Sandy Bridge-E Benchmarks Leaked: Disappointing Gaming Performance?

[John Doe]

Gulftown was 32nm vs the 45 of Nehalem... When Gulftown was received, many were criticizing Gulftown much like you are SB-E right now... Gulftown did not bring higher memory bandwidth or more PCI-E lanes... SB-E will, much like Nehalem did in the past.

What do you mean by that bolded part?

More complicated internal die so it could be harder to OC, which is what TH also says.

The Gulftown that was criticized was the 980x due to it's weak price/performance (hence the Extreme chip). The 970 was considerable after the first price cut and people bought it for it's lower temps, higher potential and so. In fact, if you can score one on the cheap from eBay (like me), it still is. I got a 4 core Westmere at 4.25 although it's crippled by 18x multi. Got the UD9 with binned IOH so I can do 220-240 BCLK.

That said, you can have those lanes with NF200, and memory bandwidth doesn't mean much either. For example, you can use dual channel on X58 with a %1-5 performance hit.

Sandy is the best choice for everyday users/for games. Sandy-E leaves to be... for those that want to show off or need the extra cores.

 

[n-ster]

More complicated internal die so it could be harder to OC, which is what TH also says.

The Gulftown that was criticized was the 980x due to it's weak price/performance (hence the Extreme chip). The 970 was considerable after the first price cut and people bought it for it's lower temps, higher potential and so. In fact, if you can score one on the cheap from eBay (like me), it still is. I got a 4 core Westmere at 4.25 although it's crippled by 18x multi. Got the UD9 with binned IOH so I can do 220-240 BCLK.

That said, you can have those lanes with NF200, and memory bandwidth doesn't mean much either. For example, you can use dual channel on X58 with a minimal performance hit.

Sandy is the best choice for everyday users/for games. Sandy-E leaves to be... for those that want to show off or need the extra cores.

I didn't go 970 as I think I would be wasting my money and I would rather get X79 than get stuck with X58 for a while.

One of the reasons I am going X79 is definitively the 8 DIMM slots.

The 2500K is obviously the king for games/everyday use here for sure. The only exception I can see to that is perhaps multi-GPU and upgradability and if games ever being fully multi-threaded to take advantage of the extra cores. For the short term, like most people in TPU think because they upgrade more often than the average, 2500K can't be beat, especially if you have a Microcenter near you

 

[John Doe]

One of the reasons I am going X79 is definitively the 8 DIMM slots.


The 2500K is obviously the king for games/everyday use here for sure. The only exception I can see to that is perhaps multi-GPU and upgradability and if games ever being fully multi-threaded to take advantage of the extra cores. For the short term, like most people in TPU think because they upgrade more often than the average, 2500K can't be beat, especially if you have a Microcenter near you

RAMdisk? 8 GB is more than sufficient right now so eh.

And yes, the second paragraph is what I've been trying to say from the beginning. Cheers.

 

[Wile E]

It doesn't. Point is, most apps don't scale over 6 cores, which is why Sandy outdoes Gulftown at 4 cores. So yes, per thread means a lot.

SB-E seems to have (from the given info) worse IPC than SB, and could have been worked on further to improve per-clock performance. Like Nehalem to Gulftown. But Intel didn't because they have no reason to. They're standing on top of AMD. Just add 2 more cores and leave it at that, right? Then jack up the price. There you have it, a $1000 CPU with higher TDP that, most the time, doesn't outperform a $350 chip.



Says the guy who were watching you when you were trying to talk with pipe in mouth. There was no need for that, please. I think you need to wake up a little before posting.

SB only beats Gultown in tests that do not take advantage of the available threading. All of the fully threaded tests prove my point nicely. Lower clock speed on the 990X and it lacking the new AVX instruction set and architectural improvements the SB has, and it still comes out on top. Now, take that same AVX and improvements, and up the clock speeds like SB-E is likely to do, and the 2600/2700k stands no chance.

And most apps not scaling over 4 cores is again irrelevant when talking about how powerful a cpu is. As long as there is ANY application that is capable of using the cpu to it's full potential, the additional potential power is fully relevant.

Gaming, in no way, leverages the full potential of these cpus, (or even Gultown, for that matter).

