Cherry picking a 5960X

Status
Not open for further replies.

dr/owned

Gawd
Joined
Feb 21, 2012
Messages
577


Going to find out which one overclocks the best.



I won't explain this distribution (nor the axes scale or units), but ideally you want to be as far to the right and as far down as possible. It's an interesting exponential correlation, and I'm not sure which axis matters more so that's what I'm here to find out (ie would you trade off being further to the right if it means you're further up?). Yellow marks where my current 5960X sits at. The orange dots are the three 5960X's I'll be physically testing. Green is just a random subset of data so I could visualize what's going on within the solid wall of blue.



I'll give you a few minutes to figure out what's wrong with this configuration.

Nope you're wrong. The correct answer is I can't remove the CPU waterblock without removing the entire board from the case. Well...I could I guess drain the CPU loop and detach the tube connecting the VRM block, but I don't really trust being able to get a 100% drain and not have enough still in the block that spills on the mobo. So yay for giant pain in the ass having to remove the entire board to swap test a new cpu!

So I'm waiting until this weekend. Plus I have a cheapo GT 210 in the mail so I can unplug my video cards as one piece, leave them filled with water, and just use the 210 to give me some video support to see what I'm doing (easier to do that then unplug and replug in 10 lbs worth of SLI video cards + waterblocks 4 times over)
 
Last edited:
It looks like a mad scientist contraption. I look forward to seeing your results. Mine is stable at 4.3GHz with a closed loop cooler.
 
I said screw it and decided to see if I could bang this experiment out on a Wed. night.

Currently loaded up with the chip circled in red. It should be better than the old / first chip in every way because it's both lower on the Y axis and further to the right on the X axis.



Here's the first chip (look at max temperature), and core 2 would die after about 2 minutes of P95 @ 1.35V:



Here's the second / current chip and 10 minutes later it's still got all cores 100% in P95.



Here's my "ghetto slamming CPUs in and out" setup. 2 screws holding the mobo in, GT210 graphics card:



It's bright because I have a desk lamp down there so I can see what's going on.

EDIT: And one of the P95 threads on core 4 died after 12 minutes or so. Moving on to the next chip.
 
Third one suck balls with threads dying seconds after starting P95. I even tried +.05 core voltage and it didn't really help:





Some of the core temps aren't relevant because threads are dead.
 
If you want the best, highest-yielding, coolest-running 8C die, simply buy an E5-1680 V3. No point in buying a ton of i7-5960Xs to try an find one as good as an E5-1680 V3. I'd be willing to bet that if you bought one E5-1680 V3, it would beat most of your i7 collection and might be equalled by one of your i7s. The E5-1680 V3s uses only the pick of the litter when it comes to 8C dies...;)

Just buy the E5-1680 V3 and forget about it. My E5-1680 V2 is running very happily at [email protected], once I got a hold of more Arctic Silver 5 (I had been using Alumina for my initial tests) and the CPU peaks under load at 55C (as indicated in XTU) with my Corsair H110. It could probably get to 4.7 or 4.8GHz if I was inclined to push the envelope.

Buy one and test it as a comparison...;)
 
4th and final chip (the one furthest to the left on the distribution) sucked the worst. Couldn't even get it to boot at 4.6 Ghz.

So conclusion: golden die that overclock like monsters are furthest to the right possible on the distribution and Y axis position plays less of a role. I've put in the chip from post #3 and that's what I'm going to roll with.

The other 8 chips weren't anywhere remarkable on the distribution, so not even worth testing out physically.

^^ 1680v3 doesn't appear to be for sale anywhere.
 
4th and final chip (the one furthest to the left on the distribution) sucked the worst. Couldn't even get it to boot at 4.6 Ghz.

So conclusion: golden die that overclock like monsters are furthest to the right possible on the distribution and Y axis position plays less of a role. I've put in the chip from post #3 and that's what I'm going to roll with.

The other 8 chips weren't anywhere remarkable on the distribution, so not even worth testing out physically.

