1366 x58 Xeon Enthusiast overclocks club

[King Icewind]

My 5650 is a bitch. I can't get it to work with 200 blck yet. I have gotten two different issues. 1. It will post sometimes and when it does it hangs on building DMI and the BIOS won't load. You have to reset the CMOS manually. 2. It will give me a 0x0007E error which means windows is corrupt. Resetting the cmos let's me boot into windows again.

The most stable I have gotten is 170 blck. I'll try 180 next. Maybe I should try 180×220 or whatever

 

[Cecil]

Got the Big Bang X-Power and testing it out. So far I can't seem to make it allow UCLK multi changes. It stays greyed out. It seems to default to 1.75x ram speed.

I think this is what is stopping it from being stable over 190ish BCLK. Using a dmm measuring voltages, Vcore I've gone as high as 1.4v, QPI 1.348v, VDimm 1.603v and IOH 1.208v.

Thinking it wants more QPI voltage to get the UCLK higher.

In reviews for the board the UCLK is adjustable, so not sure if it just doesn't give the option with this CPU, or if there is something I'm missing.

Also, CPU-Z and CPU Tweaker are reporting the uncore at 3800Mhz with an x20 multi, regardless what is being set in BIOS.

 

[Zoson]

Got the Big Bang X-Power and testing it out. So far I can't seem to make it allow UCLK multi changes. It stays greyed out. It seems to default to 1.75x ram speed.

I think this is what is stopping it from being stable over 190ish BCLK. Using a dmm measuring voltages, Vcore I've gone as high as 1.4v, QPI 1.348v, VDimm 1.603v and IOH 1.208v.

Thinking it wants more QPI voltage to get the UCLK higher.

In reviews for the board the UCLK is adjustable, so not sure if it just doesn't give the option with this CPU, or if there is something I'm missing.

Also, CPU-Z and CPU Tweaker are reporting the uncore at 3800Mhz with an x20 multi, regardless what is being set in BIOS.

Have you updated the BIOS? That's usually why things like that don't work right. I'd trust CPUz and CPUt on the UCLK. It would explain your problems with getting a higher BCLK.

 

[Cecil]

Yes, its on the newest BIOS.

Lowered CPU and RAM multi to get 200BCLK, and took 1.38QPI to get to windows. So, it seems even though the board reports a low uncore speed in BIOS, it is forcing 20x at all times.

Going to settle around 185-190BCLK to top the ram out. 24GB rated at 1866 9-10-9-28 in 8GB kits, was able to get 1860Mhz 10-11-10-28 and 1900Mhz 11-11-11-28. Need to fiddle with them some more.

 

[primetime]

My 5650 is a bitch. I can't get it to work with 200 blck yet. I have gotten two different issues. 1. It will post sometimes and when it does it hangs on building DMI and the BIOS won't load. You have to reset the CMOS manually. 2. It will give me a 0x0007E error which means windows is corrupt. Resetting the cmos let's me boot into windows again.

The most stable I have gotten is 170 blck. I'll try 180 next. Maybe I should try 180×220 or whatever

have you boosted all the voltages to the max safe levels that zoson and quite a few of us have been doing...if you have then im thinking those evga boards are a bit trickier than the asus boards.....i think mine does 200blck easier than 133 lol...and im not kidding

 

[combat fighter]

Your wrong about this point, the test isn't running (in the screen shot is the message at the end of the test) and the memory is only allocated when the test is running, so in his screen shot he only actually has 4242 free, IBT doesn't deduct the setting from your available memory so you can't add on the imaginary 1024 that the test will allocate when it's running.

I knew I was right with what I was trying to explain. Yet I get called a troll! WTF?

I must be confused about exactly what you guys are debating, but i would think just setting it to max (sometimes you have to start over since it trying to use to much the first time and gives that error message about not having enough ram) but its always worked good for me in the past just setting it to max (of course us with 12 gigs set it to 9216mb since we have been told that was the max it could use properly) according to zoson, but it makes sense the way he explains that part

but it wouldn't be the first time i was confused what people were talking about

Glad I'm not the only one that was confused.

 

[GonzoP]

Inspected the CPU and socket they appear mint to me.

