Torn between 2011 and 1055

[izusaga]

I'm looking for a motherboard recommendation, primarily. I need reliable SATA-III which will allow me to RAID0 120GB OCZ Vertex 3's for maximum performance from a read/write perspective, and allow me to SLI 2 580GTX's at 16x/16x.

Right now I'd really like to purchase a i7-3930K - is there any specific reason I should not?


Specs on the machine I'm upgrading from (which will turn into my "secondary" machine for quad-boxing games):

i7-970 @ 4.23ghz
12GB 2000mhz Mushkin (3x4GB)
180GB Vertex 2 Raid0
eVGA x58 Classified


I would like to know if the i7-3930k would yield significant results in high-end maxed out gaming situations, or if I'm better off sitting tight with my 970 setup and waiting for Intel's next processor lineup.

Thanks in advance!

 

[Tsumi]

1. You don't need 16x/16x SLI. 8x/8x has no significant performance loss with 580s.

2. Your 970 is more than enough. You don't need to upgrade it with just two 580s.

 

[Trackr]

1. You don't need 16x/16x SLI. 8x/8x has no significant performance loss with 580s.

2. Your 970 is more than enough. You don't need to upgrade it with just two 580s.

Agreed on both.

Also, I'd figure it stupid to buy a 32nm CPU two months before the 22nm CPUs are to be released.

 

[izusaga]

1. You don't need 16x/16x SLI. 8x/8x has no significant performance loss with 580s.

2. Your 970 is more than enough. You don't need to upgrade it with just two 580s.

Re: 2) When would I need to upgrade? IE when will I be CPU bottle necked with the 970?

Thanks.

 

[Trackr]

Re: 2) When would I need to upgrade? IE when will I be CPU bottle necked with the 970?

Thanks.

You're always CPU bottlenecked.

My 2600k @ 5.1Ghz will squeeze out 51FPS in Crysis @ 2560x1600, where as yours will only get 46-47FPS.

When will a game be unplayable? Not in the next 3 years.

 

[izusaga]

Alright. Allow me to pose a logistical question.

I multi-box games. Right now, 5-box Rift, and in the future, I will be 5-boxing TERA.

Would the upgrade to the 3930k and quad channel memory dramatically increase my ability to smoothly do this?

 

[Zomoa]

Alright. Allow me to pose a logistical question.

I multi-box games. Right now, 5-box Rift, and in the future, I will be 5-boxing TERA.

Would the upgrade to the 3930k and quad channel memory dramatically increase my ability to smoothly do this?

Dramatic improvement? Nah. Noticeable? Probably. Worth your money? No. If you really want to multibox on a single system, build up an sr-2 box, or wait for sr-3. It will be a better use of your money. I run into a problem with some cores running to 100% usage and others chilling at 30-60% when 6-boxing.

Really the only way to noticably improve performance and reduce the potential of hard lag spikes would be to give each client more cores (without sacrificing IPC). Upgrading to a 3930k wont give you that - merely a small boost in IPC that will be noticable in some cases, but wont solve the core-starved nature of multiboxing.

 

[izusaga]

Dramatic improvement? Nah. Noticeable? Probably. Worth your money? No. If you really want to multibox on a single system, build up an sr-2 box, or wait for sr-3. It will be a better use of your money. I run into a problem with some cores running to 100% usage and others chilling at 30-60% when 6-boxing.

Really the only way to noticably improve performance and reduce the potential of hard lag spikes would be to give each client more cores (without sacrificing IPC). Upgrading to a 3930k wont give you that - merely a small boost in IPC that will be noticable in some cases, but wont solve the core-starved nature of multiboxing.

I utilize innerspace, so I can specifically assign cores and GPU's to whatever instance of the game I want, allowing me to run the primary with the most performance and the rest with lowest settings nearly flawlessly. There is still a lot of CPU bottlenecking going on, however.

 

[Redbeard80]

Just to throw out the raid question... I did 2x Crucial M4 SSD's raid 0 on a P67 board and hit 1GB/s reads... I am sure either platform will serve you in that dept (as long as it is intel)

 

[izusaga]

Just to throw out the raid question... I did 2x Crucial M4 SSD's raid 0 on a P67 board and hit 1GB/s reads... I am sure either platform will serve you in that dept (as long as it is intel)

Good to know. Part of my current issue is I have about 12 Vertex 3's at my disposal but currently I'm running on x58 chipsets only. I picked up a RocketRAID cart to try and squeeze some SATA-3 bandwidth and just ended up getting 750mb/s read and 500mb/s write, at the loss of 16x graphics channel. Wasn't worth it. Getting the actual Vertex 3 speeds is definitely worth it to me.