^^ 1680v3 doesn't appear to be for sale anywhere.

They aren't as common as an i7-5960X, but can be found...


At Amazon through DECTrader. DECTrader is a decent company...ordered from them in the past.

http://www.amazon.com/Intel-CM8064401547809-XEON-E5-1680V3-2133MHZ/dp/B00QX0W59A


NCIX's business side, prices in CDN$

http://www.ncixbusiness.com/product...Processor_Socket_FCLGA2011OEM_Pack-1028243638


Listed as preorder at Acmemicro...give 'em a call/e-mail and see...theie site often doesn't update when they get them in stock.

http://www.acmemicro.com/Product/14...0M-Cache-3-20GHz-8-Core-16Thread-TRAY?c_id=17


The bottom line is that the i7s are reject chips that are offered at a cheap price. The truly exceptional quality (high speed, low leakage) dies are nearly always reserved for the Xeons, so you need to go Xeon if you want the best..;).
 
As an Amazon Associate, HardForum may earn from qualifying purchases.

The bottom line is that the i7s are reject chips that are offered at a cheap price. The truly exceptional quality (high speed, low leakage) dies are nearly always reserved for the Xeons, so you need to go Xeon if you want the best..;)
.

I'm not so sure of that. In the distribution posted above, those are all either Xeon or Extreme Edition chips. The 1680's might come from the area around the red X, but every other point in the distribution is put on shelves as some flavor. Presumably Xeons significantly outsell Extreme Edition parts since they're ordered in bulk vs. 1 at a time. I'm sure the shitty 1.8 Ghz parts are probably from the short-bus part of the distribution :)

EDIT: Also overclocking update, Chip #2 seems to like 4.6 @ 1.325V more than 4.6 @ 1.35V. 5960X seems to start glitching past 90C so lower voltage -> lower temperature seems to be the correlation that is currently giving me a more stable overclock at a lower voltage. As predicted, this chip is better than Chip #1 (what I was running for the past 4 months) because it wouldn't survive at 4.6 @ 1.35V for more than a few minutes. @ 1.325V it would bsod in P95. I was able to boot Chip #2 @ 4.8 Ghz 1.35V, but half the threads would die in P95 immediately. I tried 4.8Ghz @ 1.5V, and still 4 threads died immediately. I could probably CPU-z validate @ 5 Ghz, but it wouldn't be stable for anything.
 
Last edited:
I'm not so sure of that. In the distribution posted above, those are all either Xeon or Extreme Edition chips. The 1680's might come from the area around the red X, but every other point in the distribution is put on shelves as some flavor. Presumably Xeons significantly outsell Extreme Edition parts since they're ordered in bulk vs. 1 at a time. I'm sure the shitty 1.8 Ghz parts are probably from the short-bus part of the distribution :)

EDIT: Also overclocking update, Chip #2 seems to like 4.6 @ 1.325V more than 4.6 @ 1.35V. 5960X seems to start glitching past 90C so lower voltage -> lower temperature seems to be the correlation that is currently giving me a more stable overclock at a lower voltage. As predicted, this chip is better than Chip #1 (what I was running for the past 4 months) because it wouldn't survive at 4.6 @ 1.35V for more than a few minutes. @ 1.325V it would bsod in P95. I was able to boot Chip #2 @ 4.8 Ghz 1.35V, but half the threads would die in P95 immediately. I tried 4.8Ghz @ 1.5V, and still 4 threads died immediately. I could probably CPU-z validate @ 5 Ghz, but it wouldn't be stable for anything.