Re-seated the heatsink again and still 8GB Ram in BIOS and Windows yet CPUZ 12.
I was able to get CPUTweaker to function so channel A but Tweaker sees only 8GB ram as well. Just can't figure out why BIOS and Windows cannot see the ram.

 

[GonzoP]

 

[Zoson]

I knew I was right with what I was trying to explain. Yet I get called a troll! WTF?



Glad I'm not the only one that was confused.

You're actually still totally wrong, restart diagnostic mode and run 5120.

 

[Zoson]

First, I see at least two pads with damage, possibly four.


Things to try:
Setting ram back to 2T
Raise TRFC (all my 12GB kits have an XMP of 124 or higher), 88 seems very low.
Raise tRRD to 6.

 

[primetime]

You're actually still totally wrong, restart diagnostic mode and run 5120.

Zoson when you say Diagnostic mode is that the same thing as safe mode? Is that so the app has more free memory as to make a more accurate check of the ram and cpu when running the stress test? Is this something we all should be doing or just certain systems?

Now that i have asked the question i did notice they are 2 different options in the system config and was just kinda curious how its different from safe mode. Btw thnak you for the tip even i didn't know that was the correct was of doing it as i had been doing it in safe mode and even then only when i had issues with windows, but ill be doing it the correct way from here on out

 

[primetime]

First, I see at least two pads with damage, possibly four.


Things to try:
Setting ram back to 2T
Raise TRFC (all my 12GB kits have an XMP of 124 or higher), 88 seems very low.
Raise tRRD to 6.

I think someone else mentioned this before, but its a shot type dark type of idea of resetting the bios to defaults (possibly even clear the cmos) and then reload the setting from a saved config or just manually setting them back.....not saying it will fix the issue but in the end i think its worth trying if nothing else works

 

[Zoson]

Zoson when you say Diagnostic mode is that the same thing as safe mode? Is that so the app has more free memory as to make a more accurate check of the ram and cpu when running the stress test? Is this something we all should be doing or just certain systems?

Now that i have asked the question i did notice they are 2 different options in the system config and was just kinda curious how its different from safe mode. Btw thnak you for the tip even i didn't know that was the correct was of doing it as i had been doing it in safe mode and even then only when i had issues with windows, but ill be doing it the correct way from here on out

It's not safe mode, but it's also not full boot.
Start -> Run -> MSCONFIG -> click the diagnostic mode bullet, hit ok, and restart.
Likewise, to get back to a normal boot.
Start -> Run -> MSCONFIG -> click the normal mode bullet, hit ok, and restart.

It doesn't load any services, and doesn't load anything in your startup/run/runonce registry keys. You still get network access, and driver loading so you can do full benching/testing without your other installed stuff compromising your testing. That way, you don't end up with ~60GFlops and an invalid test due to insufficient loading when you should be approaching 80GFlops.

I think someone else mentioned this before, but its a shot type dark type of idea of resetting the bios to defaults (possibly even clear the cmos) and then reload the setting from a saved config or just manually setting them back.....not saying it will fix the issue but in the end i think its worth trying if nothing else works

Yeah, I floated a full failsafe reset. But just from his CPUt, those things seemed off to me.

 

[primetime]

better late than never but i updated the first post with zoson's max voltage settings.......hopefully it saves somebody from destroying their cpu

 

[King Icewind]

My EVGA board uses +25mv increments for the CPU VTT voltage. I read that the starting voltage for non classified boards is 1.1 on the CPU VTT which means the max I could go is +250mv (1.35v). Is that all correct?

 

[Deimos]

Tweaking this system is getting a little bit WTF? why?

I was having trouble getting 4.4Ghz stable so I want back to testing the memory just to be sure it wasn't that.

Figured out that my ram just won't run tighter timings apart from a 1T CPC (default is 2) but this is what is most interesting.

To test the ram I dropped the QPI to 1.25v, mem volts to 1.55v, left the PLL at 1.35v, CPU volts at auto. 200BCLK with a 15 multi (for 3Ghz) and ran memtest86 over night, no errors.

Bumped the vcore to 1.35v (max is 1.402v in HWMonitor), changed the multi to 22, left everything else as is and so far so good!?

Were the extra memory and QPI volts making my overclock unstable?

I'm still fiddling, I'm going to try lowering the CPU volts to see if I can't get the temps down a bit more but with these settings it's already a couple of degrees cooler than the last time I tested it.

