X79 and/or rampage IV OC'ing thread for those of us still left....

[Zyll Goliat]

A problem with too many non-standard benchmarks is they could be optimized for any arch (intentionally or not.) Common benchmarks tend to be more transparent with general performance. Here found my Passmark 9 results for E5-1680V2 at 4.65GHz: 19557 vs 17607 of the Ryzen 2700X.

View attachment 162594

PassMark - AMD Ryzen 7 2700X - Price performance comparison

The benchmarks you reference from Phoronix do not reveal the full potential of these Xeon chips, as everyone in this thread is already well aware. The chips nominally perform at 4.5-4.6GHz, which needs an overclock to match free boost clocks of the Ryzen refresh CPUs. I don't compare with Ryzen 2, because there's a drastic change in strategy with those to increase cache and pull away (not a bad thing, but it does add cost).

Well then you will be disappointed if you try the latest passmark 10.0...they really crippled X79 platform and some other older platforms a lot.....here my comparison between passmark 9Vs10 bellow:

Xeon@2650 V2*OC*3.4Ghz All Cores

 

[mouacyk]

Well then you will be disappointed if you try the latest passmark 10.0...they really crippled X79 platform and some other older platforms a lot.....here my comparison between passmark 9Vs10 bellow:

View attachment 162613
Xeon@2650 V2*OC*3.4Ghz All Cores

Your conclusion that they crippled X79 platform with v10.0 is invalid from just what you showed. All benchmarks go through a re-basing with new versions, and thus you cannot meaningfully compare cross-version scores of different CPUs (or even the same CPU, unless you are benchmarking the software itself). If you were to show that scores remained consistent for the same Ryzen CPU between version 9 and 10 of passmark, then your conclusion would mean something.

 

[Zyll Goliat]

Your conclusion that they crippled X79 platform with v10.0 is invalid from just what you showed. All benchmarks go through a re-basing with new versions, and thus you cannot meaningfully compare cross-version scores of different CPUs (or even the same CPU, unless you are benchmarking the software itself). If you were to show that scores remained consistent for the same Ryzen CPU between version 9 and 10 of passmark, then your conclusion would mean something.

They changed test algorithms and instead that newer CPU's only gain more points they decided to also "punish"those CPU's without those new instruction by taking the points from them so new CPU gain points comapared to the Passmark 9 test and certain older CPU's(not all) lost points....I really don't care that much but Passmark almost for a decade have more or less same algorithms calculation and their chart was cool to look for comparison between different types of platforms I guess not for long anymore tho' they still using passmark 9 for their main chart but soon they plan to switch on new passmark 10....

 

[mouacyk]

They changed test algorithms and instead that newer CPU's only gain more points they decided to also "punish"those CPU's without those new instruction by taking the points from them so new CPU gain points comapared to the Passmark 9 test and certain older CPU's(not all) lost points....I really don't care that much but Passmark almost for a decade have more or less same algorithms calculation and their chart was cool to look for comparison between different types of platforms I guess not for long anymore tho' they still using passmark 9 for their main chart but soon they plan to switch on new passmark 10....

Regardless... I'm only interested in "apples to apples" comparisons.

 

[phill]

Well then you will be disappointed if you try the latest passmark 10.0...they really crippled X79 platform and some other older platforms a lot.....here my comparison between passmark 9Vs10 bellow:

View attachment 162613
Xeon@2650 V2*OC*3.4Ghz All Cores

Is it me or are all of the results on the right, higher (bar one - CPU Single Threaded)??

 

[Zyll Goliat]

Is it me or are all of the results on the right, higher (bar one - CPU Single Threaded)??

Yeah top right bar(score 2057-2015)is the single thread result why?

 

[phill]

I wish I knew @Zyll Goliath I don't get some of these tests as things are never consistent enough...

 

[Aquinus]

A problem with too many non-standard benchmarks is they could be optimized for any arch (intentionally or not.) Common benchmarks tend to be more transparent with general performance. Here found my Passmark 9 results for E5-1680V2 at 4.65GHz: 19557 vs 17607 of the Ryzen 2700X.