The E5-1680 V3 probably isn't a huge seller, due mainly to it's high price, and is used in mainly OEM workstations where the increased base clock gives an advantage. The EE is also sold in OEM form and is purchased in that form quite often, with the DIY market being the predominant purchaser of the retail variant . The E5-1680 V3's increased base clock means that Intel needs to bin the chip higher to maintain the advertised TDP (which is the same as the lower clocked 3.0GHz chips), making it a superior quality chip. When overclocked, this superior binning should make it the best overclocker. Any surplus dies that are produced in excess of the market's requirements for E5-1680 V3's and E5-2667 V3's (the two top 8-core SKUs) are made into lesser SKUs like the E5-1660 V3 and the i7-5960X, with the E5-1660 V3 very likely getting first dibs on these surplus dies. So in reality, it's only after the E5-1680V3, the E5-2667V3 and the E5-1660V3 have had their demand all satisfied does the i7-5960X receive a top bin die, making this likelihood exceedingly remote. The i7 therefore typically makes do with lesser dies that may make the advertised clock speed, but lack the top pedigree of the best dies (ie: don't clock as high, leakier transistors, etc).

The reason why we have the i7 at all is that enthusiasts, after having a 6-core top SKU for both Sandy Bridge-E and Ivy Bridge-E would be unlikely to purchase Haswell-E if it topped out at 6 cores again, given DDR4's immaturity and the introduction of undesirable aspects such as the FIVR. Having the i7-5960X gives Intel renewed enthusiast demand for their top i7 SKUs and allows them to maximize their margins. Intel uses this demand to dispose of dies that didn't make the grade to be used as Xeons, with increased yields from the mature 22nm process making sufficient lower grade 8-core dies available to meet i7 demand. If there was no 8-core SKU, there would be much less demand for the Extreme Edition product, with more people buying the mid-grade (and much lower margin) SKU, forcing Intel to neuter fully functional 8-core dies to satisfy demand for this SKU. It's more profitable for them sell a fully-functional lower grade 8-core die as an EE than to neuter it down to a 6-core and sell it as a mid-grade hex core SKU.

Most enthusiasts who want an 8-core CPU have no real choice except to accept FIVR and the currently questionable value of DDR4. The only other option is the E5-1680V2, which is the only unlocked 8-core for X79/DDR3 and is such a non-mainstream and expensive SKU as to not really be an option for most people. This encourages a more rapid adoption of DDR4 and allows them to set the stage for Skylake in the future.
 
Interestingly, the E5-1660 V3 and the 5960X seem to have the same stock speeds and same amount of L3 cache, and the Xeon costs "only" about $200 more on Amazon.

Would you still say that the E5-1660 V3 is binned a bit higher than the 5960X? If the E5-1660 V3 is also fully unlocked, that could make it a very viable alternative to the 5960X if one didn't want to play the silicon lottery. I'd be pretty enticed if I was upgrading to X99, but I'm waiting for Skylake-E so it'll be a while.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Interestingly, the E5-1660 V3 and the 5960X seem to have the same stock speeds and same amount of L3 cache, and the Xeon costs "only" about $200 more on Amazon.

Would you still say that the E5-1660 V3 is binned a bit higher than the 5960X? If the E5-1660 V3 is also fully unlocked, that could make it a very viable alternative to the 5960X if one didn't want to play the silicon lottery. I'd be pretty enticed if I was upgrading to X99, but I'm waiting for Skylake-E so it'll be a while.

I would think that one would have a better chance of getting one of the top dies from an E5-1660 V3 than an i7-5960X, but they are the same base clock and same TDP, so my guess is that the E5-1660 V3's are a mix of surplus top dies and lower grade dies. The exact ratio is probably reliant on Intel's week-to-week yields. Given the premium price of the E5-1660 V3 over that of the i7-5960X, I'd be more inclined to simply buy an E5-1680 V3 once and be assured of a top die than to roll the dice on an E5-1660 V3.