Zoson there are some funky voltage settings in the R3E bios like the option to set the PWM voltage (between something like 3 to 10 volts) PWM frequency, Digi+ PWR Mode, do you mess with any of these?

 

[Deimos]

How much memory should I be testing in IBT and why? I hear 9216MB but was there some reason a specific amount needs to be used? something about HT or something?

 

[Rock&Roll]

errrr

 

[Zoson]

My EVGA board uses +25mv increments for the CPU VTT voltage. I read that the starting voltage for non classified boards is 1.1 on the CPU VTT which means the max I could go is +250mv (1.35v). Is that all correct?

That would be right, but check with HWMonitor to be sure.

Tweaking this system is getting a little bit WTF? why?

Were the extra memory and QPI volts making my overclock unstable?

Zoson there are some funky voltage settings in the R3E bios like the option to set the PWM voltage (between something like 3 to 10 volts) PWM frequency, Digi+ PWR Mode, do you mess with any of these?

Yes, having your QPI too high can cause instability.
I only change PWM freq, amplitude, and skew when over 5.2ghz. And that requires an extra 3x120mm rad with 6x 2150rpm GT's in push/pull on a cool day.

How much memory should I be testing in IBT and why? I hear 9216MB but was there some reason a specific amount needs to be used? something about HT or something?

So, the bug is actually in hyperthreading. When you have a very large problem set size the scheduler starts core thrashing, where there are too many threads for the number of instructions. This results in significantly less load on the cpu. The actual max is around 10GB, but 9216 was what we decided on on XS. Really, it's about standardizing the problem size so you can actually compare results against each other.

The consensus was 5120MB and 9216MB for 6GB and >=12GB systems.

If you disable HT, you can use any size you want.

 

[King Icewind]

I thought the max multiplier was 20 when under a full load on the x5650, and 21 in turbo. Is it just a program reading error that I'm running at 22 then? Three different programs report it when under 100% load. 22 x180 for 4ghz.

 

[Cecil]

I thought the max multiplier was 20 when under a full load on the x5650, and 21 in turbo. Is it just a program reading error that I'm running at 22 then? Three different programs report it when under 100% load. 22 x180 for 4ghz.

Turbo is 22x. When in the BIOS, you can set it (for a lot of boards) to run at constant 22x. It goes to 23x if 1-2 cores are being used.

 

[King Icewind]

Well I only have set to a x20 multiplier.. but hey, if this isn't an program reading error I'm happy with it. Although I'm sitting at 1.365v vcore. I think my board just isn't very good. I couldn't get my L5639 past 3.2ghz with a higher vcore either. I wish I could have your guy's vcores, ha.

 

[primetime]

my max goes up to 23-24 but i really can only use 21 100% stable anyway...i think the 5660 are the real sweet spots imo, but for the money if i had to do it over would just get a 5650....i bet there's hardly any difference above 4ghz in games/apps anyway......course for some reason mine likes 200 blck the best so i guess that would make a difference also

 

[GonzoP]

my max goes up to 23-24 but i really can only use 21 100% stable anyway...i think the 5660 are the real sweet spots imo, but for the money if i had to do it over would just get a 5650....i bet there's hardly any difference above 4ghz in games/apps anyway......course for some reason mine likes 200 blck the best so i guess that would make a difference also

I'll politely disagree here. The sweet spot clearly is the 5650. As we have seen now from dozens of guys OCing their Xeon's. 5650, 5660 and 5670 are all settling within a 100Mhz window when the same voltages are applied. The higher multiplier is truly the main advantage of the higher end model.
If we use the 200BCLK as a basepoint the 5650 can get you to 4.4Ghz. Sure the 60 and 70 could then get you 4.6 and 4.8 respectively. But who here is actually running 4.6+? Most of the 5660 and 5670 guys [like yourself] are settling in at, or even below, 4.4. The few guys who do seem to use 4.6+ speeds 24/7usually have some kinda water cooling system.
Myself I run 4.5 24/7 and based on my cooling and from what I've seen others achieve with their 5660 or 5670's I doubt with one of those chips I'd be running much higher than I am now. I mean yes it does seem the median average is fractionally higher with a higher model number and again median average voltage needed at higher end speeds also seems fractionally better.
I'm absolutely not looking to pick a fight here at all prime. Just saying if a buyer is intending to settle in somewhere in the 4.2-4.4 neighborhood there is no reason to look beyond the 5650. If you have a custom loop or maybe something like a H100i or Kraken AIO and are looking for 4.6+ would you be better off with a 5660, 5670? I think the evidence point to the most accurate answer being...maybe, even probably, but not definitely. I mean I can run 4.8 although then I'm above 1.45v and that's a no-no for me and my cooling set up and my desire to have this cpu last me till Skylake. But sure there's guys over at OCN and XS running 5Ghz Xeon's with 1.5v but then they usually have custom water loops.
Bottom line for the masses 5650 is all someone needs

