Because they have so much incredible power and influence that they are doing nothing to influence software companies to develop at the correct pace.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
Sandy E benches - disappointing?
- Thread starter ShuttleLuv
- Start date
Because they have so much incredible power and influence that they are doing nothing to influence software companies to develop at the correct pace.
How about this:
dual core went "enthusiast mainstream" in ~2005
quad core went "enthusiast mainstream" in ~2008
It's 3 years later, 2011 is almost over, and the only 8 core "enthusiast mainstream", Bulldozer, is lukewarm, at best. And, Intel's "enthusiast mainstream", is still very much quad core.
Sure the "enthusiast mainstream" quads today are a bit quicker than the quads from 2008, but my nearly brand new e3-1220 is not going to double the performance of my q9450 in the majority of my software(nor would a 2600k or a e3-1275). I only bought it, because I'm tired of swapping, when I exceed 8 gigs of memory usage. Computation ability wise, it's still quite decent(70-80%) compared to a similar system today.
With single threaded performance stalling, the majority of software is not going to start getting markedly better until "good" 8-10+ core machines are "enthusiast mainstream"(really more like "mainstream mainstream"), and the software vendors are dragged kicking and screaming, to start developing proper generalized multi-threaded apps. There's a big difference in how you think about optimizing software for 2, 4, even 6 cores, and how you think about optimizing software for 8-10+ cores. Compared to the advancement rate in 2008, I'd guess, we've now fallen at least a year behind, on the hardware being able to begin to force the software development/optimization process to make that jump.
dual core went "enthusiast mainstream" in ~2005
quad core went "enthusiast mainstream" in ~2008
It's 3 years later, 2011 is almost over, and the only 8 core "enthusiast mainstream", Bulldozer, is lukewarm, at best. And, Intel's "enthusiast mainstream", is still very much quad core.
Sure the "enthusiast mainstream" quads today are a bit quicker than the quads from 2008, but my nearly brand new e3-1220 is not going to double the performance of my q9450 in the majority of my software(nor would a 2600k or a e3-1275). I only bought it, because I'm tired of swapping, when I exceed 8 gigs of memory usage. Computation ability wise, it's still quite decent(70-80%) compared to a similar system today.
With single threaded performance stalling, the majority of software is not going to start getting markedly better until "good" 8-10+ core machines are "enthusiast mainstream"(really more like "mainstream mainstream"), and the software vendors are dragged kicking and screaming, to start developing proper generalized multi-threaded apps. There's a big difference in how you think about optimizing software for 2, 4, even 6 cores, and how you think about optimizing software for 8-10+ cores. Compared to the advancement rate in 2008, I'd guess, we've now fallen at least a year behind, on the hardware being able to begin to force the software development/optimization process to make that jump.
On the gaming side, it is because developers are developing for consoles. On the productivity side, I think there have been some pretty remarkable strides made in the past few years. Think photo and video editing, rendering, high level simulations - all (or almost all) those things are highly multithreaded and seriously outperform stuff from just a few years ago.
I have set aside 2k to upgrade my I7-860 system that I have been running since launch. I am extremely disappointed. Socket 2011 with the water gun heatsinks, on an enterprise level board, with a processor that should be in a server is confused at best, way over priced for the target audience. And they did not develop this for the enthusiast.
It's not a bad product, it's more like buying a loaf of white bread opening it up and getting wheat. Having something put out there and saying this is what you wanted, I give you Sandy Bridge Extreme... Umm ok - not really. It's more like Sandy Bridge Stretched.
It's not a bad product, it's more like buying a loaf of white bread opening it up and getting wheat. Having something put out there and saying this is what you wanted, I give you Sandy Bridge Extreme... Umm ok - not really. It's more like Sandy Bridge Stretched.
DeathPrincess
Fully [H]
- Joined
- May 15, 2010
- Messages
- 18,205
Don't the PS3/xbox not both have multi core CPUs? Its probably more due to laziness. Or to the vast number of PC users with their crap 7 year old systems and laptops.
One of the bigger strides is GPU rather than CPU processing. the hundreds of "cores" in a GPU allow for mass multi threadedness to great effect. Being able to realtime edit things means workflows are easier to work with, and being able to do things with a sub $400 graphics card than dual 6 core xeons can't means costs are much lower.
- Joined
- Mar 18, 2010
- Messages
- 14,454
It is an enthusiast product. They just made it more high level than 1366 was, since the 2600k and 2500k were already extremely well performing CPUs. The gap between 2011 and 1155 is wider than the gap between 1366 and 1156 was.