View attachment 162594

PassMark - AMD Ryzen 7 2700X - Price performance comparison

The benchmarks you reference from Phoronix do not reveal the full potential of these Xeon chips, as everyone in this thread is already well aware. The chips nominally perform at 4.5-4.6GHz, which needs an overclock to match free boost clocks of the Ryzen refresh CPUs. I don't compare with Ryzen 2, because there's a drastic change in strategy with those to increase cache and pull away (not a bad thing, but it does add cost).

Did you even look at what I posted? Michael used over 60 different tests to come up with these results. I'll give you that if you observe a single benchmark, that you might not be seeing what you think you're seeing, but when you're coming up with results like this, it's not just a matter of compiler optimizations.

 

[mouacyk]

Did you even look at what I posted? Michael used over 60 different tests to come up with these results. I'll give you that if you observe a single benchmark, that you might not be seeing what you think you're seeing, but when you're coming up with results like this, it's not just a matter of compiler optimizations.
View attachment 162777

I didn't find the clocks for the Ivy-E CPU when I "looked", so one has to assume it's at stock, while the Ryzen can be assumed to boost naturally to its max clocks at stock settings. If you already forgot, my claim is only relevant to overclocked Ivy-E, so it's more apples-to-apples compared to your benchmarks (which is hardly fair comparing stock to "stock" CPUs that are years apart.)

 

[Aquinus]

I didn't find the clocks for the Ivy-E CPU when I "looked", so one has to assume it's at stock, while the Ryzen can be assumed to boost naturally to its max clocks at stock settings. If you already forgot, my claim is only relevant to overclocked Ivy-E, so it's more apples-to-apples compared to your benchmarks (which is hardly fair comparing stock to "stock" CPUs that are years apart.)

Sure, I might get close if I push my 3930k to the absolute max. I might get within arms reach of a modern 6c/12t CPU at stock.

 

[dalekdukesboy]

I have seen this data before...however they did pick a Sandy 6 core when now it is easy to obtain Ivy 6 cores which might show a bit better and certainly the 1680 8 core would show considerably better than a 3960x to say the least. So those results are what they are but I'd be more interested in seeing an ivy particularly 8 cores + tested with same tests...

Case in point....I've got a large OC on this 1680 granted but by no means have I ever bothered to even try to dial in memory, bclk, etc to get the max out of it...I just left it at 125 strap and got the voltages somewhat optimized. But being almost in the 80% range isn't as good as results with old test granted, but still quite good. Like I said I'd be interested to see what the other tests from that article would be like with my setup.

 

[phill]

Is there a 1680 V3 that would be unlocked at all? (I pressume that's Ivy?)

 

[dalekdukesboy]

Is there a 1680 V3 that would be unlocked at all? (I pressume that's Ivy?)

That would be X99 not X79, but I believe there is.

 

[phill]

Must be a very similar model name or something then if it would be V2?? Too many numbers in my head!!

 

[Zyll Goliat]

V1,V2=X79
V3,V4=X99

Must be a very similar model name or something then if it would be V2?? Too many numbers in my head!!

Maybe 1660 V2 tho' that's the 6c/12t CPU........

 

[phill]

I remember the X range but I would have to look up which Xeon is gone from the Sandy to the Ivy... I'm sure there's a few digits change lol

 

[Zyll Goliat]

I remember the X range but I would have to look up which Xeon is gone from the Sandy to the Ivy... I'm sure there's a few digits change lol

Ahh...that is what you looking for I didn't get what exactly you meant at first... Sandy Bridge latest CPU's was made in 2012 april/may/june something like that.....what exact cpu made the first "transition"to the Ivy bridge I have no idea.....

 

[dalekdukesboy]

Did you even look at what I posted? Michael used over 60 different tests to come up with these results. I'll give you that if you observe a single benchmark, that you might not be seeing what you think you're seeing, but when you're coming up with results like this, it's not just a matter of compiler optimizations.
View attachment 162777

I'd love to see this battery of tests for a 4960x and the 1680 v2. I didn't look recently but I thought someone said the methods of this test was explicit enough where you can replicate the tests and run them exact same way?

 

[X79User]

Hi,
i just bought an X79 Extreme11 (Bios F3.40) and was looking to use it with an E5 2690. The cheapest DDR3 32GB (4x8GB, 2 sets of 2x8) i could find is an avexir 2666mhz kit and i am unsure if those speeds would be supported on the motherboard i just bought (officially supports up to 2500). I have not over- or underclocked RAM before but know the general idea behind it (through watching some RAM OC videos).
I would appreciate some advice, thanks.