EDIT: OK technically with SpeedStep I don't actually run 4.5 24/7 but you get my point

 

[YBS1]

If we use the 200BCLK as a basepoint the 5650 can get you to 4.4Ghz. Sure the 60 and 70 could then get you 4.6 and 4.8 respectively. But who here is actually running 4.6+? Most of the 5660 and 5670 guys [like yourself] are settling in at, or even below, 4.4. The few guys who do seem to use 4.6+ speeds 24/7usually have some kinda water cooling system.

I am, 195x24(25) is my everyday setting. Now mine is water cooled, but it's an H80i, only marginally more expensive than most of the upper end air coolers. I do agree with much of what you are saying but I'd still take a 5660 or 5670 over the 5650 simply to eliminate any chance that my motherboard would be the limiting factor in a possible overclock. The 5660 pretty much takes care of this, but with the current price difference between it and the 5670 I'd just go 5670. The 5650 clearly hold the price/performance trophy though (assuming prices haven't changed much in the past week).

 

[GonzoP]

I am, 195x24(25) is my everyday setting. Now mine is water cooled, but it's an H80i, only marginally more expensive than most of the upper end air coolers. I do agree with much of what you are saying but I'd still take a 5660 or 5670 over the 5650 simply to eliminate any chance that my motherboard would be the limiting factor in a possible overclock. The 5660 pretty much takes care of this, but with the current price difference between it and the 5670 I'd just go 5670. The 5650 clearly hold the price/performance trophy though (assuming prices haven't changed much in the past week).

Yeah we're not in disagreement here. In fact you build my case for me. I mean its all good here; I'm not genuinely bickering. You have an above average 24/7 speed with above average cooling. You also are chosing to run a sub 200BCLK. As I previously stated the main draw of the higher priced models was their higher multi's. Its very established that many/most x58 mobos struggle past 215/220BCLK. Now I could match your total clockspeed with a BCLK of 213. That's very near my personal ceiling but still do-able. My problem of going above 4.6 is the vcore I needed. If you are doing 195x24 and you're at less than 1.45v than I tip my cap. But again as I stated in my previous post I did acknowledge that there is a slight marginial or fractional median average of say 100Mhz or a few voltage ticks. I mean I can show you my cpuz validation at 4.7Ghz on 1.44v vcore but it turns out that was not a stress test proof voltage and I would need an extra .25v to stablize that. Same with 4.8Ghz. I have my meaningless proof shot on 1.48v vcore but again I would need over 1.5v for stress test passing stability and its pure nonsense to pump 1.5v thru my Xeon with my budget air cooler.
Now again having seen my own 5650 do 4.8 I was originally tempted by the 5660 or 5670 thinking I could pull off 5Ghz with the extra multiplier. At 4.8 I was at 219 BLCK and that is about my ceiling. I didn't even try for 4.9...I think. Maybe I did and but 223 not going to happen if I did. What ended my temptation was remembering why I got the 5650. It was $70 and I was done with upgrading costs. I mean for me to truly take a run at 5Ghz I genuine would need a better cooler than I have now. But what am I going to do spend $70 on a cpu and be done or $160 on a cpu and then $75-125 on its cooler; all for 5-10% more speed than I can have for $70? To me that was silly. The draw of this Xeon for myself and most other x58'ers is that is one step and done. No new mobo, no new ram, no new PSU, no new cpu cooler.
At the end of the day I've reminded myself to be content with a 24/7 4.5Ghz for $70. Is this better or worse than a 24/7 4.7Ghz for $150? Maybe, maybe not. Or a 24/7 5Ghz for $250 [assuming a 5670 and a H110i could even achieve such a feat]?
I'm not against the higher priced models. Some guys who might have mobo problems or difficulty with 200+BCLK might steer towards the 5660 just for the extra multi. I mean YBS1 I see you're choosing 195 and I wonder what the reason for that is? Again we're all in the same club, one big family Xeon-X58 family so I'm not trying to defend some kinda turf here. And yes YBS1 you are in the minority at least here at [H] of guys over 4.6 24/7