Yes, they do have multi-core CPUs. However, due to the design of consoles, almost all work is done by the CPU, and hardly any work is done by the GPU. It's the opposite in computers, most of the work should be done by the GPU, but it isn't. Things that would be more effectively run on the GPU are instead run on the CPU because that's how they designed it on the consoles. At least, that is what I think is happening.
So why the hell does Intel design this CPU/motherboard combos to let you pack on even more memory, but still doesn't support ECC RAM? I'm sorry, but at this CPU pricing level, I can afford the couple of extra bucks for ECC sticks and not have to worry about bit errors, especially when you are pushing the > 16+GB RAM amounts. Filling all the DIMMs with high density RAM just increases the bit error likelihoods exponentially and not having ECC RAM when it barely costs more than standard RAM seems kinda stupid..
Going to a quad channel config also makes up for any tiny reductions of ECC memory performance.
Going to a quad channel config also makes up for any tiny reductions of ECC memory performance.
With regards to the various comments pro and con on Intel knowing about BD waaaaaaaaay before any of us did. I don't want to insult anyone so I'm not directing this to any particular person but it boggles the imagination that anyone in the year 2011 would still believe that ANY large corporation operates "by the rules." Anyone thinking that is either incredibly naive, has never held a job higher than gas station attendant, or actually believes that corporate integrity is not an oxymoron. I've been involved in mid and high level management of corporations either as an exec, a consultant, or a freelancer since I was in my early 20s and I can assure you that the dirty tricks and outright espionage/sabotage that goes on in EVERY large corporation would make your head twirl. This is not just silicon, but everywhere! So if you think that Paul O. finds out about FX benchies by reading forums, then I'll meet you on the roof and we can go flying over London with Peter Pan tonight! 
That was my conclusion as well. With a Xeon(for ECC support), and some decent 2 socket workstation boards, this platform could be worth the premium. But without ECC support, and only 6 cores, it's just an expensive toy. And it even fails at being a toy, since it's no faster in 99% of games than socket 1155. Slower in fact, since it's overclocking ceiling is lower.
Here's hoping the Xeons come out, preferably alongside some nice 2 socket workstation boards, in early 2012.
burningrave101
[H]F Junkie
- Joined
- Sep 9, 2003
- Messages
- 11,825
Yeah I beg to differ on that. The 3930k for example is basically a 990x with Sandy Bridge's overclocking ceiling. It runs hotter and your less likely to find a chip that will hit 5Ghz on air-cooling but the average overclock for a 2500k/2600k is not 5Ghz, it is more around 4.6-4.8Ghz and the 3930k/3960x are perfectly capable of hitting that. And even against the 2700k which is clocked higher SB-E still comes out ahead in gaming benches when it's actually a game that's not GPU limited like Far Cry 2 and Crysis 2 which seem to be about the only two games half the reviewers like to use since they come with built-in benchmarks.
You can see here that the 3960X outperforms the 2700k by a fair margin in HardwareHeaven's benches for MW3, Skyrim, and F1 2011:
http://www.hardwareheaven.com/revie...-3960x-processor-review-modern-warfare-3.html
You can see the same thing in testing World of Warcraft:
http://www.tomshardware.com/reviews/core-i7-3960x-x79-sandy-bridge-e,3071-17.html
Is SB-E's additional cost worth it for a gaming only rig? No, probably not unless you just want the very best there is available, but if your PC is actually used for real work that requires real processing power then the additional cores on the 3930k/3960x can make a very worthwhile difference in multithreaded applications. If your PC is part of your livelihood those extra seconds and minutes shaved off your work means more money.
Last edited:
If your PC is used for work beyond running office, then it's worth getting ECC memory for it. ECC memory tells you straight up, oops I just detected a bit error, and heck I even corrected it, so you might want to replace me. Non ECC memory just silently corrupts things when it fails. Yes usually it just results in a miscolored pixel, but it can and does cause unnecessary crashes, and file system corruption/loss. There's a reason all serious iron on the planet uses ECC memory.
To the extent you choose to trust non-ECC memory with your livelyhood, you can verify it with memtest86+. But that takes hours to days to do. If you have hours to days to play with verifying your memory every few months, then you don't need to make every second count, and you don't need to pay twice as much for 33% more performance.
And these problems only increase as the amount and density of memory increases. So a platform, that has extra DIMMs as one of it's main advantages, and doesn't offer ECC, is just brain dead.
And for that matter, if you really need 33-50% more performance, you could probably really use 100% or 200% more performance, and it's time to go dual socket, or start looking at clusters. This platform doesn't yet offer dual socket, and it's not cost effective for clusters.