My $0.02...;)
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Eh when Intel comes out with that 10 core E5-1680 V5 (or V4 if they can Broadwell) I might have some interest. ;)
 
If you want the best, highest-yielding, coolest-running 8C die, simply buy an E5-1680 V3. No point in buying a ton of i7-5960Xs to try an find one as good as an E5-1680 V3. I'd be willing to bet that if you bought one E5-1680 V3, it would beat most of your i7 collection and might be equalled by one of your i7s. The E5-1680 V3s uses only the pick of the litter when it comes to 8C dies...;). Just buy the E5-1680 V3 and forget about it. My E5-1680 V2 is running very happily at [email protected], once I got a hold of more Arctic Silver 5 (I had been using Alumina for my initial tests) and the CPU peaks under load at 55C (as indicated in XTU) with my Corsair H110. It could probably get to 4.7 or 4.8GHz if I was inclined to push the envelope. Buy one and test it as a comparison...;)

Maybe too much entusiasm here. I have the E5-1680v3 with H240-X watercooling and stock paste in the Asus X99-E WS mobo plus 128 GB ECC RAM. Not done extensive overclocking yet just standard Asus optimization which is getting its up to 4.4 GHz. This is not different from the 5960X.
 
Maybe too much entusiasm here. I have the E5-1680v3 with H240-X watercooling and stock paste in the Asus X99-E WS mobo plus 128 GB ECC RAM. Not done extensive overclocking yet just standard Asus optimization which is getting its up to 4.4 GHz. This is not different from the 5960X.

Give it a try with manual overclocking and let us know what you get. Your board probably is just using default overclocking values for an unlocked 8-core...I'll bet your chip is capable of much more.
 
Give it a try with manual overclocking and let us know what you get. Your board probably is just using default overclocking values for an unlocked 8-core...I'll bet your chip is capable of much more.

Yeah - that's a pretty amazing "auto" OC (4.4 GHz). My 5960X is seeming to be stable at only 4.2 GHz. I've been testing back from 4.5 GHz and going up to 1.35v. All with manual tuning.
 
I managed to steal some time in my day and play a little more with my E5-1680 V2. Stable at [email protected] (BIOS Vcore), but I need to play with the Vcore LLC a bit more and maybe take it up a touch (Vcore droops to 1.264V under full load at 4.6GHz). Temps peaked at 59°C just once, and were typically around 55°C ±4°C. All on a run-of-the-mill, 1.5 year-old Corsair H110 with stock fans...;)
 
I managed to steal some time in my day and play a little more with my E5-1680 V2. Stable at [email protected] (BIOS Vcore), but I need to play with the Vcore LLC a bit more and maybe take it up a touch (Vcore droops to 1.264V under full load at 4.6GHz). Temps peaked at 59°C just once, and were typically around 55°C ±4°C. All on a run-of-the-mill, 1.5 year-old Corsair H110 with stock fans...;)

Dear God man, just bit the bullet and do a custom loop and give that chip 1.4V and see 8 5Ghz cores!!!:D:D..The investment of a custom loop can be done on the cheap, especially if you go used...But you have deep pockets so just get building:p!
 
Dear God man, just bit the bullet and do a custom loop and give that chip 1.4V and see 8 5Ghz cores!!!:D:D..The investment of a custom loop can be done on the cheap, especially if you go used...But you have deep pockets so just get building:p!

The cost of a custom loop is no problem...the time necessary to source all the parts, do the build, deal with the inevitable build snafus, reorder a few odds and ends...big problem. And getting any appreciable time to devote to such a project between now and September is looking exceedingly remote as well...:(
 
The cost of a custom loop is no problem...the time necessary to source all the parts, do the build, deal with the inevitable build snafus, reorder a few odds and ends...big problem. And getting any appreciable time to devote to such a project between now and September is looking exceedingly remote as well...:(

I just realized you are running TRI-SLI, and that is indeed not a small feat when it comes to building a loop for that many components. Would make for a nice project next fall no? In the mean time, crank that voltage up. There are many of us that ran 32mn parts with 1.4~1.5V for years now without an issue, and you certainly have the thermal headroom..Even temperatures in the 70s at load are chilly for these parts..