Since YBS1 mentioned it I did just check the ebay prices and see the still at the same pricepoint I got mine for a month ago. The 5660 and 5670 seem to be only about $20-30 apart which yes does tend to make me think the 5670 would be the better choice there if someone was going to spend more than $70. I mean if you're committed to the $130 for a 5660 why not spend $20 more on the 5670? The fact that the 5650's haven't moved down for a month tells me this is their bottom line. 5670 is $10-15 cheaper than a month ago from the two big sellers and now you even have individuals selling their's for less. Even less still are these from China ones but I'd pass on those myself some of the screen shots look nasty

 

[Zoson]

FYI higher BCLK generally requries higher voltages across the board.

My every day driver is 4665MHz core with 3733MHz uncore and 1866MHz memory. I only need 1.46v core and 1.3625v uncore in bios to drive this overclock with 24GB ram.

If you look at the 200BCLK 4.6GHz sample I created for you guys the other day, you'll notice that my voltages are noticably higher than that, and I'm only using my 6GB kit(although I was running 4000MHz uncore for this, which requires higher voltages too).

I never really took the time to understand why exactly this is, but generally you want the higher multipliers to run lower BCLK so your entire voltage requirements are less all around.

 

[primetime]

FYI higher BCLK generally requires higher voltages across the board.

As a hole i generally agree with this statement, with some exceptions, and by that im just referring to what i thought i read GonzoP mentioning that in regards to just the vcore voltage, that part was mainly in relation to the overclocked clock speed in use, likewise well for myself it appears to be that way as well. Now all the rest of the voltages at least in my mind make more sense to need to be higher with higher bclk but i admit im far from an expert

Speaking of voltages when i look at your voltage sticky on the first page i notice 1.65 max for the DDR voltage and since my board only does in even numbers than 1.66 would still be inside safe margin correct?

 

[YBS1]

My problem of going above 4.6 is the vcore I needed. If you are doing 195x24 and you're at less than 1.45v than I tip my cap.

I can check later to confirm, but I think my 195x24 setting is running 1.39v to 1.41v. Now I do have LLC enabled for that setting so under a load I usually see ~1.45v. To be honest, unless someone is wanting to just make a run at 5GHz for benchmarking e-peen purposes there is no point to going over the 5650 assuming you have a board capable of over 200BCLK. It may increase your chances of getting an excellent chip, or maybe not, I don't think there is really enough data to say as of yet as there aren't as many people with 5670/5675s. This particular chip cruised to 4.8GHz, but I had to bear down on it a bit for 4.9, and have yet to get a benchable 5GHz. So it has a pretty hard ceiling, and I've went up to 1.62v trying. My Bloodrage itself is capable of at least 220 that I know of, that's as far as either the board could go or my W3520 could carry it.

I mean YBS1 I see you're choosing 195 and I wonder what the reason for that is?

That's as fast as I can go at a voltage I feel comfortable with for extended durations.

 

[Rock&Roll]

My 5650 is a bitch. I can't get it to work with 200 blck yet. I have gotten two different issues. 1. It will post sometimes and when it does it hangs on building DMI and the BIOS won't load. You have to reset the CMOS manually. 2. It will give me a 0x0007E error which means windows is corrupt. Resetting the cmos let's me boot into windows again.

The most stable I have gotten is 170 blck. I'll try 180 next. Maybe I should try 180×220 or whatever

Post your BIOS template? Curious what kind of settings you're running. Voltages. Enabled features. Etc...

There should be a reasonable amount of overlap between my older E760 and your newer E767

 

[adobian]

This is an interesting thread.

I am torn between buying a $70 dollar x5650 and overclock it to 4.4ghz with a $250 X58 motherboard, as vs buying a 4790K and a lga1150 motherboard.

