To answer my own question: Nootbookcheck ran the numbers:
"Intel also listed the new Core i7-4558U, which features 28W of TDP and is designed for larger 15" devices. This CPU managed performance that was 2.25 times that of the 3687U, garnering a score of around 1,350 points. Astonishingly, this is not only faster than AMD's Radeon HD 7660G, but also rivals that of the dedicated GeForce GT 640M LE"
You can compare the old HD4000 vs the 650M using AnandTechs own Bench database: http://www.anandtech.com/bench/Product/601?vs=622 3dMark 11 sees the 650M as a 3x favorite, so the GT3e seems to get more or less the same score. For real games, HD4000 was trialing the nVidia card with factors between 2 and 4, so it will be interesting to see real-world benchmarks for the new parts.
I wonder how AMD's next gen APU's graphics will fare compared to Iris or Iris Pro. AMD probably prepared for Intel to step it up in GPU performance, it's just a matter of if they expected that much of a step up....
They absolutely did. They know that graphics performance scales very well with the number of execution units, and Intel has announced a long time ago that they would double the number of EUs. So twice the performance is exactly what everybody expected from Haswell.
And will these 15W parts also be going in tablets? Because i think we was all expecting better than this from Haswell. What happened to the sub 10W parts Intel was showing off?
All of Intel's ULV Haswell parts are getting released later then their high-end chips. It may seem like Haswell is imminent in "ultrabooks" and tablets but this is not the case
It was all another lie/misleading fact from Intel, and Anand, unfortunately, kept that lie alive until very recently.
Haswell is only going to achieve a 10W TDP by doing the same thing they did with that IVB with 13W TDP, where they cut the clock speed down to 800 Mhz.
So I'll tell you EXACTLY what Intel is going to do. They will release a 800 Mhz or so Haswell version for "mobile". They will start lying about its real clock speed, though, just like they are doing for Atom now, and say it's actually a 1.3 Ghz or higher chip (with Turbo-Boost, but of course they'll 'forget" to mention this in their press releases). Then they'll probably tell the second lie, that it's a "7W chip" ("whoops, did we also "forget" to mention it's SDP, not TDP?")
Re-read what I said again, remember it, then comeback in fall and tell me how wrong I was. But I'm thinking I'm not wrong at all. I've been paying VERY careful attention to Intel's press releases in the past 2 years, and they are anything if not EXTREMELY MISLEADING. And Anand is not helping by repeating their lies in his podcasts and posts.
For years we've been reading how Intel's Haswell architecture would finally bring mobile TDPs down to the 10-20W range. Now it all seems like a bunch of marketing talk - what we actually get is +2W on standard voltage parts and -2W on ultra-low voltage parts. Would Intel even include a 10W one-chip ULV part as reported recently?
What could have Intel possibly meant by putting Haswell and 10-20W in the same sentence in the past, that Haswell is the first step towards such TDPs? If so it doesn't seem likely we'd see such thermals across the board until a mature 10nm process, which pretty much means 2017.
This is why we need AMD. Intel no longer has reason to deliver anymore with no competition. I hope Kaveri, Kabini and Temash make Intel actually innovate for once. I personally am waiting for Kaveri
TDP is just a design power. If I have a 17w or 15w TDP, I can make it limit itself to 10w by reducing clockspeed. The more important thing is the lower power states, since that's where a CPU will be most of the time in tablet type workloads. How often are you going to max your tablet CPU?
Sure, it might be a little disappointing that TDPs of what's been shown so far aren't that far down, but that doesn't mean battery life won't be potentially significantly improved.
Don't forget that the ultrabook parts at 15w also include the PCH and the integrated voltage regulator - components that are not included in the 17w on previous generation. Factoring those extra ingredients in means that the CPU/gpu/cache/memory controller/display power is more like 11w budget for apples-to-apples vs IvyBridge.
The 15w Haswells have U sku names; There may also be Y skus that hit lower power envelopes.
