Saturday, November 12th 2011
Sandy Bridge-E Benchmarks Leaked: Disappointing Gaming Performance?
Just a handful of days ahead of Sandy Bridge-E's launch, a Chinese tech website, www.inpai.com.cn (Google translation) has done what Chinese tech websites do best and that's leak benchmarks and slides, Intel's NDA be damned. They pit the current i7-2600K quad core CPU against the upcoming i7-3960X hexa core CPU and compare them in several ways. The take home message appears to be that gaming performance on BF3 & Crysis 2 is identical, while the i7-3960X uses considerably more power, as one might expect from an extra two cores. The only advantage appears to come from the x264 & Cinebench tests. If these benchmarks prove accurate, then gamers might as well stick with the current generation Sandy Bridge CPUs, especially as they will drop in price, before being end of life'd. While this is all rather disappointing, it's best to take leaked benchmarks like this with a (big) grain of salt and wait for the usual gang of reputable websites to publish their reviews on launch day, November 14th. Softpedia reckons that these results are the real deal, however. There's more benchmarks and pictures after the jump.
Source:
wccftech.com
171 Comments on Sandy Bridge-E Benchmarks Leaked: Disappointing Gaming Performance?
The fact the lwoer-binned chips have cache disabled should speak VOLUMES as to what the story is with power consumption, nevermind the seemingly more efficient utilization of memory bandwidth, no matter how slight.
At least with the EE, you get all the cache enabled. It will eb interesting to see how the lack fo cache pans out performance-wise for the "K" 6-core chip, and the "locked" quad.
Tech Report predicts that the 3930K would be around the "point" of the i7 970 (if following along the x-axis) but be way higher up (if following along the y-axis), making it a more likely buy than the new EE.
I didn't expect this. :eek:
www.anandtech.com/show/5091/intel-core-i7-3960x-sandy-bridge-e-review-keeping-the-high-end-alive/7
hardocp.com/article/2011/11/14/intel_core_i73960x_sandy_bridge_e_processor_review/1
www.legitreviews.com/article/1773/1/
www.tweaktown.com/reviews/4414/intel_core_i7_3960x_extreme_edition_lga_2011_cpu_review/index1.html
techreport.com/articles.x/21987
hothardware.com/Reviews/Intel-Core-i73960X-Extreme-Edition-Sandy-BridgeE-Review/?page=1
hexus.net/tech/reviews/cpu/32591-intel-core-i7-3960x-extreme-edition-cpu/
www.tomshardware.com/reviews/core-i7-3960x-x79-sandy-bridge-e,3071.html
www.anandtech.com/show/5091/intel-core-i7-3960x-sandy-bridge-e-review-keeping-the-high-end-alive
pcper.com/reviews/Processors/Intel-Sandy-Bridge-E-Review-Core-i7-3960X-and-X79-Chipset-Tested
www.overclockersclub.com/reviews/intel_core_i7_3960/
www.guru3d.com/article/core-i7-3960x-processor--msi-x79agd65-review/
I do not think that's of any importance, other than that it should be noted that retail sample clocking is still in question at this point.
That said, many ES chips feature overly high power consumption, and aren't stellar overclockers, as they are intended to test functionality of the platform, including cooling.
I really wish AMD and Intel would stop giving out ES chips to reviewers, and start giving out the same chips end users will get. I suppose that's just me though.
www.hardwareheaven.com/reviews/1316/pg1/intel-sandy-bridge-extreme-and-x79-chipset-launch-core-i7-3960x-processor-review-introduction.html
Looked at a handful of reviews and yeah, gaming performance is no better than a 2600K, in some cases just a teeny bit worse, so it's disappointing alright. I had suspected that the Chinese website I reported on in the news article was right. :ohwell:
I'd have expected a new revision to show some improvement, surely? Basically, it's working like plain vanilla SB with a couple of extra cores. It does overclock better, though. I'm not commenting on other types of app however, because gaming performance is the only thing I'm personally interested in.
I'll be buying that 2700K system Real Soon Now.
HardOCP had an IPC comparison here, which shows that SB & SB-E are the same.
This is what happens when Intel doesn't have any head-to-head competition. :shadedshu
I'm sure I'll get flamed for this post by ye fanbois out there, so I've put on my strongest flameproof jacket. :p
EDIT: Here's more IPC comparisons by Hardware Canucks.
It's a multi-threading chewer beast, and that's what I wanted it to be... Honestly I'm not going to buy this for gaming, that would be crazy!
I'm buying this because that's what I need for my work :) and believe me, it's not a disappointment :D
This is what happens when there's no competition in the marketplace. And believe me, Bulldozer and its derivatives are zero competition to Intel right now and for the forseable future.
Intel are just sitting on there laurels and laughing all the way to the bank.
To be honest I have a little disappointment...
I would have loved the 3960X to be an 8 core(16t), without the 2 fused off cores... that would have justified the price tag over the 3930K!
Also I think you are not being reasonable with the timeframes, "resting on their laurels"? Of course they are relaxed, but to say as much as that they are sitting on their laurels... it's just not true since they started with the tick-tock. Have you been paying attention these last years (decade I'd say)? The days of 1 year cycle for each new architecture were over more than 10 years ago, now it's typically 2-3 years if things go perfectly and you are Intel, or just simply ask AMD, Via or hell even the quite successful ARM how easy it is to create a new arch every <insert random number> months. SB was introduced 10 months ago, expecting another new architecture this soon is delusional. 5 years and look what happened with Bulldozer, it's not easy.
I agree that Intel could advance faster if they took some capital risks, like the rest are doing, but why should they take any risk? Why abandon the tick-tock strategy that is working so well fr them? They are already 50%+ faster than competition and the difference grows with every tick-tock cycle. AMD just can't keep up at this pace, no one can really, so Intel is not objectively sitting on their laurels.
I understand your desire to get better and better CPUs (in all fronts, i.e. gaming) with every release, but I think it's not very realistic to expect huge improvements on every chip, at least when they belong to the same architecture. Like many people have said already, SB-E delivers where it was designed to deliver: in hevily threaded apps.
Sorry for the long post and sorry if it seems I'm picking at you, not at all my intention.
However, then Bene "picks" on me, lol, with this terribly well reasoned post and he's actually perfectly right. :) I was expecting the wrong thing at the wrong time - but at least I was in good company. :pYes, as you're quite right about the release cycle, as SB was released less than a year ago. It's still true though that if Bulldozer had been any competition, then we would have seen some single threaded improvement and that's what I mean by resting on their laurels. But you're right, it's too small a time frame to see any architectural improvement other than small refinements, so I take the resting on laurels bit back. :)
By the looks of it, they've basically improved memory bandwidth to handle the extra cores. So, are we going to actually see an 8 core SB-E CPU next year then?Yes you would see that scaling, but as I've explained above, I'd hoped single threaded performance would have been improved, which would speed up everything. Obviously, my expectations were unrealistic though. :ohwell:
What does AMD need to do to better compete? FIX Branch Prediction, Pipeline Flushing, Cache Trashing, increase the Decode unit's width, resolve this scheduling issue? Can a B3 stepping fix this? AMD has a chance to compete now.
:laugh:
I don't care to speculate on Bulldozer.
Frankly, the only disappointment for me is the lack of a 8-core/16 thread monster.
Hey ....did I miss the overclocking? 600 for a six core.....intels learning. Slowly.....but learning
Maybe it's kind of better this way. While it would have been interesting to see a direct comparison between both 2 billion transistor behemoths (SB-E vs BD), it would have been a bloodbath.