Could you advise?

 

[King Icewind]

I'd go with the newer tech. I like having my 6 core 12 thread processor just because I am a geek, but a newer 22nm lower TDP processor with stock 4-4.4ghz freq. that could be overclocked extra is a better decision. Plus you will want to find triple channel kits of ram which is harder to find. Clock wise they will be the same and I don't think you'll notice a difference really with any games unless you were doing encoding type of stuff.

And I will post pictures of my bios for you rock&roll tomorrow.

 

[Deimos]

I'd go with the newer tech. I like having my 6 core 12 thread processor just because I am a geek, but a newer 22nm lower TDP processor with stock 4-4.4ghz freq. that could be overclocked extra is a better decision. Plus you will want to find triple channel kits of ram which is harder to find. Clock wise they will be the same and I don't think you'll notice a difference really with any games unless you were doing encoding type of stuff.

And I will post pictures of my bios for you rock&roll tomorrow.

I don't think it's that clear cut, firstly, its not particularly easy to overclock any of the available processors beyond 4Ghz without serious cooling. Things have not changed a great deal from 32nm to 22nm die size has degreased which means more heat per mm2 so clock speeds are hardly any faster in the newer 4770 and 4790s.

I recently purchased a 4770k for my HTPC and that system with a top end air cooler is very difficult to get to 4.2Ghz from 3.9! the chip gets uncomfortably hot, 90C at stock with memory at 2400Mhz (at that speed it only just matches my X5670 for memory bandwidth). Unless you're willing to go water you wont get very far.

The only thing that would sway me for a new system is the newer tech such as PCIe v3.0 (Increased bandwidth that has yet to be used in practice), native USB3 and SATA 6GB/s and many of the shortcomings can be remedied with add in cards or RAID.

With 6 cores and a very fast subsystem the X58 easily bests any 1150 system.

 

[Cecil]

This is an interesting thread.

I am torn between buying a $70 dollar x5650 and overclock it to 4.4ghz with a $250 X58 motherboard, as vs buying a 4790K and a lga1150 motherboard.

Could you advise?

I don't think it's that clear cut, firstly, its not particularly easy to overclock any of the available processors beyond 4Ghz without serious cooling. Things have not changed a great deal from 32nm to 22nm die size has degreased which means more heat per mm2 so clock speeds are hardly any faster in the newer 4770 and 4790s.

I recently purchased a 4770k for my HTPC and that system with a top end air cooler is very difficult to get to 4.2Ghz from 3.9! the chip gets uncomfortably hot, 90C at stock with memory at 2400Mhz (at that speed it only just matches my X5670 for memory bandwidth). Unless you're willing to go water you wont get very far.

The only thing that would sway me for a new system is the newer tech such as PCIe v3.0 (Increased bandwidth that has yet to be used in practice), native USB3 and SATA 6GB/s and many of the shortcomings can be remedied with add in cards or RAID.

With 6 cores and a very fast subsystem the X58 easily bests any 1150 system.

It depends on usage. For video compression, Quicksync is still faster then a 6 core even with HT.

A Z97 board will also (as far as we know) support Broadwell, so there is still an upgrade in the future.

Z97s have M.2 and SATAe. Some have Ultra M.2, which is extremely fast. The advantage being it goes directly through PCIe, and no RAID screens to slow down boot times.
http://www.thessdreview.com/our-reviews/asrock-z97-extreme6-tests-pcie-m2-and-sata-ssds/

Even before we get into the ASRock Z97 Extreme6 motherboard, the simple fact that we are testing three totally different SSD interfaces merits an understanding of how the speed of each is achieved. SATA 3 for example, does not run directly through PCIe lanes to the CPU, and as such is limited to SATA 3 (6Gbps) throughput of somewhere in the area of 575MB/s maximum in reality. Both the Samsung XP941 and the Plextor M6e M.2 SSDs use PCIe lanes which eliminated that SATA barrier, 4 lanes for the XP941 and 2 lanes for the M6e. In understanding PCIe 2.0 lane throughput, a theoretical 500MB/s is stated as throughput per lane. This means that the M6e can reach 1GB/s (theoretically), whereas the XP941 might reach twice that at 2GB/s. This is never quite possible as a result of hardware components and their constraints.

