i7-4980HQ vs i7-5950HQ

i7-4980HQ vs i7-5950HQ

I'm about to buy the MSI GT80 Titan SLI and there are quite a few models out there, one with the i7-4980HQ and the other with the i7-5950HQ they have both the same SLI GTX 980m graphics card but which of the CPU should I choose? They have the same price as well but I wonder really what would be the difference. I did make a research but I want to know more.

 

http://www.futuremark.com/hardware/cpu/Intel+Core+i7-4980HQ+Processor/review

There aren't many scores available for the 5950HQ, but reviews say it performs like a 4940MX. If that is the case, I would just go with the 4980HQ. It actually has a higher FutureMark score than the 4940MX. However, you may want to take a look at Haswell vs. Broadwell reviews.

 

I've been slightly out of touch but from what I've read, Broadwell seems to hold it's clocks for much longer then the Haswell CPU's, at least for the ULV's... That would tempt me to go for the 5750HQ however the 5950HQ is slightly lower clocked...I think with all the improvements with Broadwell, they'll probably end up with similar scores... I would personally go for the newer 5950HQ however you can't go wrong with either...

 

The 5950HQ has a higher base frequency, neither CPU is going to hold their max 4 core turbo at 47W but the 5950HQ should perform at a higher point at the same TDP.

 

Definitely 5950HQ. Will perform better despite TDP regulation. 5700HQ proved to be faster all around compared to 4720HQ, despite being largely the same CPU.

 

Care to explain why? I know the basics, but for some reason I don't quite understand why it'll perform better ..

 

Like I said, there weren't any scores to compare, so I can't exactly suggest the 5950HQ with certainty that it'll be a better CPU. It doesn't really matter at this point. We're talking 5% difference, maximum. It won't impact GPU performance or anything.

Isn't 59 50 lower in the food chain than 49 80? There may be a 59 80? Naming schemes really annoy me.

NotebookCheck has the 5950HQ rated lower than the 4980HQ.

 

You can't compare models between generations.

Higher performance per watt means under the same power restriction you get more performance.

 

Just as Meaker said, since the new architecture is more efficient, it runs at the same speed while consuming less power. So, all current Core i7 mobile CPUs from the HQ family are power controlled by their designed TDP of 47W. Since all HQ processors consume more than 47W under heavy load (I can reach near 60W on mine at 3.6ghz) you get a turbo boost window for a minute or less. After that turbo window, you drop down to 3.4ghz or so, depending on the load, to keep it at 47W or under it.

Benchmarks have shown that compared to Haswell, Broadwell scores a bit higher all around in CPU benchmarks, including 3dmark, cinebench etc. This means it can maintain higher clocks at the same TDP, or at least maintain them for a longer period of time.

Also, GT80 titan broadwell consistently scores higher physics scores in general, and if I remember correctly, it is the 5950HQ GT80 that currently has the highest firestrike score among GT80 users, at 17K.

 

ryzeki, I just really want to ask you since you have the GT80 Titan with the 4720HQ and I guess it is a bit cheaper than the GT80's with the 4980HQ and 5950HQ, would it be sufficient to run ultra modded games such as Skyrim with ENB's installed at 60 FPS? I just want to know if CPU really affects the gameplay for most games because what I read if the truth CPU power will be very useful for video editing and strategy games and I hardly play those type of games. So if a 4720HQ with a GTX 980m SLI is more than sufficient then I wouldn't really mind about the CPU and I could save more money for upgrades down the road.

 

Personally, anything that pushes the CPU will run very similarly regardless of the CPU, because of what I mentioned of the TDP control. There are games right now, like Final Fantasy 14, that I push beyond 120 fps on the benchmark at max settings.

I run witcher 3 maxed out without vsync and no AA in hairworks (but hairworks enabled), at 60fps. I think a 4720HQ is sufficient for games. The problem I see with higher end CPUs is that if you do push the CPU, it will always downclock itself to regulate at 47watts TDP, so you will have very similar performance anyways. 5950HQ excels at short bursts of extra CPU power, and a higher integrated GPU (which is nothing compared to SLI 980m).

I think 4720hp should be enough for 60fps to a long time. If you do find the GT80 with 5700HQ, I would go for that one myself.

Unless we find a way to overcome the TDP regulation, I think 5950HQ are wasted potential.

 

The GT80 - 263 with the 5950HQ is running great at stock settings, and here are some OC vs. stock results for comparison.

Used XTU to set CPU at 43x/42x/42x/42x core's, ran XTU Benchmark, and got 1131
http://hwbot.org/submission/2942474_hmscott_xtu_core_i7_5950hq_1131_marks/



At the stock CPU settings the XTU Benchmark result is 1013.