I'm well aware of this, as mentioned in the original post, however, Intel led the public to believe (or is that just me?) Haswell's ULV parts would be 10W in a recent presentation; they even went as far as to demo a 7W part (Y). And while it's undeniable we're getting faster, more efficient hardware, it's all going a bit slower than expected, at least for me.
Any word on what resolutions these comparisons are made on? I'm hoping for next gen notebooks to be at minimum 1920x1080 or better (although I don't hold my breath) 1920x1200. Not to speak of retina displays. And I don't care about gaming, OS UI, video, web rendering is just fine for me.
Highly unlikely. Iris Pro starts at 47W; and while the Edge is capable of cooling two chips with a total power consumption of about 37W, it wouldn't be able to handle a single 'big' chip at 47W. It would be able to handle Iris (non-Pro) in the form of the 28W chip, but this part probably wouldn't be significantly faster (if at all) than the 640 LE. Therefore, it makes more sense to offer the 2nd gen Edge with a 15W (or even lower) Haswell ULV coupled with a 700-series or higher GeForce.
It is good to be a customer next year and end of this year. Performance pressure from Intel GPUs mean that the bottom falls out for discrete GPUs in laptops. Means that discrete vendors will have to put out much better parts than they care to or shrink away from that market totally. AMD will be sucking wind on their APUs. They are close to losing the only advantage they have. AMD needs a game changer. Will their ARM license start showing up?
While it's good that intel is significantly increasing iGPU performance with Haswell there is little to no chance they will rival Kaveri especially with HSA and GCN. They will also lose to richland in iGPU performance. The other problem is that CPU performance is only about 5% better, 10% in certain benchmarks. AMD is increasing CPU performance much faster then Intel and their iGPU lead is not evaporating.
The 28w ones could go into 13.3" ultrabooks as well, the MacBook retina 13" has a 35w ivy bridge, and haswell has an integrated voltage regulator, which should decrease the power consumption.
just before I saw the check which had said $4099, I have faith that...my... cousin was truley erning money part time on there computar.. there sisters neighbour haz done this for only twentey months and just now took care of the dept on their villa and bought Ariel Atom. this is where I went, Bow6.comCHECK IT OUT
I said it even before Anand thought about it, that if anything the "tablet" version of Haswell will be a VERY underpowered version of the chip., but for some reason Anand kept pushing the idea that Haswell will "get efficient enough" to be used in tablets.
But it's not "getting that efficient". It's only going to cut the performance dramatically to even achieve a 10W TDP, which is still far too high for a tablet - unless you want a 15,000 mAh battery in it, but good luck with the pricing and weight on that one.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
30 Comments
Back to Article
GuniGuGu - Thursday, May 2, 2013 - link
How do these benchmark compare to the current GPU offerings such as the popular Nvidia 650m?GuniGuGu - Thursday, May 2, 2013 - link
To answer my own question: Nootbookcheck ran the numbers:"Intel also listed the new Core i7-4558U, which features 28W of TDP and is designed for larger 15" devices. This CPU managed performance that was 2.25 times that of the 3687U, garnering a score of around 1,350 points. Astonishingly, this is not only faster than AMD's Radeon HD 7660G, but also rivals that of the dedicated GeForce GT 640M LE"
http://www.notebookcheck.net/Intel-releases-perfor...
ShieTar - Thursday, May 2, 2013 - link
You can compare the old HD4000 vs the 650M using AnandTechs own Bench database:http://www.anandtech.com/bench/Product/601?vs=622
3dMark 11 sees the 650M as a 3x favorite, so the GT3e seems to get more or less the same score. For real games, HD4000 was trialing the nVidia card with factors between 2 and 4, so it will be interesting to see real-world benchmarks for the new parts.
Kabij2289 - Thursday, May 2, 2013 - link
I wonder how AMD's next gen APU's graphics will fare compared to Iris or Iris Pro.AMD probably prepared for Intel to step it up in GPU performance, it's just a matter of if they expected that much of a step up....