Most games that are actually effected by CPU, still get most gains from IPC or higher single threaded performance in most cases.


And lastly, Devils Canyon does not run as hot as Haswell (4770K). A 4.5-4.7Ghz OC is reasonable with an AIO water kit or top end air.

There are plenty reasons to go Z97/DC over X58/Westmere.


It comes down to what the user is planning to do with the system.



That's not to take away what X58/Westmere does offer though. But for many users, and me personally, would get more out of a newer setup for the same money.

 

[Rock&Roll]

And I will post pictures of my bios for you rock&roll tomorrow.

Text will actually be easier to digest. What I mean is just fill out a template like the one below that is corrected to match your settings. There may be features missing on this list that exist on your board. So add remove as needed.

Mother Board ( EVGA X58 Classified E-760 )
CPU ( Lapped i7 920 D0 )
CPU Cooler ( Corsair H50 )
Memory ( 3x2GB Patriot Viper 2 - 2000mhz 8-8-8-24 )
PSU ( Corsair HX1000 )
GPU ( 2x GTX-280 SLI )
Drivers ( 191.07 )
Operating System ( Windows 7 Ultimate 64)

Frequency Control
CPU Multiplier ( 21X )
CPU Host Frequency (Mhz) ( 201 )
MCH Strap ( Auto )
CPU Uncore Frequency (Mhz) ( 4020Mhz { 20X } )
CPU Clock Skew ( 0 ps )
Spread Spectrum ( Disabled )
PCIE Frequency (Mhz) ( 100 )

Memory Feature        
Memory Speed ( Standard )
Memory Control Setting ( Enabled )
Memory Frequency ( 1333Mhz / 2:10 )
Channel Interleave Setting ( 6 Way )
Rank Interleave Setting ( 4 Way )
Memory Low Gap ( Auto )
tCL Setting ( 8 )
tRCD Setting ( 8 )
tRP Setting ( 8 )
tRAS Setting ( 24 )
tRFC Setting ( 74 )
Command Rate ( 1t )

Voltage Control
EVGA VDroop Control ( Without VDroop )
CPU VCore ( 1.45000 )
CPU VTT Voltage ( +350 )
CPU PLL VCore ( 1.5 )
IOH PLL VCore ( 1.5 )
DIMM Voltage  ( 1.65 )
QPI PLL VCore ( 1.3 )
IOH VCore ( auto )
IOH/ICH I/O Voltage ( auto )
ICH VCore ( auto )
**VTT PWM Frequency ( 370  )
**CPU PWM Frequency ( 1210 )
**CPU Impedance ( Auto)
**QPI Signal Compensation ( Less )

CPU Feature
Intel SpeedStep ( Disabled )
Turbo Mode Function ( Enabled )
CxE Function ( Disabled )
Execute Disable Bit ( Disabled )
Virtualization Technology ( Disabled )
Intel HT Technology ( Disabled )
Active Processor Cores ( All )
QPI Control Settings ( Enabled )
QPI Link Fast Mode ( Enabled )
QPI Frequency Selection ( 4.8gt )
OC Recovery ( Enabled )

 

[GonzoP]

I can check later to confirm, but I think my 195x24 setting is running 1.39v to 1.41v. Now I do have LLC enabled for that setting so under a load I usually see ~1.45v. To be honest, unless someone is wanting to just make a run at 5GHz for benchmarking e-peen purposes there is no point to going over the 5650 assuming you have a board capable of over 200BCLK. It may increase your chances of getting an excellent chip, or maybe not, I don't think there is really enough data to say as of yet as there aren't as many people with 5670/5675s. This particular chip cruised to 4.8GHz, but I had to bear down on it a bit for 4.9, and have yet to get a benchable 5GHz. So it has a pretty hard ceiling, and I've went up to 1.62v trying. My Bloodrage itself is capable of at least 220 that I know of, that's as far as either the board could go or my W3520 could carry it.



That's as fast as I can go at a voltage I feel comfortable with for extended durations.

1.62v yikes. My courage ends at 1.5v

 

[Zoson]

I have to admit, I'm tempted to buy an x5670 and kill it going beast mode.

 

[Zoson]

It depends on usage. For video compression, Quicksync is still faster then a 6 core even with HT.