It's not stable at this high of a setting without adding voltage, +100mV.

There isn't a lot of stability above the stock settings of 38x/37x/37x/37x core's and 34x cache without adding voltage.

It is a bit early, I haven't settled on high XTU settings to share, but these seemed interesting enough, along with these comments and benchmarks to post.

It would be nice to get the 44x/43x/43x/43x core's and 40x cache settings available to be stable, but so far the GT80 won't allow it.

Here are the passmark results at the same XTU settings, 43x/42x/42x/42x core's 34x cache +100mV.

http://www.passmark.com/baselines/V8/display.php?id=45394510801
Passmark Rating 5,847
CPU Mark 12,321
2D Graphics Mark 882
3D Graphics Mark 7,842
Memory Mark 2,894
Disk Mark 12,865

Here are the passmark results with CPU stock settings, 38x/37x/37x/37x core's 34x cache +0mV. passmark doesn't let you submit if the previous submitted passmark score is within 5%.

Passmark Rating 5,406
CPU Mark 10,937
2D Graphics Mark 794
3D Graphics Mark 7,821
Memory Mark 2,817
Disk Mark 13,226 (PerfectDisk SmartOptimize ran between passmark runs).

What can the GT80 with the 4980HQ set / benchmark through XTU / passmark?

Passmark lists i7-4980HQ CPU Mark as 10062 vs. 10,937/12,321 OC 5950HQ

Intel Core i7-4980HQ @ 2.80GHz Passmark CPU Mark Rating
http://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Core+i7-4980HQ+@+2.80GHz&id=2327

 

Weird, I forgot what to ask.....

 

SagerFanatic, you could also wait for the Skylake CPU model of the GT80.

I doubt the Skylake CPU will be enough of an improvement to wait, but the new chipset might give you more future-proofing - what with the PCIE x4 slots (2?) and the USB 3.1 ports.

The GT80 model 263 with 4x 512GB M.2 SATA gives me 2TB of fast storage. The 1x/2x PCIE will max out at 512GB/1TB. For me more storage was more important than a bit faster.



The DDR4 SODIMM's are an unknown, so far DDR4 hasn't really been worth the effort on the desktop, but that has been changing, time will tell how DDR4 SODIMM's perform.

My g.skill 4x 8GB 2133mhz DDR3L is performing much better than expected, better than in the G750, so for me the current model with DDR3L is the best choice.

For me I was happy enough with the 14nm Broadwell 5950HQ GT80 to purchase now.

 

Yeah, you may as well wait. They are approaching.

 

I'm not an expert so I will take your words on it, if its worth the wait and for me to save up more budget then I'll wait.

 

SagerFanatic, no one really knows

I bought the Broadwell GT80 263, it works for me right now just fine, and has a long life with GPU upgrades for the next few years.

For me Skylake is too new, and the foreseeable improvements over the GT80 263 that a Skylake GT80 can provide aren't interesting enough to wait for. Months will go by before it releases and is out there long enough to hear of the problems and watch them get fixed.

If I like what I see over the next year with Skylake releases, into the Cannonlake range, I will likely wait for Cannonlake to mature to replace the GT80 frame.

One way to look at it is that Broadwell is the culmination of an architecture that has been evolving for years through Haswell, and now is at it's peak of implementation performance in a 14nm die shrink. The chipset is optimized, the drivers are well debugged, and the reliability is known.

Skylake brings a completely new chipset, with a new CPU architecture - grown out of Haswell / Broadwell - but new enough to be an unknown.

Whatever comes out first will be the first implementation of hardware, drivers, and configuration / tuning for Skylake.

Skylake will likely have a few releases of CPU's and chipsets, improving and growing better over time, like Haswell / Broadwell.

It might be that Skylake -H mobile, what we are interested in, will be 100% out of the gate without problems, and then again - it might be ok and iterate toward 100% over time.

Just like the desktop Skylake, it is likely the Mobile Skylake will ship with entry level CPU's first and release higher end CPU's later.

I am sure MSI will do a good job getting their GT80 Skylake released as quick as they can and make it as good as it can be upon release.

Besides, if you can wait a couple of months or more to see how the Skylake GT80 performs you might find a Broadwell/Haswell GT80 at a small discount. I think the Haswell models are $100-400 off now...

http://www.newegg.com/Product/Produ...ription=gt80&bop=And&Order=PRICED&PageSize=30

 

@SagerFanatic There was a nice review of the GT72 using the i7-5950HQ here.

 

Thanks Dufus!! I have been looking for other 5950HQ reviews/notes to compare against my results.