ShieTar - Thursday, May 2, 2013 - link
They absolutely did. They know that graphics performance scales very well with the number of execution units, and Intel has announced a long time ago that they would double the number of EUs. So twice the performance is exactly what everybody expected from Haswell.B3an - Thursday, May 2, 2013 - link
So if the 15W parts don't have Iris, then what do they have? HD4000 still?B3an - Thursday, May 2, 2013 - link
And will these 15W parts also be going in tablets? Because i think we was all expecting better than this from Haswell. What happened to the sub 10W parts Intel was showing off?kirilmatt - Saturday, May 4, 2013 - link
All of Intel's ULV Haswell parts are getting released later then their high-end chips. It may seem like Haswell is imminent in "ultrabooks" and tablets but this is not the caseKrysto - Tuesday, May 14, 2013 - link
It was all another lie/misleading fact from Intel, and Anand, unfortunately, kept that lie alive until very recently.Haswell is only going to achieve a 10W TDP by doing the same thing they did with that IVB with 13W TDP, where they cut the clock speed down to 800 Mhz.
So I'll tell you EXACTLY what Intel is going to do. They will release a 800 Mhz or so Haswell version for "mobile". They will start lying about its real clock speed, though, just like they are doing for Atom now, and say it's actually a 1.3 Ghz or higher chip (with Turbo-Boost, but of course they'll 'forget" to mention this in their press releases). Then they'll probably tell the second lie, that it's a "7W chip" ("whoops, did we also "forget" to mention it's SDP, not TDP?")
Re-read what I said again, remember it, then comeback in fall and tell me how wrong I was. But I'm thinking I'm not wrong at all. I've been paying VERY careful attention to Intel's press releases in the past 2 years, and they are anything if not EXTREMELY MISLEADING. And Anand is not helping by repeating their lies in his podcasts and posts.
tipoo - Thursday, May 2, 2013 - link
4600 maybe.Zandros - Thursday, May 2, 2013 - link
A few have GT3 graphics (HD5000), most have HD4600 IIRC. As we saw in the previous article, only HD5100 and up get Iris branding.yhselp - Thursday, May 2, 2013 - link
For years we've been reading how Intel's Haswell architecture would finally bring mobile TDPs down to the 10-20W range. Now it all seems like a bunch of marketing talk - what we actually get is +2W on standard voltage parts and -2W on ultra-low voltage parts. Would Intel even include a 10W one-chip ULV part as reported recently?What could have Intel possibly meant by putting Haswell and 10-20W in the same sentence in the past, that Haswell is the first step towards such TDPs? If so it doesn't seem likely we'd see such thermals across the board until a mature 10nm process, which pretty much means 2017.
kirilmatt - Saturday, May 4, 2013 - link
This is why we need AMD. Intel no longer has reason to deliver anymore with no competition. I hope Kaveri, Kabini and Temash make Intel actually innovate for once. I personally am waiting for KaveriLonyo - Thursday, May 2, 2013 - link
TDP is just a design power.If I have a 17w or 15w TDP, I can make it limit itself to 10w by reducing clockspeed.
The more important thing is the lower power states, since that's where a CPU will be most of the time in tablet type workloads.
How often are you going to max your tablet CPU?
Sure, it might be a little disappointing that TDPs of what's been shown so far aren't that far down, but that doesn't mean battery life won't be potentially significantly improved.
rootheday - Thursday, May 2, 2013 - link
Don't forget that the ultrabook parts at 15w also include the PCH and the integrated voltage regulator - components that are not included in the 17w on previous generation. Factoring those extra ingredients in means that the CPU/gpu/cache/memory controller/display power is more like 11w budget for apples-to-apples vs IvyBridge.The 15w Haswells have U sku names; There may also be Y skus that hit lower power envelopes.
