You know I distinctly remember reading a review where they showed Haswell's L1 cache is almost twice as fast as Ivy's, but L2 and L3 cache were just about tied. Let me see if I can dig it up...
EDIT: ah ok here we go, it was actually from a review by Maximum PC on Ivy-Bridge E. No wonder I couldn't find it easily lol
Bonus: here's another chart from Tom's, and yet a third one from ExtremeTech,
-
HopelesslyFaithful Notebook Virtuoso
-
dude just ignore the words and read the charts if you're in a hurry lol, but in any case
Compared to Ivy Bridge quad cores, Haswell L1 cache = ~2x bandwidth; L2 cache 10-18% more bandwidth; L3 cache tied.
Haswell quad core L1 cache still 21% faster than IB hex core, HOWEVER IB hex core L2 cache is 36% faster and L3 cache 48% faster than Haswell quad core. -
HopelesslyFaithful Notebook Virtuoso
BTW was this the thread we were talking about CPU TDP and energy usage...was that topic over. I can't recall if that was finished. Also the OC part was what i was remembering was not finished. I think i forgot about it.
Someone posted 6 or 7 GHz for a chip but thats not common. There is a thread on a site that shows tens or 100s over OCers show their best scores that are actually stable and SB is 5-5.5GHz and HW was 4.5-4.7 GHz for stable 24/7 load speeds. -
yeah it's this thread, the discussion was whether we've made any real improvements besides efficiency at stock since SB can clock higher but IPC is not as good as Haswell, but Haswell can't OC as much as SB, so does it all just come out a wash in the end
I should probably add stock for stock Haswell is definitely superior, and probably what matters for Joe Average and possibly 90% of the computer users. For people that like to tinker and get the most out of their hardware though it's not a clear cut choice. -
HopelesslyFaithful Notebook Virtuoso
What's interesting are the bandwidth changes. I wonder how that affects performance
-
So does anyone have any idea when the new broadwell chips will be in the notebooks? When do they want to release it for notebooks?
I read many forums and discussions but i dont really understand if i should wait for it and buy my notebook with the new cpu. 1-2 months is totally fine but half year no. I want to get an msi gt72 or asus g751 with 980m but i ve heard that the haswell i7-4710 bottlenecks 980m. Hopefully the new broadwell chips wont... -
tilleroftheearth Wisdom listens quietly...
All notebook cpu's will bottleneck a 980m, even the new platforms/cpus.
If you have something now that is doing the job for you, I think that even 6 months is reasonable to wait for Broadwell (yeah, it will be that good). -
Qing Dao and HopelesslyFaithful like this.
-
tilleroftheearth Wisdom listens quietly...
Unless a notebook is running a desktop cpu and it is sufficiently cooled, then it is not BS - any notebook cpu in the constraints of a notebook chassis will throttle any highend component.
And where do you get that Broadwell will run hotter than Haswell? lol... -
If by bottleneck you mean "thermal throttling due to laptop not being to handle heat" then sure.
-
And unless the notebook design is bad, CPU should not significantly affect GPU in terms of cooling.
Broadwell is 14nm, durrr... -
Ok. So finally will we see a bigger increase in perfomance with broadwell? I didnt get finally when it comes out and if it worths the wait. For gaming, with 980m...
-
-
HopelesslyFaithful Notebook Virtuoso
the advantage of broadwell is that we hope to see higher clocks because it is at a lower process. Since it is the same CPU but a smaller process we hope to see more work dome in the same 45/47W TDP window. Right now you are seeing 3.5GHz for all cores maxed out. I personally am hoping to see 4 GHz on all 4 cores with the shrink.
If gaming is your only concern it isn't an issue UNLESS. You play a lot of single threaded games. Total War, Source, RCT 3, AOE3, and any other game that was made single thread. If you use a lot of peograms that are single thread that would be a reason to wait as well. But again if you do not have a large amount of stuff in this window then don't wait. -
An i7-4940X with GTX980m SLI and 660W dual adapter is way outside my price range at US$7000 but maybe you could get it down to US$4000 - 5000. For instance using the i7-4910MQ would save $600.