To the extent you choose to trust non-ECC memory with your livelyhood, you can verify it with memtest86+. But that takes hours to days to do. If you have hours to days to play with verifying your memory every few months, then you don't need to make every second count, and you don't need to pay twice as much for 33% more performance.
And these problems only increase as the amount and density of memory increases. So a platform, that has extra DIMMs as one of it's main advantages, and doesn't offer ECC, is just brain dead.
And for that matter, if you really need 33-50% more performance, you could probably really use 100% or 200% more performance, and it's time to go dual socket, or start looking at clusters. This platform doesn't yet offer dual socket, and it's not cost effective for clusters.
I agree whole heartily. This was a huge misstep by Intel. A missed opportunity and a platform that really does not fill any segment well while being way out of reach for many, and too over priced with it's short comings for the crowd that may see value in it.
No it wasn't. This is all about Xeon, just like 1366 was. i7 on these sockets is an abomination. The real target market will run one, two or four of these bad boys in a single computer in order to make money - loads of money - from computations (from scientific research to web hosting), not one at crazy clocks with four GPUs with the whole point of drawing fancy pixels (gaming).
I skipped the 1366/1156 generation so I can't speak from experience, but the gist I got was that the only reason 1366 was so popular was firstly because it came out before its less expensive brother 1156, and secondly because the way Intel binned the original 920s to meet demand it was very likely you would receive a much higher quality chip than what they were pricing it as. Had the i5-750/860 come out before the 920 I can imagine it being less successful.
ECC and server grade components regardless of the density of the RAM is no longer an issue like it was 5+ years ago. Ram failures are so ridiculously low that the need for ECC is just not justifiable in a home user platform no matter how enthusiast it is. Justifiable I said, but that doesn't mean you cant get it if you want it of course.
But we know the answer why Intel didnt offer it on X79. Its because they want you to buy their server hardware instead to get that stuff. AMD on the other hand for the large part of time offered ECC on their platforms. Thats why when I build a cost effective mid-range small business server I tend to go with AMD due to the added ECC support. I do not believe in running ECC on a computer that is going to be a video gaming platform, heavy multithreading home user, etc....
IF this guy we are referencing needed ECC that bad on socket 2011 just wait until Januarry when the server segments are released and get a Xeo and a Xeon board.
Or wait on EVGA SR3 and get it on there since its using the Patsburg server chip set I believe.
And lastly MY GOD there are some people that are really mad with socket 2011 so far. Why? Were you ever going to buy it anyways? And if so why you are expecting a delivery from God himself?
Actually X79 DOES support ECC, 2 of the launch motherboards list it in their spec sheets and users manuals. It's just the brain dead consumer level processors that have had that functionality lasered off or microcoded out. And that would make sense, from a market segmentation POV, IF Intel sold these significantly cheaper than Xeons. But, they don't do that, after the "newness tax" wore off, the 1366, 1156, and 1155 Xeons were all the same price or even cheaper than their equivalent consumer model. With a possible exception of the very highest binned CPU from each category.
And ECC is not just about protection from bit errors caused by cosmic rays, it provides a simple continuous and comprehensive check on system stability(motherboard, processor, RAM and PSU), that can only be equalled in a non-ECC system with a 1+ day long burn in test.
ECC should've become "standard" 10 years ago... I blame the bean counters. Though admittedly the need for ECC is somewhat hidden by how crappy most software was and still is.... Though it's getting better, or at least Windows has...
Gotcha. Yeah I personally as well as others like me do not need to have ECC however if it were a little less expensive than it is I would certainly go with 32GB of ECC 1600 if they made it but its at base speeds of 1333 so for now non ecc it is for my consumer grade desktop.
I was running an Openfiler appliance I built with 8GB of ECC memory and it ran for 7 months without a single performance degradation. I did not reboot or turn it off in 7 looong months. I love ECC its just not up to par with high performance overclocking style rigs yet. It could be though I wish it was.
I thought to some extent the CPU memory controller on sandy bridge and now E series has some amount of error control and correction in the controller. Not like ECC ram does but to some minor extent a form of ECC in the CPU.
ECC is not up to par with overclockable memory only because the same people who would actually purchase ECC that could run at 1600+ are not allowed to use it by Intel since it's not supported on the processors that make sense for this crowd. So there is no incentive by manufacturers to make it.
I daresay that overclockers who would spend en extra $5 per DIMM for overclockable RAM would not hesitate to add $2-4 extra per stick if it would also include ECC. The difference in price between Unregistered ECC standard RAM and regular non-ECC RAM is generally less than the differences between highly overclockable RAM and standard non-ECC RAM. Basically on a DIMM, all you're paying for is an extra memory chip, so the cost differences is not high (compared to adding RAM blingy LED lights and heatspreaders).