While this IS true, it's also not a good reason to upgrade. As a streamer, I was really excited for QSV, but it wasn't all it was cracked up to be. The quality of the video is not anywhere near the quality you get when doing a CPU encode.

H/x.264 have many parameters you can use to control what the encoder spends bitrate on, and how well it analyses the picture. QSV is hardware on-chip that creates results worse than the 'veryfast' preset, which is the default preset for both OBS and xSplit. Basically, any 4 core(including like q6600) can encode veryfast in real time with a negligible performance impact.

I encode on my CPU using a modified medium preset, at 60FPS, so there's no way QSV could ever match the quality produced.

QSV is great for grandparents who need easy conversion, and decidedly not great for anyone on this forum who dabbles in overclocking, as the depth of knowledge to create good CPU encodes is significantly less than the depth of knowledge to get a good, stable overclock.

A Z97 board will also (as far as we know) support Broadwell, so there is still an upgrade in the future.

Z97s have M.2 and SATAe. Some have Ultra M.2, which is extremely fast. The advantage being it goes directly through PCIe, and no RAID screens to slow down boot times.
http://www.thessdreview.com/our-reviews/asrock-z97-extreme6-tests-pcie-m2-and-sata-ssds/

This, however, is a really good reason to buy new tech. We're at the end of the line for x58 as far as storage is concerned. Real world benefit hasn't quite got there yet, but I expect it to soon.

Most games that are actually effected by CPU, still get most gains from IPC or higher single threaded performance in most cases.

The IPC difference is very small, and really almost not worth talking about. Intels primary focus has been on the iGPUs. Really case in point is I *had* an x79 system with a 3930k, and it did 4.4GHz totally stable with a Cooler Master GeminII S524. So, under water I would have expected 4.7-4.8GHz stable.

However, at the end of the day the performance difference was simply not worth breaking down my system, swapping all the parts, and then going through the pain of tuning such an overclock.

And lastly, Devils Canyon does not run as hot as Haswell (4770K). A 4.5-4.7Ghz OC is reasonable with an AIO water kit or top end air.

Talking about heat/power savings, did you forget how to [H]ard?

 

[Makaveli@BETA]

It depends on usage. For video compression, Quicksync is still faster then a 6 core even with HT.

At the cost of image quality.

 

[Cecil]

While this IS true, it's also not a good reason to upgrade. As a streamer, I was really excited for QSV, but it wasn't all it was cracked up to be. The quality of the video is not anywhere near the quality you get when doing a CPU encode.

H/x.264 have many parameters you can use to control what the encoder spends bitrate on, and how well it analyses the picture. QSV is hardware on-chip that creates results worse than the 'veryfast' preset, which is the default preset for both OBS and xSplit. Basically, any 4 core(including like q6600) can encode veryfast in real time with a negligible performance impact.

I encode on my CPU using a modified medium preset, at 60FPS, so there's no way QSV could ever match the quality produced.

QSV is great for grandparents who need easy conversion, and decidedly not great for anyone on this forum who dabbles in overclocking, as the depth of knowledge to create good CPU encodes is significantly less than the depth of knowledge to get a good, stable overclock.


This, however, is a really good reason to buy new tech. We're at the end of the line for x58 as far as storage is concerned. Real world benefit hasn't quite got there yet, but I expect it to soon.


The IPC difference is very small, and really almost not worth talking about. Intels primary focus has been on the iGPUs. Really case in point is I *had* an x79 system with a 3930k, and it did 4.4GHz totally stable with a Cooler Master GeminII S524. So, under water I would have expected 4.7-4.8GHz stable.

However, at the end of the day the performance difference was simply not worth breaking down my system, swapping all the parts, and then going through the pain of tuning such an overclock.


Talking about heat/power savings, did you forget how to [H]ard?

While I haven't personally used Quicksync yet, Anandtech's article says its the best way to transcode.
http://www.anandtech.com/show/4083/...-core-i7-2600k-i5-2500k-core-i3-2100-tested/9

That the quality is indistinguishable from using the CPU, while being much faster.


There has been at least a 25-30% improvement in same clock / single threaded performance since Gullftown. While that's not much over the span of 4 years, it still is a substantial enough difference to be counted.


And the heat was mentioned in direct reply to Deimos' concern of Haswell heat.