I've been putting some serious thought into an eVGA SR-2 and running dual quad-core Xeons with a few 12GB RAM kits. Would possibly let me run all of my characters on one computer instead of the current 2 system setup.

Bonus - it's got 2 SATA3 ports..

 

[Dan_D]

Just to throw out the raid question... I did 2x Crucial M4 SSD's raid 0 on a P67 board and hit 1GB/s reads... I am sure either platform will serve you in that dept (as long as it is intel)

The Z68 and X79 chipsets are identical in terms of storage. Both use Intel's C600 SATA controller.

 

[RanceJustice]

I've had a great experience with my "old" 1366 kit. If I was running a 970 or 980 instead of a 920 I wouldn't even have been tempted by what has come off of 2011 thus far with its outrageous pricing. However, I do believe that 2011 will be the best socket for enthusiasts in the future, possibly at the arrival of Ivy Bridge if Intel gets their head removed from their corporate, collective ass and prices them reasonably.

I didn't like watch 1155's fumbling around, the issues withe P67 and even Z68 chipsets (requiring GEN2, GEN3 etc...) and generally how swiftly for the performance the mainstream chipset kit is outmoded. My current kit has basically kept me near the top of the pack since the launch of the i7-920, in terms of performance and it wasn't until SB-E's launch that I really felt I'd get a significant performance increase...but mind you on their $600-800 hex-core part! Even now, I still feel the increase isn't worth the expense of the insanely expensive SB-E platform.

WIth a kit like yours, I'm guessing this isn't something where you HAVE to buy now and you're lamenting the loss of power. You're still matching or wiping the floor with everything that isn't a SB-E hex + high end X79 + shitloads O quad-channel RAM etc... setup. You have a high end later generation 1366 motherboard, enabling SATA3 etc... you'll be fine. If you don't have to buy this minute and are interested in saving any money at all, I highly suggest waiting until Ivy Bridge-E 2011 and buying then.

If you HAD to buy right this minute, regardless of value, I'd still say 2011. Its a more powerful and featured platform that is going to only be increasing in power with IB-E's arrival. Mainstream is on its way out...yet again, just like those who bought Socket 1156.

 

[izusaga]

I'm running a 970 @ 4.4ghz on air side by side with a 950 @ 4ghz on air. I've honestly found that the quad-core processor tends to handle multiple instances of a specific game better than the hex-core. Any specific reason why I see those results?

Additionally, if I went hex-core 2011 instead of going dual hex-core (or dual quad core) 1366 on a SR-2, how significant would the gains be from the memory bandwidth quad channel vs. tri channel compare to the gains of doubling the processing power? Which would accomplish my multiple gaming needs better?

 

[Tsumi]

Quad-channel memory performance on 2011 is actually slower than dual-channel on 1155 from what I heard. The only benefit quad-channel gives you over tri-channel I think is the larger memory capacity, that's it. I'm not sure if the memory controller on 2011 is faster than 1366.

I don't see why the 970 would run worse than the 950. Is there possibly some resource sharing issues going on? Try assigning each instance of the game to their own cores.

You need more raw cpu performance than memory performance. The memory controllers on Intel CPUs are quite fast.

 

[izusaga]

So if I am looking for maximum processor performance, ideally I should wait for the SR-Z to be released and throw two hex core monsters in it?

 

[Dan_D]

I'm running a 970 @ 4.4ghz on air side by side with a 950 @ 4ghz on air. I've honestly found that the quad-core processor tends to handle multiple instances of a specific game better than the hex-core. Any specific reason why I see those results?

Additionally, if I went hex-core 2011 instead of going dual hex-core (or dual quad core) 1366 on a SR-2, how significant would the gains be from the memory bandwidth quad channel vs. tri channel compare to the gains of doubling the processing power? Which would accomplish my multiple gaming needs better?

For anyone thinking about LGA2011, consider this. Let's put aside cost / performance arguements for now. Performance wise, LGA2011 and LGA1155 perform about the same. More memory bandwidth really only helps in very specific cases most of which the home user and gamer won't see. X58 based systems have more memory bandwidth than P67/Z68 based systems do, yet aside from rare cases where the extra cores offered by the Core i7 980X/990X make a difference, the newer P67/Z68 platform ends up being faster (and sometimes significantly faster) most of the time. X79 based systems have more bandwidth than either, but this hasn't so far really translated into vastly improved performance over P67/Z68 in most cases. Again situations where the bandwidth matters are few and far between and will only be encountered by the professional workstation crowd or the hobbyist who runs one or two applications on occasion which benefits from the added bandwidth. I encode movies for example, and sometimes do it alot. But it isn't the primary use for my machine. The time I spend doing general office tasks on it and gaming far outstrip the time I spend encoding movies. So while I can take advantage of the extra CPU cores and to a lesser extent some memory bandwidth, the time I spend doing these things wouldn't justify the cost for MOST people.