Base System and Display EUROCOM X7 Gaming Edition Choose your LCD Display 120Hz; 17.3-inch (43.9cm) FHD 1920x1080; MATTE (Non-Glare) Wide Viewing Angle; eDP; LP173WF2 Professional Display Calibration Service Optional - choose from the options below: Biometric Fingerprint Device Integrated Fingerprint Reader for Secure Data Access and Easy Management of Passwords Embedded Security Chip None, Optional Security Encryption TPM 1.2 Version Available Built-in Web Camera Two webcams: 5M (front cover) and 1M (backcover); FHD VGA Graphics Options - Choose the one that you need and/or want SLI; 16GB; 2x 8GB GDDR5; NVIDIA GTX 980M (N16E-GX); 1536 CUDA; GPU/VRAM Clock 1038MHz/2500MHz; Maxwell (28nm); MXM 3.0b; 2x 100W GPU / Graphics Technology Discrete GPU; Modular MXM 3.0b; Upgradeable; Dual VGA support (SLI and CF); up to 100W per card Processor (CPU) with High Quality Diamond 24 Thermal Paste 3.10 GHz (up to 4.00); Intel Mobile Core i7-4940MX Extreme; 4C/8T; 8MB L3; 22nm; FCPGA946; 57W; DDR3L-1600/1866 Memory (RAM) Configuration - Fully Upgradeable 32GB; 4x 8GB; DDR3L-1600; PC3-12800; 204pin; Micron, 1.35V - 4 SODIMMs Graphics eXpansion Modules for 2-3 External Displays None - Choose from the options below DisplayPort Adapters, Cables and Converters None - choose from the options below: USB 3.0 Built-in 4x USB 3.0 and 1x USB 2.0 Ports Internal mSATA SSD (1st) 120GB mSATA3 SSD; MLC; Micron/Crucial M500 SATA3 (6Gb/s); 20nm; Marvell 88SS9187 Internal mSATA SSD (2nd) 120GB mSATA3 SSD; MLC; Micron/Crucial M500 SATA3 (6Gb/s); 20nm; Marvell 88SS9187 1st Hard Drive. Choose Regular, Solid State or Hybrid Drive: 1TB (1000GB); 7200rpm; SATA3 (6Gb/s); 32MB cache; Hitachi Travelstar 7K1000; 9.5mm 2nd Hard Drive Choose your 2nd HDD/SSD/SHHD from the options below: 3rd Hard Drive 3rd HDD co-shared with Optical Drive; no internal ODD available if 3rd HDD selected; external ODD available Hard Drive(s) Setup i.e. Multiple Partitions, RAID 0/1/5/10 RAID 0 setup - for maximum performance Optical Drive 6x Blu-Ray WRITER; Tray; SATA; Panasonic UJ-260; BDXL 100/128GB media support; w/ software Choose Your Keyboard (Language) Backlit; English; USA/Canada; QWERTY External Optical Drive None - Choose from the options below Operating System(s) Microsoft Windows 7 Ultimate 64-bit; OEM; w/ Media DVD Disk Wireless LAN 2-in-1; 802.11 ac/a/b/g/n + Bluetooth 4.0;Intel Dual Band Wireless-AC 7260; 7260HMW TV Tuner (Internal and/or External) None - Choose from the options below Bluetooth for Wireless Devices Optional available via WLAN / Wireless Option Office and Business Productivity Software None - Choose from the options below Digital Content Creation; Multimedia & Graphics; CAD, CAE & CAM; 3D & Animation None - Choose from the options below Built-in Subwoofer Built-in Subwoofer Module Internal Card Reader 9-in-1 Card Reader; MMC/RSMMC; SD/MiniSD/SDHC/SDXC; MS/MS Pro/MS Duo Standard Battery (internal) 8-cells; Smart Li-Ion Battery Standard AC Adapter 660W (2x 330W) AC Adapter; Auto-switching ;100-240V w/Converter Box (for Maximum XTU/OC/SLI Performance) Standard Power Cord - included USA / Canada Carrying Case None - Choose from the options below Choose Your Warranty Plan 3 Years Return to Factory Depot with 3 Years of Tech Support Last edited by a moderator: May 12, 2015 -
HopelesslyFaithful Notebook Virtuoso
buy ram and hardrives on your own and wait until broadwell to buy XM/MX chip. Buy it on ebay for 300-500 bucks when broadwell hits the shore.
That'll save you 1 maybe 2 grand i bet -
tilleroftheearth Wisdom listens quietly...
No, by bottleneck I mean that any desktop cpu easily surpasses any true notebook cpu for performance. And that is what throttles any high end component like a top shelf gpu, for example. -
HopelesslyFaithful Notebook Virtuoso
-
tilleroftheearth Wisdom listens quietly...
The bottleneck exists on all notebooks, whether the user notices it or not.
-
HopelesslyFaithful Notebook Virtuoso
-
-
Last edited by a moderator: May 7, 2015
-
-
"Bottleneck" = a component is incapable of reaching its designed performance due to interference from another component directly related to it.
Example: the SATA rev. 3 SSD in my notebook's mSATA port cannot go faster than 300 MB/s because it is bottlenecked by the SATA rev. 2 speeds of the port.
"Throttling" = a component lowers performance by design as a result of external factors.
Example: my processor throttled when it went beyond its designed thermal capacity.
Neither of these terms accurately describe what occurs between notebook processors and graphics cards, because (in a well-designed system) there is no overheating (throttling), and these systems are not directly related (so no bottleneck). While it is true that desktop processors and graphics cards are more powerful than notebook processors and graphics cards, the difference between them continues to get smaller. In my opinion, there is no need to point out that a desktop processor or graphics card would be more powerful than its notebook version as this is obvious and not related to notebook discussion unless specifically discussing notebooks with desktop processors.Charles P. Jefferies likes this. -
Actually, there is very much a relationship between the CPU and GPU in the context of something like gaming if you think about what they do respectively. The CPU simulates the game, the GPU renders it on-screen. If the CPU can't do its job fast enough, the GPU has to wait for it, which lowers performance (frame rate).