 

[n-ster]

RAMdisk? 8 GB is more than sufficient right now so eh.

And yes, the second paragraph is what I've been trying to say from the beginning. Cheers.

RAMDisk, cache, VMs, extreme multi task... I have 24GB and 8GB more on the way to use the total of 32GB on X79

But I don't think there is any disagreement here about the 2nd paragraph... I believe Wile E's point to be completely different and definitively something to consider.

 

[John Doe]

SB only beats Gultown in tests that do not take advantage of the available threading. All of the fully threaded tests prove my point nicely. Lower clock speed on the 990X and it lacking the new AVX instruction set and architectural improvements the SB has, and it still comes out on top. Now, take that same AVX and improvements, and up the clock speeds like SB-E is likely to do, and the 2600/2700k stands no chance.

And most apps not scaling over 4 cores is again irrelevant when talking about how powerful a cpu is. As long as there is ANY application that is capable of using the cpu to it's full potential, the additional potential power is fully relevant.

Gaming, in no way, leverages the full potential of these cpus, (or even Gultown, for that matter).

What I'm saying is, if you don't do much multi-threaded work (like most here), go Sandy. It's a better bang for buck and can be highly OC'ed on a cheap cooler. If you do (yes, you do), then go SB-E. But you're in a minority. Seems like majority or minority doesn't matter as long as it's your opinion. Because your perspective is the best of all.

 

[n-ster]

What I'm saying is, if you don't do much multi-threaded work (like most here), go Sandy. It's a better bang for buck and can be highly OC'ed on a cheap cooler. If you do (yes, you do), then go SB-E. But you're in a minority. Seems like majority or minority doesn't matter as long as it's your opinion. Because your perspective is the best of all.

I don't think that is his argument... He's arguing the proper way of assessing a CPU's performance. I think he'll agree that for most gamer/normal use SB is better than SB-E

 

[Wile E]

What I'm saying is, if you don't do much multi-threaded work (like most here), go Sandy. It's a better bang for buck and can be highly OC'ed on a cheap cooler. If you do (yes, you do), then go SB-E. But you're in a minority. Seems like majority or minority doesn't matter as long as it's your opinion. Because your perspective is the best of all.

I never said SB-E was the choice for everybody. I said games are a terrible cpu test. I don't see how you are confusing this. Any test that does not fully stress a cpu is not a good cpu test.

 

[John Doe]

I never said SB-E was the choice for everybody. I said games are a terrible cpu test. I don't see how you are confusing this. Any test that does not fully stress a cpu is not a good cpu test.

A test doesn't have to load a CPU to %100. It comes down on what you're testing the CPU for. If for playing just older, single threaded games, you can look at tests with one core then make your decision from there. A %100 loaded CPU wouldn't have any power left, yes. But that's not what I'm saying. It's about how effective a CPU's threads are. Not how many threads is has. An SR-2 gets beaten by Sandy. Why? Because most apps don't make use of the second chip (Windows just throw loads across), it has lower build quality with coil-chokes, and struggles to do beyond 185-200 BCLK.

 

[Wile E]

SR-2 does not get beat when used with apps that use all of it's available threads, or under heavy multitasking. Does that mean you have to buy it to game? No. But that still doesn't validate games as a good indicator of a cpu's performance.

 

[John Doe]

SR-2 does not get beat when used with apps that use all of it's available threads, or under heavy multitasking. Does that mean you have to buy it to game? No. But that still doesn't validate games as a good indicator of a cpu's performance.

No, it depends on your goal. What you're testing/need the CPU for. But yeah, this isn't going anywhere. It's only your opinion that matters. Fact or fiction or in between it doesn't matter as long as it's your opinion.

 

[n-ster]

No, it depends on your goal. What you're testing/need the CPU for. But yeah, this isn't going anywhere. It's only your opinion that matters. Fact or fiction or in between it doesn't matter as long as it's your opinion.

To have a representative idea of a CPU, you need something that stresses it to 100%, to think otherwise is insane. Just because most people will browse the internet with their PCs, does not mean people should benchmark that. Technically, to have a full assessment of a CPU, you have to have many tests, including gaming and application, but that's not the point. To have a a test that doesn't stress the CPU completely doesn't show you it's real performance.