Also consider that the extra PCIe lanes offered by X79 will usually go unused by most. Even if they didn't, almost no difference exists between SLI or Crossfire setupsusing an 8x8 vs. 16x16 lane PCIe lane configuration. Aside from that X79 from a feature standpoint only has PCI-Express 3.0 support over Z68 and only a handful of devices currently exist which support it. And even then no performance benefit has really been shown to exist for it. Video cards have still not saturated PCI-Express x16 2.0 slots yet. PCIe Gen3 is overkill for the time being. It will probably continue to be for some time and even if it isn't, Ivy Bridge is just around the corner for LGA1155/P67/Z68. So it isn't as if LGA1155 won't have that in the future.

LGA2011 has advantages. Don't kid yourself, but those advantages will not be utilized by most people on the forums here. You'd really need a lot of expansion cards and very specific usage scenarios to leverage it. X79 is almost a workstation chipset masquerading as a high end desktop part. There is some overlap between the high end desktop market and the workstation market, which X79 can handle on both sides of the fence. Still most won't be able to leverage the extra CPU cores. extra memory capacity, bandwidthmuch or even the added PCIe lanes the platform offers.

Adding cost or price vs. performance concepts into the mix makes the situation even clearer. X79 chipset motherboards are $50 or more expensive than their Z68 counterparts. As far as the chipsets other features, USB ports, SATA ports, etc. are identical on both. They use the same drive controller. All Z68 has that X79 doesn't is support for integrated GPUs. If you need or rather want that then X79 most likely isn't for you anyway. You end up almost losing the ability to use air coolers with X79, without at least loosing DIMM slots. Which defeats one of the biggest advantages LGA2011 boards have over LGA1155 boards. LGA2011 CPUs use more power and run hotter than LGA1155 parts do. They tend not to overclock as well either. So far in my experience I can get between 100-300MHz more out of LGA1155, and sometimes more compared to LGA2011 setups. The 3930K is what? $600+ and the Core i5 2500K is only $179.99 at Microcenter or $209.99 at Newegg. The Core i7 2600K costs around $300 if I recall correctly. Let's not forget that you'll need twice as many memory modules for LGA2011 to take advantage of quad-channel RAM. That doubles the cost of RAM. So we go for what, $100 to $200? So at a minimum, all said and done you are talking about a cost difference of at least $400+ on the conservative side between LGA1155 and LGA2011. LGA1155 is 95% of what you get with LGA2011 at about half the cost.

Seems simple to me. Unless you absolutely need what LGA2011 offers, I wouldn't bother with it over LGA1155. It just doesn't make sense.

 

[Tsumi]

Ideally, you would wait for IB-E and throw two 8-core monsters in there, for major epeen rights

 

[Dan_D]

So if I am looking for maximum processor performance, ideally I should wait for the SR-Z to be released and throw two hex core monsters in it?

ASUS also has a board coming out with dual LGA2011 sockets designed for the performance / workstation enthusiast. But yes, if money isn't really an object and maximum performance is desired, then yeah, something like that is where it's at.

 

[izusaga]

I appreciate the detailed replies (especially from Dan_D). This entire project has me thinking that for the purposes I envision (in this a case semi-automated innerspace/isboxer based gaming rig running 10 copies of a game) it will be wise for me to wait for the SR-Z or it's ASUS counterpart to be released and then drop in a pair of 6 or 8 core Xeon's into it and overclocking them to the 4.5-5ghz range for optimal performance, and then pairing it with 2-3 580GTX's (or whichever 590-ish variant is the best buy at the time, the VRAM is probably most important here anyway for what I'm doing). In this case, money is no object so I have no problem dropping 3-4 grand on the motherboard-processor alone. This will be for a secondary machine that will act as a slave to an IB-E primary gaming rig regardless, so it's purpose is to have HEAVY CPU power so I can specify game instances to specific cores.

I really feel I was putting too much weight on memory, and I appreciate the education on the multiple variants of RAM channels being used - it was the biggest gap in my understanding holding me back from a decision. Looks like I'm safer off waiting a little bit and doing the "go big or go home" route.