I could I'd give ED +100 rep just for being so experimental in releasing A10C to its community and allowing us input. I think this must be the way to go in future for all gaming communities with such informed, intelligent members.

 

Anyway that's all OT so back to the debate ;)

[Kuky]

may I add something here, I just downloaded and tried this Cinebench 10 software and I got 5849 for single core test and 21152 for 4 cores test, my CPU is i7 950 overclocked to 4GHz.

 

What I am also thinking is that when they state i7 2600 is at 3.4GHz in this test is not correct as these CPU's auto-overclock and also when only 1 core is in use and others are idle the CPU has ability to shut down cores that are not in use so that core that is in use can overclock more. The default stated turbo freq for i7 2600(K) is 3.8GHz and for i7 2500(K) its 3.7GHz so they would have to be running at least at those frquencies when this test was done, so I am confused as to why they are not saying that and if I am correct this test is misleading.

 

also price performance wise i7 950 that can easily overclock to 4GHz is still cheaper then i7 2600K that can also easily overclock to 4GHz but judging from this cinebeck 10 test for 1 core my CPU is the same as the i7 2500K (again which auto overclocks to 3.7GHz) but granted the i7 2500K is about the same price also

Edited January 28, 2011 by Kuky

[JaseGill]

You may have a point there Kuky that I hadn't considered.

I think you'd also have to see the background spec that's involved (motherboard, memory, BIOS Settings) before you could say that single (amongst many) benchmark has any significance.

 

I know that the BIOS on my P8P67 PRO allows you to turn off the automatic over-clocking that's built in, whether this is the case across the board (no pun intended) I couldn't say.

 

As in all things PC, its about much more than the CPU, no matter how good the CPU.

[EtherealN]

Well, tests at stock frequencies are usually more about having the processor "stock" than measuring a specific frequency. So when comparing an i7 950 at stock and an i7 2600K at stock, both would have some TurboBoost happening and we're comparing processors that have not been tinkered with at all, so to speak.

 

That said, there is indeed an issue there as far as measuring overclocked systems goes, but it seems fairly easy to control for in my opinion.

[GregP]

UPDATE: Added overclocking results.

 

Well guys, I just built a new system last night and ran some quick tests, and am excited to share the results. I've done what I think is as close to an apples-to-apples comparison as is possible, and I'm amazed at the performance bump I'm seeing. I ran these tests under Windows 7 Professional 64-bit and DCS Warthog Beta 4 64-bit. Using Method 1 from the instructions (here), I "sysprep'ed" my existing Windows install and simply moved it to the new build with no problems at all, a Windows first for me, saving countless tedious hours of program reinstallation. This should minimize the differences between the OS configuration on my old and new systems, making a direct performance comparison all the more meaningful.

 

Old system:

Core 2 Quad Extreme QX6800, stock 2.97 GHz overclocked to 3.47 GHz

eVGA nForce4 680i SLI

eVGA GeForce GTX 470 SC

8 GB DDR2 PC2-8500 (1066 MHz)

 

New system:

Core i5-2500K, stock 3.3 GHz overclocked to 4.6 GHz

Asus P8P67 Pro

eVGA GeForce GTX 470 SC

8 GB DDR3-1600 PC3-12800 (currently running at 1333 MHz)

 

My standard framerate tests use FRAPS in a simple 'sitting on the runway' mission I created and then a 1-minute framerate record in the Litening pod training mission. I also run two different video configurations for testing: a single monitor at 1920x1080 and 3 monitors with SoftTH for a total resolution of 4630x1080. A caveat with these two video configs, though, is that FRAPS will only display the framerate for the left or right monitor when running 3 screens; given that these tend to be rendering at higher rates than the central monitor (I believe this is simply the way that SoftTH works) the results below appear to show the unexpected result that I get higher framerates when running 4630x1080 than I do when running 1920x1080. This is not actually the case, it's just that I can't measure the central monitor's FPS with FRAPS nor with DCS's built-in framerate counter as it doesn't seem to be working. As such, it makes more sense to look at framerate percentage changes rather than raw numbers to get an idea of the performance increase I'll see.

 

So far I've been able to push it up to 4.6 GHz without much work, and I have a feeling I could probably safely get it to 4.8, but there may not be much additional payoff.

 

In-game graphics settings:

Textures: High

Scenes: High

Civ Traffic: Low

Water: Medium

Visibility Range: High

Heat Blur: On

Shadows: Medium

Cockpit Res: 1024

MSAA: 2x

HDR: On

TSSAA: Off

Clutter/Bushes: 250m

 

And here are my results:

 

Sitting on runway:

Old

4630x1080: 39

1920x1080: 24

 

New, stock 3.3 GHz

4630x1080: 48

1920x1080: 43

 

New, o/c to 4.6 GHz

4630x1080: 52

1920x1080: 49

 

Litening pod training mission:

Old

4630x1080: 39

1920x1080: 27

 

New, stock 3.3 GHz

4630x1080: 52

1920x1080: 46

 

New, o/c to 4.6 GHz

4630x1080: 57

1920x1080: 60

 

So basically I'm seeing an incredible 25-74% framerate increase (25-33% increase in the configuration that I actually run) going from my old system to the stock 3.3 GHz of the i5-2500K, but when I crank it up to 4.6 GHz I'm seeing an insane 33-122% increase (33-46% in my actual config). Of course, the "FRAPS only measuring outside screen framerate" issue makes it hard to tell if the the percentage increases at 4630x1080 are being measured on the same scale as those at 1920x1080, but either way, I'm quite happy.

 

Bottom line: woo-hoo!

Edited January 28, 2011 by GregP