I really wish I were able to work full time. My last job was with a firm that dealt with large commercial/industrial construction, and I lived with AutoCAD open. It loved my heavily OC'd AMD 1055T hexcore @ 4.3Ghz, and the 970 @ 4.4Ghz that replaced it (was a pointless upgrade to satisfy the itch I had)..

I spend days/weeks making sure my OC's are 48~72 hour Prime Stable, and once they get to that point they have never let me down. Even with Haswell, this not using Prime95 crap doesn't sit well with me. It has always been a torture test for a reason. It's the best there is..Thankfully dr/owned seems to believe that as well.:D

dr/owned;1041450136[SIZE="4" said:
]4th and final chip[/SIZE] (the one furthest to the left on the distribution) sucked the worst. Couldn't even get it to boot at 4.6 Ghz.

I'm confused brother. You had a picture showing 12 CPU's, but you only tested 4 of them?? I assume you worked with a local supplier instead of putting $12K+ outta pocket or on a CC..If not, well can I come live with you?:p
 
Last edited:
I just realized you are running TRI-SLI, and that is indeed not a small feat when it comes to building a loop for that many components. Would make for a nice project next fall no? In the mean time, crank that voltage up. There are many of us that ran 32mn parts with 1.4~1.5V for years now without an issue, and you certainly have the thermal headroom..Even temperatures in the 70s at load are chilly for these parts..

I'd like to devote more time to computer-related pursuits, but at this time, I lean heavily toward fire-and-forget, idiot-proof, reliable solutions like my AIO Corsair H110. Solutions of this nature may not be on the absolute bleeding edge of performance, but they are usually not too far away from it and save me a great deal of time messing around with hardware for what are often incremental gains. It may also pain you to learn this, but my Titans have never been overclocked, not even slightly...;)

I'm definitely not keen to plow that much voltage through this chip, especially considering it's $2100 USD price (to get to my door). I simply plan to use the headroom in the chip that's present at a reasonable voltage. For me, 1.3V is the absolute maximum for 24/7 usage. Can I replace the chip if it dies? Sure I can, but to me it seems unwise (and pointless) to destroy (or even risk) a $2100 USD chip simply to get a couple of hundred more MHz out of it.

I really wish I were able to work full time. My last job was with a firm that dealt with large commercial/industrial construction, and I lived with AutoCAD open. It loved my heavily OC'd AMD 1055T hexcore @ 4.3Ghz, and the 970 @ 4.4Ghz that replaced it (was a pointless upgrade to satisfy the itch I had)..

Sometimes (on days that aren't payday:D), I really wish I could work part time for a year or two, or take a year or two off to pursue hobbies like my computers. I worked nearly 3000 hours last year. This year is looking like it'll put that number to shame...

I spend days/weeks making sure my OC's are 48~72 hour Prime Stable, and once they get to that point they have never let me down. Even with Haswell, this not using Prime95 crap doesn't sit well with me. It has always been a torture test for a reason. It's the best there is..Thankfully dr/owned seems to believe that as well.:D

Agreed. Prime95 has been the standard for many years. I've never been a fan of FIVR, hence my purchase of the E5-1680 V2. However, it seems from what I've read that a CPU can pass Prime95, but be incurring degradation to the FIVR itself due to the very high current draw through the CPU. Definitely not the most resounding endorsement of FIVR, which seems overall to be a very poor idea by Intel. It looks like for FIVR-equipped CPUs, a safe, long term CPU current maximum value needs to be found to avoid FIVR damage, much like 1.4V is the de facto recommended maximum voltage for SB-E/IV-E. I imagine that for Intel to improve the current handling ability of FIVR (for either Haswell-E or Broadwell-E) would require a major re-engineering effort, one that they are highly unlikely to even consider given their current dominating position in x86 CPU performance.

I'm confused brother. You had a picture showing 12 CPU's, but you only tested 4 of them?? I assume you worked with a local supplier instead of putting $12K+ outta pocket or on a CC..If not, well can I come live with you?:p

I was puzzled also, and was quite surprised that someone would spend that much on i7-5960Xs when the E5-1680 V3s are better binned. If one were to blow $12K on CPUs, I'd have gone with 6x E5-1680 V3s, took the top chip of the 6 and then sold the rest (which should still all do better than a run-of-the-mill i7-5960X).
 