I was able to duplicate their numbers, but to do it I had to add positive voltage offset for both CPU and cache - and so my CPU temp higher - no thermal throttling, but I seem to not be able to bump up anything on my CPU through XTU without adding some positive offset voltage for the CPU.

 

Well, this is disappointing, and not a good omen for the Skylake mobile CPU/chipset:

Well, This is Disappointing
http://www.techspot.com/review/1041-intel-core-i7-6700k-skylake/page14.html

The Skylake i7-6700K CPU/iGPU was buried by the Broadwell i7-5775C CPU/iGPU
http://www.techspot.com/review/1041-intel-core-i7-6700k-skylake/page9.html

 

5950HQ have a baseclock of 2.9GHz and can go up to 3.8GHz
6700HQ have a baseclock of 2.6GHz and can go up to 3.5GHz

Add 10% IPC, 5950HQ and 6700HQ should be very similar in CPU performance.

6700HQ will have support for PCIe SSDs through M2.
5850HQ have a much faster IGP than 6700HQ.

Your call

 

Well they did test with ambient at 20C, if you ran anything higher than that then add the difference to their result for a better comparison. Might just be luck of the draw with voltages.

AFAIK the 5950HQ has L4 cache like the 5775C but not the 5700HQ or 6700K, which may help with performance.

@Cloudfire, where did you get the clock specs for the 6700HQ?

 

Oh you know, here and there

 

Dufus, my ambient was 25c, so knocking off 5c from 92c = 87c - all due to the Voltage bump.

My 5950HQ is a voltage hound above 40x, and it does it without Thermal Throttling.

At stock settings, up to 40x/40x/40x/40x, the 5950HQ is running cool and happy with no additional +V. XTU Bench 85c @ 40x x 4, 79c @ 38x/37x/37x/37x, 77c with -75mV

The performance difference at and above 42x is measurable, but not necessary - all games are well beyond needing any additional CPU help.

I will be happy to run it cool clocked below the need for +V offset, 40x x 4.

Update: Part of the higher temps might be because I bench using 100%/100% Power to the CPU, High Performance, so the CPU never downclocks between load spikes - 0%/100% or Balanced will produce lower results, both in scores and temperatures.

 

I seriously doubt there is even a 10% IPC difference between skylake and Broadwell at this point, considering the lackluster reviews so far.

But I would still go for the x700HQ variations, due to TDP. Unless the rumors are true and they released an unlocked skylake processor for laptops. Well, I think even current broadwell options are great, unless you want the extras with skylake, like DDR4, PCIe storage etc.

I am fine with my lowly 4720HQ even if I only score like 730 in XTU benchmark But it would be so awesome if we could unlock the current hasswell/broadwell CPUs to run uo to 60w TDP. There is a lot of performance to be hardnessed from these CPUs.

 

A couple of Skylake performance notes...URL's below.

1) 20 PCIe 3.0 lanes in the new 6700k are really 4 lanes in/out of the CPU and 16 lanes competing for access to the CPU. Max throughput after overhead in/out of the CPU is only 3.5GB/sec - not enough to meet dreams of 2-3x PCIe x4 NVME speeds. Also, ops/sec are limited, no number cited, but said much less than 200k ops/sec total.

http://www.pcper.com/reviews/Storag...logy-Tested-PCIe-and-SATA-RAID/PCIe-RAID-Resu

http://www.pcper.com/reviews/Storag...-Technology-Tested-PCIe-and-SATA-RAID-updated

 

Usually they don't change the PCI-E from the desktop, so 16x from the CPU to anything you like and 4x to the chipset (which has more connections).

Yes that's a total allowance of 3.5GB/sec from the chipset which I don't think leaves much to cry over.

 

Ambients don't correlate to a linear increase/decrease. I've had a hot humid day of 34c ambients product 15 degrees hotter CPU than a nighttime ambient temp of 28c with less humidity

Because it isn't. There's at best 5% from broadwell, which is at best 10% from haswell, but even so it's still barely anything, since the "at best" doesn't mean "most of the time".

Would still be worth it to buy the P37xSM-A or P7xxZM models though. Especially if the TDP limit is there.

XTU Bench is so RAM dependent it's a joke. Try tossing in 2 x 2133MHz sticks and make sure they're in dual channel with good timings and see how much your XTU Bench score gets bumped.

 

can you do a cinebench score for the cpu too?

 

Phase, at some point I will get around to running and posting benchmarks, Cinebench included. Working on optimizing IO

 

D2 Ultima, hit's on both, it's very humid here, and I have 4 x 8GB 2133mhz g.skill memory installed

 

Meaker, nothing to cheer about either, the SATA RAID performance still hits a 1.8GB/sec ceiling, and the PCIe RAID implementation is jerry-rigged and doesn't even fully scale for 2x PCIe RAID0 performance.