For instance, I can fire up Battlefield 4 and limit my CPU to 1.2 GHz instead of its normal 3.2 GHz. My FPS tanks, and my GPU usage also drops because it has so little work to do while it's waiting on the CPU to finish telling it what to draw. That's called a CPU bottleneck, where the CPU is the component that's limiting the performance of the game.HopelesslyFaithful likes this. -
The only real corner cases are the ones where the game engine is fully multithreaded like Watchdogs where it may scale across 6 or 8 cores in which case the Desktop CPUs are unquestionably superior. The other corner case is the 4790K which has a monstrous default 4Ghz clock and 4.4ghz turboboost. Aside from those, the pedestrian desktop i7 and i5s are in the same ballpark as the mobile units. -
It's also easier to OC desktop chips given the better clock/voltage setting access, assuming you have the right mobo.
I would not be surprised if in the future, the performance of mobile and desktop chips start to converge due to Intel's focus on tablets and smartphones (aka, lower power consumption over performance). -
Jayayess1190 Waiting on Intel Cannonlake
-
So the L4 is even smaller now?
-
HopelesslyFaithful Notebook Virtuoso
wait so BW is only getting an extra 8 GPU cores over HW...thats dumb at least skylake will get a decent boost to keep the ~30% a year. Basically they are waiting to boost GPU until skylake :/
-
-
-
Skylake AVX512 would be tempting but having played with SDE it seems AVX512 might only be for the Skylake Xeons after all -
-
HopelesslyFaithful Notebook Virtuoso
1) Intel will loose a lot of customers due to:
Principle
AMD
AMD isn't very good but its better then a severely throttled intel or at least comparable.
2) FTC and DOF would flip a . I don't expect much out of government (i also loathe it overall) but with the history of US and monopolies that won't last long especially with people in the industry being the people smart enough to protest and file lawsuits. -
idk, perhaps just a way to make all those ES chips that find their way on to e-bay less appealing or the process needs a little more refining for the higher clocks.
-
HopelesslyFaithful Notebook Virtuoso
Also did it specify what type of chip it was?
-
Holy hell 95W TDP?!
-
HopelesslyFaithful Notebook Virtuoso
TDP is an irrelevant number without basis. My car may show it can go 120 MPH (really 105 then governor kicks in...i tried down my street) So skylake maybe able to pull 95W TDP on a decent voltage but thats not what stock will do. Same thing for every gen. -
2600/2700K was 95W, 3770K was 77W, 4770K 84W, 4790K 88W, and now there's a Skylake that will be 95W. So yeah I dunno.
-
HopelesslyFaithful Notebook Virtuoso
Remember those TDPs really are meaning less. Its like a governor or guide. It doesn't really mean much in of it self. A higher TDP would make me happy means less throttling on locked chips and potentially intel being serious about having more power CPUs but that's speculation of course. -
Why would you expect a CPU manufacturer to give you realistic clock speeds and TDP on early engineering samples? First, the process is certainly still being improved and second, they probably don't want to give their competition more information than they have to so they'd degrade the results intentionally if necessary.
-
HopelesslyFaithful Notebook Virtuoso
you do raise an interesting point but ES samples are for internal uses AND OEMs to test with. So depending on what version ES it is it should show market TDP. If its an early ES sample then its probably only internal uses and not likely to be accurate but if its late ES sample then its practically Retail.
-
That's true, but given that we don't even have retail Broadwell chips yet, I think it's a safe bet that these are early Skylake ES rather than late ones.
-
Charles P. Jefferies Lead Moderator Super Moderator
TDP isn't meaningless but is taken out of context.
TDP = Thermal Design Power. It's a number published by computer chip makers so that OEMs know what kind of heatsink to pair with it. A chip with a 35W TDP must be paired with a heatsink capable of dissipating 35W of thermal energy in order for the chip to operate within its operating specs. The meaning ends there. -
And that's why seeing 95W on a 14nm chip really scares me. Although both 2600K and 2700K were 95W chips, they were 32nm so the thermal density should be less than half, especially since Skylake is likely to pack more transistors than SB.
-
HopelesslyFaithful Notebook Virtuoso
seriously? Charles? its more so a governor then anything else. A chip can't pass a set TDP. A TDP is a "not random" but "loose number" that is set up for a Large range of CPUs. All CPUs get the same TDP in their class even though their power usage is all over the place. Its loose at best, maybe even arbitrary.
I doubt any OEM even pays attention to it anyways. Its basically an arbitry number when it comes to OEMs and it falls down to just being a governor.
Ivy Bridge (microarchitecture) - Wikipedia, the free encyclopedia
all quads have same TDP for their class. 77w down the list.
-
That was IB-E though, 3770K was still 77W.
Yes I understand what you're saying, but if the TDP was simply a governor, then it's likely based on the highest chip in class. I'm saying the Skylake equivalent of 4770K may end up running even hotter due to the increased TDP that's all. (and 4770K ran like a blast furnace already...)
Forget Intel Broadwell, Skylake On the Way
Discussion in 'Hardware Components and Aftermarket Upgrades' started by Jayayess1190, Jul 3, 2013.