While RAM tests can ensure that ram is working at 100% in a system, most bit errors come out of the blue and are not caused by RAM functionality. The RAM could be 100% functional and still have an induced bit error (from radiation, power glitch, etc), and yet still be 100% for testing after a bit error has occurred.
Funny thing is that even with procs like the Xeon's that generally use ECC ram, you can still use non-ECC ram if you want in most non-server type systems. It would just make more sense for higher end "gamer" procs to allow this option as Intel probably actually has to spend more money "removing" the ability than just allowing it.
http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf
Data on RAM errors from Google.
For the average user out there, non-ECC should be fine as long as you are not doing something mission critical. But RAM errors happen and at a higher frequency than prior reporting.
Data on RAM errors from Google.
For the average user out there, non-ECC should be fine as long as you are not doing something mission critical. But RAM errors happen and at a higher frequency than prior reporting.
Interesting! Good read. So it seems that the Google study shows about an 8% of DIMMs are likely to have bit errors, but they can't break it down further between hard errors (hardware issues) versus soft errors (random could-be-caused-by-lots-of-things type). I would guess that Google does not overclock their machines so In the OC/performance community where RAM and associated circuitry is pushed even harder and in greater amounts than standard systems, the failure percentages can only go upwards. Seems to me to make the case for ECC support with processors geared towards the "enthusiast" crowd..
Afaict intel won't put ECC support on their "desktop" processors because they use it as the main distinguishing factor between "desktop" and "UP server" parts and they want to be able to charge more for "UP server" parts than "desktop" parts.
X79 is a server based chipset, repackaged for the "enthusiast"..... Go figure.
And I personally am not mad about socket 2011, I am hugely disappointed. And have already started to spend the money I set aside for 2011 on other things.
Given that Skyrim only uses 2 threads, I can't fathom how those numbers could possible be accurate...unless it is THAT desperate for memory bandwidth and/or on-die cache.
But...wow. If so...I don't even know what to make of that.
- Joined
- Mar 18, 2010
- Messages
- 14,454
Of course it's top. It's just not as top as some people expected it to be.
$600 for the low end processor... I think for $250 with 2500K and Z68 motherboard from Microcenter combo, I can hold off until Ivy comes out. Can't see a few games being worth all that extra cost.
Just right around the corner, new video cards and Ivy bridge. Next year is gonna hurt my wallet big time as I have two computers now I have to keep up with.
Just right around the corner, new video cards and Ivy bridge. Next year is gonna hurt my wallet big time as I have two computers now I have to keep up with.
Probably a case where the working data set fits in the SB-E cache, but not in regular SB. Same thing happens occasionally between 2500K and 2600K.
Im still seriously laughing at how everyone thinks IVY is going to rape every electronic invention know the Human race. This is inevitably going to give me the biggest LOL moment ever when they are like 2% faster than Sandy Bridge. Oh my gosh they will use like 20 watts less.
Yeah I have a PII 965 black and a I5-2500K and an I7-3930K - I cant testify that each has it's own place and purpose. I dont understand people. Each chip plays a different role. I am dumbfounded at the mentality of people on these forums. I am as equally happy with my AMD chip as I am with my 2500K. I am even more happy with my 3930K and I havent even had a chance to turn the bastard on yet.
- Joined
- Mar 18, 2010
- Messages
- 14,454
+1.
Once again everyone is hyped on something that is nothing more than a die shrink with a potential... and my God that is a far fetched potential, given intel's history, the GPU might be an inch better than the craptastic HD3000.
Anyways their gonna be 4 cores at 22nm. Whewhoo... a little power savings is about all I can see, maybe an inch more IPC. What would have excited more was if the standard issue socket 1155 mainstream IVY were to be 6 cores with an 8/10 core EP version. That would be so worth the hype.
Now as for you person who posted this. It will be a killer upgrade for you coming from C2Q for SURE!!!!
Node jump + 3D transistors + new architecture
What do you honestly think is going to happen? Or are you just trying to justify to yourself since you bought SB or SB-E?
It's pretty much going to be the same as going from the Q6600 to the Q9550 - and that was a pretty decent bump.
also ib-e will fit lga2011 sb-e, so if have the cash go ahead enjoy the sb-e right now ur lga 2011 is the enthusiast level pinout, replace 1366, ib-e will fit it as well,. its just your 4 core 1155 or 6 core 2011, so yea buying the lga 2011 is the socket for enthusiast
Last edited:
needmorecarnitine
Supreme [H]ardness
- Joined
- Feb 10, 2004
- Messages
- 4,609