Is this what happened to make magny ES chips overclockable, but with retail interlagos?
Edit: oh, just the standard multiplier range
Last edited:
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Interlagos has landed
- Thread starter SatTech
- Start date
Edit: oh, just the standard multiplier range
Vaulter98c
[H]ard|DCer of the Month - October 2009
- Joined
- May 21, 2008
- Messages
- 5,817
Do those Tyan boards (4P) require server RAM or can you run standard DDR3 sticks at the rated speeds?
Standard. I just bought cheap Kingston value RAM 1333 1GB sticks for like $9 per dimms
Hi everbody.
Always no detailled benchmarks ?
You're in a sub-forum devoted primarily to folding and related projects. This thread contains folding numbers (ppd, ppd/w, costs per proc), and those are the ones we care about. What more "detailled benchmarks" were you hoping for?
Of course, i would like to see some benchmarks in connection with foldinghome but particularly magnycours vs interlagos.
What is the influence of ddr3 frequency ( 1333 vs 1600 ) ? Does it help a lot to increase PPD ?
Thanking you in advance for sharing results.
which is
given that
He also notes that
And then
and
Which is
So when you run the numbers,
For comparison to MC (again), we have
And as far as changing frequency/OCing on IL,
As far as memory frequency, I don't know much about interlagos other than what I've read here, so I don't know if it officially supports ddr3 1600. If it doesn't, we likely will never have those numbers. If it does, I'm sure we'll have them as soon as someone gets around to trying it.
FrEaKy
[H] Movie and TV Show Review Guy
- Joined
- Jan 31, 2003
- Messages
- 14,113
Way to break down this ENTIRE thread!
Well... the entire thread past page 5, anyway. Once you get past all the "Still waiting for chips.... still waiting for chips.... still still waiting for chips."
Hopefully it'll help someone? It really only takes like 5 minutes to read those 2 pages.
FrEaKy
[H] Movie and TV Show Review Guy
- Joined
- Jan 31, 2003
- Messages
- 14,113
Yeah, but, I didnt even need to read the first two pages, just read your post and I knew Interlago's isnt for me soly on the premises of PPD to PPW
Off course IL officially support 1600 mhz memory.
Moreover, only one member shared his results.
It is not a lot for such a great forum and that is why i asked other results if possible.
However i can understand that IL is not a success at all and that can explain why there is not more results.
My only hope was the improvement coming from better memory bandwith.
Memory wont be able to make up for the 20% or more power these thing pull. ppd/w wise, it just isn't a good idea.
R-Type
[H]ard|DCer of the Month - October 2011
- Joined
- Mar 6, 2006
- Messages
- 2,801
There aren't a lot of results because most of us have cancelled plans to upgrade to Interlagos due to the underwhelming performance. We have a basic frame of reference for how much slower and power hungry IL is compared to MC, I honestly don't think there is any merit to asking for more.
Sadly i was pretty sure of your answer.
In july, i bought this platform with four 6128 opteron, expecting a lot of interlagos.
But it seems that IL is a complet fail.
That is why i am very disappointed and do not want to buy twelve core magnycours again.
So i am going to wait for EVGA SRX ( dual socket ). With two overclocked eightcores sandy-E xeon, i suppose that i will get better PPD than four twelve cores magnycours.
4p 6282 SE: 12m30s - 12m45s P6903 (w/DLB); 770-780W at the wall.
That's with stock CL9/DDR3-1333 MHz memory. Faster memory may
shave 20, maybe 30 seconds off that. Will post numbers when I have
them.
For comparison: 4p 6174 @ 2.475 and premium memory
(6-6-5-18/DDR3-1200) achieves similar performance (10s slower)
with 115W less at the wall (same PSU model).
And $4.4k (four new 6282s) budget allows procurement of four
used 6174s or 6176s with quite some headroom...
EDIT: with 2P SB-E you'd need 4.2GHz+ OC to match that
(extrapolated from desktop SB-E)
That's with stock CL9/DDR3-1333 MHz memory. Faster memory may
shave 20, maybe 30 seconds off that. Will post numbers when I have
them.