 

[agent_x007]

@X79User I'm pretty sure a cheap 32GB
2666MHz kit will be DDR4, and NOT DDR3.
So, it's not compatible with X79 boards.

 

[X79User]

@X79User I'm pretty sure a cheap 32GB
2666MHz kit will be DDR4, and NOT DDR3.
So, it's not compatible with X79 boards.

Nah, it's a DDR3 Kit. Other 32GB DDR3 kits in my area are just super overpriced and if i were to buy on ebay, with ebay shipping, cheaper kits come to the same or often higher price as the 2666mhz one.
Pic from seller:

http://imgur.com/Eyas3ZL

I am mainly looking to find out if and how i can underclock this RAM to make it work on my board, and also what kinds of speeds the CPU (E5 2690) i am planning to combo it with would support. I would also like to confirm that these 2x8GB kits can work in quad channel (according to what i've read online, they should but i'm not too sure).
I couldn't find the stock non-xmp speeds, though i think they should work at whatever those speeds are (probably 1333 or 1600).

 

[Aquinus]

I'd love to see this battery of tests for a 4960x and the 1680 v2. I didn't look recently but I thought someone said the methods of this test was explicit enough where you can replicate the tests and run them exact same way?

You bet.

phoronix-test-suite benchmark 1909137-AS-3960XRYZE03

 

[dalekdukesboy]

Nah, it's a DDR3 Kit. Other 32GB DDR3 kits in my area are just super overpriced and if i were to buy on ebay, with ebay shipping, cheaper kits come to the same or often higher price as the 2666mhz one.
Pic from seller:

http://imgur.com/Eyas3ZL

I am mainly looking to find out if and how i can underclock this RAM to make it work on my board, and also what kinds of speeds the CPU (E5 2690) i am planning to combo it with would support. I would also like to confirm that these 2x8GB kits can work in quad channel (according to what i've read online, they should but i'm not too sure).
I couldn't find the stock non-xmp speeds, though i think they should work at whatever those speeds are (probably 1333 or 1600).

Wow that's really fast ram with good timings....

You bet.

phoronix-test-suite benchmark 1909137-AS-3960XRYZE03

Yeah I downloaded it....I kept trying various things even read some of the manual but didn't get it to run any tests.

 

[X79User]

Update:
I got the memory and as i suspected, the motherboard doesn't support the XMP profile.
It works fine at 1333mhz but when i select the XMP profile and save n exit, it resets back to 1333mhz.
Would it be possible to just enable the XMP and then manually set the speed to 2400mhz (a supported speed by my mobo), or would i have to change other memory oc settings as well?

Also, the south bridge temps are quite high (stock, no load at 55*C with SB fan on speed setting 3 at an ambient of 18*C).

EDIT: I tried some 1600mhz dominator memory and that also resets back to 1333mhz after i try enable xmp. After some research, i think it may be my CPU (E5 2640) that is restricting the RAM speeds (as that has max supported 1333mhz), though i did find other users running 1600mhz with an E5 2640.

EDIT2: The two 2x8GB kits work together in quad channel (at 1333mhz).

 

[Zyll Goliat]

Update:
I got the memory and as i suspected, the motherboard doesn't support the XMP profile.
It works fine at 1333mhz but when i select the XMP profile and save n exit, it resets back to 1333mhz.
Would it be possible to just enable the XMP and then manually set the speed to 2400mhz (a supported speed by my mobo), or would i have to change other memory oc settings as well?

Also, the south bridge temps are quite high (stock, no load at 55*C with SB fan on speed setting 3 at an ambient of 18*C).

EDIT: I tried some 1600mhz dominator memory and that also resets back to 1333mhz after i try enable xmp. After some research, i think it may be my CPU (E5 2640) that is restricting the RAM speeds (as that has max supported 1333mhz), though i did find other users running 1600mhz with an E5 2640.

EDIT2: The two 2x8GB kits work together in quad channel (at 1333mhz).

Try manually to set timings&speeds and yeah it is difficult to get beyond 1866 Mhz with 26xx Xeons.....GL