yhselp - Thursday, May 2, 2013 - link
I'm well aware of this, as mentioned in the original post, however, Intel led the public to believe (or is that just me?) Haswell's ULV parts would be 10W in a recent presentation; they even went as far as to demo a 7W part (Y). And while it's undeniable we're getting faster, more efficient hardware, it's all going a bit slower than expected, at least for me.Amagus - Thursday, May 2, 2013 - link
10W and sub-10W parts are coming later in the year. They won't be part of the initial launch.RU482 - Thursday, May 2, 2013 - link
yes - those are the Y-SKU. the were originally slated to be 7.5W "SDP", but that has recently been reduced to 6W!kirilmatt - Saturday, May 4, 2013 - link
SDP is really a scam, they should measure everything on the same scale, "Y" series chips are just marketingConficio - Thursday, May 2, 2013 - link
Any word on what resolutions these comparisons are made on?I'm hoping for next gen notebooks to be at minimum 1920x1080 or better (although I don't hold my breath) 1920x1200. Not to speak of retina displays.
And I don't care about gaming, OS UI, video, web rendering is just fine for me.
HisDivineOrder - Thursday, May 2, 2013 - link
Here's hoping it's good enough to make a great Razer Edge tablet with Iris Pro.yhselp - Friday, May 3, 2013 - link
Highly unlikely. Iris Pro starts at 47W; and while the Edge is capable of cooling two chips with a total power consumption of about 37W, it wouldn't be able to handle a single 'big' chip at 47W. It would be able to handle Iris (non-Pro) in the form of the 28W chip, but this part probably wouldn't be significantly faster (if at all) than the 640 LE. Therefore, it makes more sense to offer the 2nd gen Edge with a 15W (or even lower) Haswell ULV coupled with a 700-series or higher GeForce.kirilmatt - Saturday, May 4, 2013 - link
Or they could offer a Kaveri ULV APU that could perform better at a lower TDP and costshing3232 - Thursday, May 2, 2013 - link
THIS IS JUST BENCHMARK...... I want real game bech.eanazag - Friday, May 3, 2013 - link
It is good to be a customer next year and end of this year. Performance pressure from Intel GPUs mean that the bottom falls out for discrete GPUs in laptops. Means that discrete vendors will have to put out much better parts than they care to or shrink away from that market totally. AMD will be sucking wind on their APUs. They are close to losing the only advantage they have. AMD needs a game changer. Will their ARM license start showing up?evilspoons - Saturday, May 4, 2013 - link
Looks like this should help out on the 13" Retina Macbook Pro. Interesting.kirilmatt - Saturday, May 4, 2013 - link
While it's good that intel is significantly increasing iGPU performance with Haswell there is little to no chance they will rival Kaveri especially with HSA and GCN. They will also lose to richland in iGPU performance. The other problem is that CPU performance is only about 5% better, 10% in certain benchmarks. AMD is increasing CPU performance much faster then Intel and their iGPU lead is not evaporating.jpal12 - Sunday, May 5, 2013 - link
The 28w ones could go into 13.3" ultrabooks as well, the MacBook retina 13" has a 35w ivy bridge, and haswell has an integrated voltage regulator, which should decrease the power consumption.MelodyRamos47 - Sunday, May 12, 2013 - link
just before I saw the check which had said $4099, I have faith that...my... cousin was truley erning money part time on there computar.. there sisters neighbour haz done this for only twentey months and just now took care of the dept on their villa and bought Ariel Atom. this is where I went, Bow6.comCHECK IT OUTKrysto - Tuesday, May 14, 2013 - link
So much for the "much more efficient" Haswell.I said it even before Anand thought about it, that if anything the "tablet" version of Haswell will be a VERY underpowered version of the chip., but for some reason Anand kept pushing the idea that Haswell will "get efficient enough" to be used in tablets.
But it's not "getting that efficient". It's only going to cut the performance dramatically to even achieve a 10W TDP, which is still far too high for a tablet - unless you want a 15,000 mAh battery in it, but good luck with the pricing and weight on that one.