For comparison: 4p 6174 @ 2.475 and premium memory
(6-6-5-18/DDR3-1200) achieves similar performance (10s slower)
with 115W less at the wall (same PSU model).
And $4.4k (four new 6282s) budget allows procurement of four
used 6174s or 6176s with quite some headroom...
EDIT: with 2P SB-E you'd need 4.2GHz+ OC to match that
(extrapolated from desktop SB-E)
R-Type
[H]ard|DCer of the Month - October 2011
- Joined
- Mar 6, 2006
- Messages
- 2,801
tear, it looks like the tight memory clocks give you a significantly greatly IPC advantage. Your 6174 rig is pulling down close enough to be equal frame times with 16 less cores and 17.5% lower clocks. This is as compared to the 8% difference we saw earlier. Am I missing something in the comparison?
I'm hopeful SR-X will give us clocks over 4.2 as we see that speed on SR-2's, in that case you really get the best of both worlds with the high ppd output while not sacrificing single threaded performance.
Thanks a lot , Tear.
i am a big fan of this thread : http://www.amdzone.com/phpbb3/viewtopic.php?f=521&t=138490
With 2P SB-E my goal is 4.5 ghz because my 3930k can do 4,9 ghz and more easily .
For example, at 4.8 ghz, the 3930k achieve 150000 ppd ( WU 6903 ).
i am a big fan of this thread : http://www.amdzone.com/phpbb3/viewtopic.php?f=521&t=138490
With 2P SB-E my goal is 4.5 ghz because my 3930k can do 4,9 ghz and more easily .
For example, at 4.8 ghz, the 3930k achieve 150000 ppd ( WU 6903 ).
Thanks a lot , Tear.
i am a big fan of this thread : http://www.amdzone.com/phpbb3/viewtopic.php?f=521&t=138490
With 2P SB-E my goal is 4.5 ghz because my 3930k can do 4,9 ghz and more easily .
For example, at 4.8 ghz, the 3930k achieve 150000 ppd ( WU 6903 ).
- I don't think they will clock that high
- 1 Ghz lower starting point
- locked low multi bus clock only
- even with perfect scaling... at that clock speed... it would match a 4p 6282 rig
I think SRX all depends on how the overclocking works. If there is just a BCLK multi (x1.25, x1.66) as on current X79 boards then overclocking could be very dependant on the turbo ratios of the CPUs you are using. I assume all of the cheap xeons will need to use the stock turbo multipliers so then a cpu like this would hopefully hit 100 x 1.66 x 26? = 4.3 ghz. A 26 multi on a 2000mhz base chip like this would be assuming you have slight headroom like with non-K i7 and i5 chips on 1155 where you are allowed to set the multi 4 bins higher than the max turbo for that number of active cores.
happiwakko
n00b
- Joined
- Dec 3, 2004
- Messages
- 15
Is it true that a 3930k at 4.8ghz can do 150k ppd?
Hi .
Is it possible to increase base clock of supermicro board with TurionPowerControl 0.41 ?
I tried but don't see any option to achieve it .
Anyway, i thought it was possible.
If it was that easy....there would be a guide for it...
TPC only sets clocks for ES chip and does so by changing the multiplier.
tear is working on tools to make the overclock easier...
because you are adjusting the bus.... it affects the ram as well...
So you have to adjust timings based on what the frequency will be and flash the spd settings to the ram before you adjust the cpu clock.
Then there is the actual cpu clock change...
Its tricky stuff... come into IRC ... tear has been around more...
ask nicely and you will probably be helped...
I think you've lost me
MCs run at 2475MHz. ILs at (on average, enabled turbo) 2770 MHz.
Though MCs run w/premium memory here so it's not exactly apples
to apples comparison (would need to remove HSFs to install and
I haven't gotten around to it yet).
Stock 6174s with stock memory should approx. give 14m49s.
Using simple arithmetic one gets clock difference factor of ~1.259 (2770/2200)
and performance difference factor of ~1.17 (14.8166/12.6666).
Dividing these two: ~1.076 -- in favor of MC, per clock (whole-chip-wise;
completely ignoring core number difference).
Is that what you meant when you said 8%?