With the ceiling a bit higher than 1x PCIe at 3.5GB/sec, it's not 2x 8GB/sec to 4x 16GB/sec - and that 3.5GB / sec is shared - with everything on PCIe. BTW, it's supposed to be 4GB/sec, but it's eaten up by overhead - even with only 1 active device.

We still need a few more iterations of chipset's before we can see 3-4x full speed PCIe throughput.

 

I know my stuff *buffs nails and blows on them all proud-like*

 

The 3.5GB is shared yes for all devices conected to the southbridge (The GPUs or any CPU connected device is not restricted like that) but people get by pretty quickly with 1.75GB/sec just now so double that leaves a lot of head room still

The extra lanes on the southbridge do let you use the speed where you need it when you need it. It's rare that people need that speed on all devices simultaneously.

 

I have to admit curiosity between Crystal Lake and Broadwell myself... Technically Broadwell should be faster... But it's only a 5% IPC over Haswell... We need to see clock vs clock.

 

just weird because with my haswell 4810mq overclocked to 4.2ghz, i only get in the 700s with xtu

 

You can't set a 4810MQ to 4.2GHz except on 1 core
Unless you set current limit to 256A you'll be throttling in XTU bench anyway
XTU bench is so RAM dependent it's a joke; you probably have 1600MHz 11-11-11-27 like I do, which explains the low score

 

well 4.2ghz with 1 core. 4.1ghz with 2, and then 3 and 4 are at 4.0ghz. should still beat that broadwell card. and yeah i have 1600mhz of 32gb of ram. that why i asked about cinebench because it's closer to realworld performance

 

Any benchmark that runs in Windows is a joke as far as CPU is concerned. Grab phoronix for Linux and test CPU vs CPU with an OS that doesn't do a bunch of useless trash in the background.

 

You should be in the 800's with that memory unless some throttling going on.

As D2Ultima says RAM can make a big difference with XTU. For instance with my i7-4700MQ with standard 2x8GB 1600MT/s CL11 RAM I scored something like 840 IIRC @ ~4GHz, by overclocking that RAM to 1866MTS pushed the score to 907. Reminds me I should try again sometime with 2400MT/s RAM.

 

eh. i mean i get 763 overclocked in cinebench. hopefully skylake desktop in the batman laptops will be a big improvement when i get one for video editing

well hopefully when i get the batman laptop with a 6700k overclocked and ddr4 fast ram, i'll see an improvement lol

 

No doubt, record last time I looked for i7-6700K was 1800, LN2 of course

 

Faster but with awful timings. Your XTU bench might actually decrease in score.

 

D1 Ultima, I was thinking the same thing when I saw DDR4 SODIMMs @ 2133mhz, that is kinda the starting speed for DDR4, but again we don't know until it all ships and we can test it

http://www.corsair.com/en-gb/corsai...mm-2133mhz-c15-memory-kit-cmso16gx4m2a2133c15

That is "value select", the low end budget priced ram from Corsair, that price is 197.86GBP = 197.86 US $ = $395.72 for 32GB

Here is 32GB of DDR3L 2133mhz g.skill memory for $259.99, or $135.73 less:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231776

So the new stuff is going to be more $ for the same or less performance?

 

LOL
15-15-15-36. And we can get 2133MHz 11-11-11-31 with DDR3L, and better with DDR3.
This is gonna be a fun time.

 

FTFY

 

DDR4 shouldn't be for notebooks yet. It's more money for worse performance, and unlike on say... X99 desktop boards, we can't just go "oh, 1.35v up from 1.2v and get 12-13-13-35 2400MHz quad channel" or something, which is actually decent (far less what can be eked out individually).

 

Phase, I got almost as much of an improvement in performance numbers for CPU on Cinebench by under-volting at stock speeds, as I did OC'ing + over-volting.

cinebench r15 38x/37x cores -85mV 34x cache -5mV #1 - 795 cb, 146.17 fps


cinebench r15 41x cores +20mV 36x cache +0.09mV #2 - 803 cb, 158.92 fps


While tuning, I ran fps as high as 168, and cpu 795-800.

notebookreview has Cinebench CPU for the 4980HQ:
www.notebookcheck.net/Intel-Core-i7-4980HQ-Notebook-Processor.122665.0.html

GT80 4980HQ/980m-SLI: 693 and 699, results from two systems.

GT80 5950HQ/980m-SLI: 803 OC / 795 stock

 

well ouch. that hurts me lol

 

Phase, what are you seeing on your current system?