I'd like to devote more time to computer-related pursuits, but at this time, I lean heavily toward fire-and-forget, idiot-proof, reliable solutions like my AIO Corsair H110. Solutions of this nature may not be on the absolute bleeding edge of performance, but they are usually not too far away from it and save me a great deal of time messing around with hardware for what are often incremental gains. It may also pain you to learn this, but my Titans have never been overclocked, not even slightly...;)

You may find this interesting:
http://www.corsair.com/en-us/blog/2015/february/i7-5960x_underwater

Obviously - a biased source. But I haven't seen anything like it before.

BTW - what is FIVR?
 
You may find this interesting:
http://www.corsair.com/en-us/blog/2015/february/i7-5960x_underwater

Obviously - a biased source. But I haven't seen anything like it before.

BTW - what is FIVR?

Very interesting. What that article illustrated clearly (at least for me) is that the i7-5960X has issues getting above 4.4GHz with any reliability, regardless of the water cooler being used, due to die heat saturation that's exacerbated as a result of it's weaker binning. This is where the E5-1680 V3 and it's better bin will excel. Wirk was already hitting 4.4GHz on an auto-tune with his E5-1680 V3. The better binning should mean less leaky transistors and less heat at that speed with the Xeon as opposed to the i7 (and a bit less heat from the FIVR as well, as the chip will be drawing less overall current at that speed). The Xeon will probably still hit the same type of heat wall, but should do so at a higher clock speed.

FIVR=Fully Integrated Voltage Regulator. Voltage regulation components that used to be on the motherboard were integrated onto the CPU by Intel. This complicates matters by making the CPU warmer than it should be at a given clock speed and can frustrate overclocking because as one adds voltage, the heat given off by the FIVR increases along with it, compounding the heat issue.
 
FIVR=Fully Integrated Voltage Regulator. Voltage regulation components that used to be on the motherboard were integrated onto the CPU by Intel. This complicates matters by making the CPU warmer than it should be at a given clock speed and can frustrate overclocking because as one adds voltage, the heat given off by the FIVR increases along with it, compounding the heat issue.

Thanks!

I don't think anyone is arguing against that the E5-1680 is superior to the 5960X. The cost alone makes the comparison dubious.
 
Very cool. I'm curious to see what you come up with.



I won't explain this distribution (nor the axes scale or units), but ideally you want to be as far to the right and as far down as possible. It's an interesting exponential correlation, and I'm not sure which axis matters more so that's what I'm here to find out (ie would you trade off being further to the right if it means you're further up?). Yellow marks where my current 5960X sits at. The orange dots are the three 5960X's I'll be physically testing. Green is just a random subset of data so I could visualize what's going on within the solid wall of blue.

Is that a scatter plot of clock speed vs voltage from a 5960X overclocking database somewhere? If so, I'd love to see where the data came from.
 
It's obviously a scatter plot of speed vs. voltage, but how would he know where his chips lie on it before he tested them. Unless this is something that got leaked directly from Intel's QA ???
 
So, you're selling these as "used and abused" after testing them?
 
So, you're selling these as "used and abused" after testing them?

My guess is it is Amazon that is going to get the nice returned units. Either that, or this is the guy that runs siliconlottery.com. :)
 
My guess is it is Amazon that is going to get the nice returned units. Either that, or this is the guy that runs siliconlottery.com. :)

Yep, going back to Amazon. 7 unopened, 3 opened without breaking the seal so as far as anyone cares they're unused. Dishonest...probably...but it gets the job done.