If you are a gamer, then yes, you look mostly at gaming benchmarks, but to say a gaming benchmark (that doesn't take full advantage of the CPU) is a good way to evaluate what CPU is best, then no, that is factually not true.

 

[Wile E]

No, it depends on your goal. What you're testing/need the CPU for. But yeah, this isn't going anywhere. It's only your opinion that matters. Fact or fiction or in between it doesn't matter as long as it's your opinion.

Pot calling the kettle black.

 

[ramcoza]

People, who say we can stress a CPU when play a game at a lower res, should take a look at this forum post and what is the recommendation he got after all.

http://www.tomshardware.com/forum/332479-33-performance

 

[jewie27]

LOL notice Power Consumption: MORE IS BETTER? FAKE!

 

[n-ster]

People, who say we can stress a CPU when play a game at a lower res, should take a look at this forum post and what is the recommendation he got after all.

http://www.tomshardware.com/forum/332479-33-performance

A lower res stresses the CPU less but eliminates in a certain sense the possibility of a GPU bottleneck

I'll take this with a grain of salt... Besides, we know many games are GPU intensive

btw, Power consumption, more is better... yup very trustworthy lol

LOL notice Power Consumption: MORE IS BETTER? FAKE!

You've been beat since a looooooong time... Over a 110 replies ago

 

[claylomax]

First one I found: http://www.hardocp.com/article/2011/11/14/intel_core_i73960x_sandy_bridge_e_processor_review

 

[cadaveca]

Crap review is crap review. No stock numbers, all CPUs benched at "max" OC = FAIL review.


Of course, if your sample of CPU doesn't clock as well, that review is useless. I really did expect better out of Kyle. But, as I said earlier, I didn't expect much out of reviews.

Sucks being right.

Kyle's conclusion:

I am not sure who is supposed to buy a 3960X. I really do not see it benefiting gamers. I do not see it being a boon too overclocking enthusiasts due to price, power usage, and subsequently heat output. I guess if I sat around all day ripping Blu-ray disks and encoding those for torrent sites, it would be awesome. Maybe that could be Intel's new 3960X motto, "Sandy Bridge E, maximizing BitTorrent ratios, one desktop at a time." Meh. Let's see what the K series brings before we totally turn our noses up at this beast of a processor...that none of us really need, or I think even want. I think we have enough cores for now. Get your noses back on the grindstone and give us stellar IPC gains or even better, 5GHz stock clocks.

So, because he can't find a use for it, the CPU is fail. HMMM...Sounds like Bulldozer.

 

[n-ster]

Crap review is crap review. No stock numbers, all CPUs benched at "max" OC = FAIL review.


Of course, if your sample of CPU doesn't clock as well, that review is useless. I really did expect better out of Kyle. But, as I said earlier, I didn't expect much out of reviews.

Sucks being right.

Kyle's conclusion:



So, because he can't find a use for it, the CPU is fail. HMMM...Sounds like Bulldozer.

Damn I guess Gulftown sucked ass

 

[claylomax]

I love this part: "But I guess if you are looking for a new way to heat your computer room this winter, Sandy Bridge E should be on your short list. Maybe they will start selling these at Home Depot?" This platform with my GPU set-up and I would save some on gas bills this winter.

 

[n-ster]

My i7 920 beats it in power consumption.... awesome

 

[cadaveca]

Damn I guess Gulftown sucked ass

At least you get where I'm coming from here. Kudos.



Oh, and...


Cache, cache, cache, cache.


That is all.

 

[entropy13]

http://www.legitreviews.com/article/1773/1/

 

[claylomax]

http://www.legitreviews.com/article/1773/1/

This is a better review than HARDOCP's, but doesn't change the picture.

 

[entropy13]

http://techreport.com/articles.x/21987
http://www.tweaktown.com/reviews/44...xtreme_edition_lga_2011_cpu_review/index.html
http://hothardware.com/Reviews/Intel-Core-i73960X-Extreme-Edition-Sandy-BridgeE-Review/
http://hexus.net/tech/reviews/cpu/32591-intel-core-i7-3960x-extreme-edition-cpu/