My "old" 4.6 Ghz 5960X I'll put up for sale eventually. I might hang on to it because I have crazy ideas involving decking the IHS or lapping it away completely. Might need to make a thread about that in the Overclocking and Cooling section. I already asked Koolance and EK if they could make a .999 silver version of their waterblocks and they said no, so plan B is delidding a 5960X which is just bonkers and kinda why I like it.

Also, I checked one of the previously sold SiliconLottery "4.6 Ghz" chips that went up on ebay - it was mediocre in its position in the distribution - at best it was slightly above average. So I'm guessing that whatever method he's using to "validate" his chips isn't P95.
 
Yep, going back to Amazon. 7 unopened, 3 opened without breaking the seal so as far as anyone cares they're unused. Dishonest...probably...but it gets the job done.

Dishonest? no, this is WELL past dishonest and right into scumbag territory.
I HATE people that do stuff like that, just buy one and take what you get.
 
Also, I checked one of the previously sold SiliconLottery "4.6 Ghz" chips that went up on ebay - it was mediocre in its position in the distribution - at best it was slightly above average. So I'm guessing that whatever method he's using to "validate" his chips isn't P95.

He says he uses ROG RealBench (http://rog.asus.com/rog-pro/realbench-v2-leaderboard/).

Can you further explain that scatter plot that you posted? You said you use Prime95 data? How did you get it? Very curious. Thanks!
 
He says he uses ROG RealBench (http://rog.asus.com/rog-pro/realbench-v2-leaderboard/).

Can you further explain that scatter plot that you posted? You said you use Prime95 data? How did you get it? Very curious. Thanks!

Generic scatterplot is a happy scatterplot. I purposely want to keep it fuzzy because the point of this thread's exercise was to test a theory of whether it would be possible to go to a store, look at a bunch of chip boxes, and pick out the one that overclocks the best. It's not something that's practical for pretty much everyone, but I wanted to bring you guys along for the ride anyways. I'm pleased though that it seems my original prediction of the red X being the place to go for the best chips was accurate.
 
Generic scatterplot is a happy scatterplot. I purposely want to keep it fuzzy because the point of this thread's exercise was to test a theory of whether it would be possible to go to a store, look at a bunch of chip boxes, and pick out the one that overclocks the best. It's not something that's practical for pretty much everyone, but I wanted to bring you guys along for the ride anyways. I'm pleased though that it seems my original prediction of the red X being the place to go for the best chips was accurate.

OK - is this:

A: Purposely vague
B: Too complicated for me to understand (I re-read your OP and this post - still not getting how you can look at a box to make a determination as to placement in a scatter plot)

Just trying to understand what you're teachin' - but if it is that you literally did just want us "along for the ride" - then maybe this look at what I'm doing but I'm not going to reveal my full methdology? If so - state that and I'll buzz off!

:D
 
OK - is this:

A: Purposely vague
B: Too complicated for me to understand (I re-read your OP and this post - still not getting how you can look at a box to make a determination as to placement in a scatter plot)

Just trying to understand what you're teachin' - but if it is that you literally did just want us "along for the ride" - then maybe this look at what I'm doing but I'm not going to reveal my full methdology? If so - state that and I'll buzz off!

:D

Correct. Sorry to blueball you though :)

Although lutjens now has my interest piqued on the quality of the W series Xeons. This experiment would be a fun repeat for those chips, but they're so damn expensive and hard to acquire. Too bad I don't work for Asus or similar...those guys probably get trays of chips on demand to play with.
 
Haha. Fair enough.

Can you give me this then?

What is your OC methodology? I've been modifying vCore and VCCIN...starting with 4.5GHz, 1.3v, 1.95v and working back (using 1 hour of ROG RealBench as a success criteria). I worked all the way back to 3.7GHz and still could not pass. I'm thinking I am not giving enough juice to something or else I have an EXTREME dog. Temps are great - Corsair H110 AIO; Gelid GC Extreme. Board is MSI X99S XPOWER AC and PSU is Seasonic X-1250 (memory: 16GB Corsair 3000MHz).
 
Status
Not open for further replies.
Back
Top