Would really have loved to see even just 3D Mark results from this, but such is the nature of restricted testing. I'm still not convinced that this CPU has a particularly attractive position in the market with its somewhat odd cost/power/performance balance.
Thanks for the CPU benchmarks. Bummer to hear there is no benefit from the eDRAM when using current applications :-(
I enjoyed your previous Isis Pro 5200 article. However, please add a DX9 game such as Starcraft 2 to the testing scenarios. Intel's latest and greatest graphics (Crystalwell) really needs more analysis as nobody seems to be able to get there hands on one yet. :-(
Clevo, you are removing fingerprint readers from most of your laptops now. Please add them back! My customers don't want to lose a fingerprint reader to gain an IPS screen. It's the biggest complaint I've had this year.
If windows doesn't utilize the L4 cache for much other than for processing graphics, is if a fair assumption that it is probably not going to be the preferred choice for hypervisor use, Ian?
Any idea of power consumption? The biggest promise for me about Iris Pro was lower power consumption due to being only one package. But an <a href="http://translate.google.com.ar/translate?sl=auto&a... review</a> puts it ahead of a Haswell/GT650m by only a handful of W when the GPU is loaded.
Knowing CB, I wouldn't give too much stock on their results. Back when I used to read them (at a time when I didn't know about AT), they tested stuff "at 50% brightness" and compared everything in one chart. Oblivious that "50%" is not an adequate comparison point at all. Unless they have all their testing methodology updated and display on a separate side, I don't see them explaining what their settings are for that test. And just for reference, they use Crysis 3 in the power consumption test, which renders identically on the Iris Pro and the 750M. So in essence, the 750M system uses 10% more power for the same performance in that game. In general, the 750M is 25% faster, but that is done in other games. To get a better sense of the part (even in such a flawed setup with 2 totally different systems), they should have tested at least 2 more settings, one where the 750M is clearly better, one where the Iris Pro is better. As it stands, that test is near worthless in my opinion. One of the reason I quit reading CB. Along with numerous factual errors.
Thanks for the details, I find it tedious to read through a very long google translate. Much appreciated. Hopefully there are second opinions not far behind.
WinRAR seems to be an oddball result here. Was the memory speed on the 4770K at 4750HQ speeds also changed to match that platform?
So far the L4 cache seems to be good for a few more % increase in IPC. I wonder if using a discrete GPU in addition to CrystalWell would further increase the IPC. The only test I've seen that showed a radical gain was in fluid dynamics. Having said that, I still want CrystalWell in a socket 1150 part with all the features enabled (overclocking, TSX, VT-d, Hyperthreading etc.).
Both were run at 1600 C11; I think that the extra L3 cache of the 4770K helped this result, and by extension suggests that WinRAR adapts its code to the amount of L3 available, and doesn't detect the L4.
Until we get a 4770R or similar in a full-on desktop model, we could only speculate at what effect the eDRAM has on anything gaming or compute. As far as I can tell, not many (if any) manufacturers are working on it - the 4x70R chips are ideally placed for AIOs, not desktops, but also would command a lot more expense.
Agreed. It is mostly speculation but I feel that there is a clear interest in the enthusiast community for CrystalWell. However Intel or manufacturers haven't shown much interest in mass market systems using it to date. Case in point is this article where it seems that you had to jump through some hops to get a mere 90 minutes of testing in (which BTW thank you).
With regards to the WinRAR test, the extra 2 MB of L3 cache would help the result but it seems like an abnormally large jump. It wouldn't have been my first guess for the performance delta.
in general, they conclude that the Iris Pro 5200 is roughly on par with the NVidia 740M. See http://www.notebookcheck.net/Intel-Iris-Pro-Graphi...">here for more comparison data vs other mobile GPUs (including some desktop Trinity and Richland, not so much on AMD mobile skus - have to look at the pages for those specifically).
I still want to see a comparison between Iris and Iris Pro, to decide if I should flame Intel or certain tablet manufacturers for not offering the latter. I'm pretty sold on flaming the ones that don't offer Iris at all and stick to GT2, but I am still not convinced about GT3 vs GT3e.
It looks like from the XTU screenshots that this laptop is overclockable?
In the past clevo has only allowed overclocking on their 17" laptops, with 15" models supporting the 17" BIOS to get overclocking that way, and reference clock overclocking required a BIOS (ME FW) mod.
Nice one Intel, put 128 MB of extra memory on chip, result in no computational benefit... I at least expected to have some influence in C++ AMP, but it seems all that memory will be entirely dedicated to a GPU that is close to useless in performance scenarios...
I wish intel used die space to put more CPU cores instead of their mediocre, GPUs - Haswell could easily pack FOUR more cores in that die area, I'd prefer DOUBLE CPU performance in the place of a crappy GPU I don't even use...
Well... I use the 'crappy GPU', so I would prefer the extra on chip 128MB be made availabe for graphics... I no longer want to pay for discrete graphics boards in terms of dollars or real-estate. So, I'm with the crowd who'd like to see a top of the line socketed I7-47XX part with GT3e for my mini-ITX build.
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.
Facebook
Twitter
RSS
28 Comments
Back to Article
Spunjji - Monday, July 29, 2013 - link
Would really have loved to see even just 3D Mark results from this, but such is the nature of restricted testing. I'm still not convinced that this CPU has a particularly attractive position in the market with its somewhat odd cost/power/performance balance.IanCutress - Monday, July 29, 2013 - link
Anand has 3DMark numbers in his analysis of the 4950HQ: http://anandtech.com/show/6993/intel-iris-pro-5200... - the difference between the two IGPs is 100 MHz turbo.Anonymuze - Tuesday, October 1, 2013 - link
GPU MHz difference is less important - the main difference will come from the bios turbo TDP limits (and whether there's any thermal throttling).coolhardware - Monday, July 29, 2013 - link
Thanks for the CPU benchmarks. Bummer to hear there is no benefit from the eDRAM when using current applications :-(I enjoyed your previous Isis Pro 5200 article. However, please add a DX9 game such as Starcraft 2 to the testing scenarios. Intel's latest and greatest graphics (Crystalwell) really needs more analysis as nobody seems to be able to get there hands on one yet. :-(
Finally, Anand previously mentioned lower image quality in Haswell based QuickSync encoding:
http://www.anandtech.com/show/7003/the-haswell-rev...
has Intel address this issue?
DanNeely - Monday, July 29, 2013 - link
It's disappointing; but it does make Intel's decision not to offer a high end desktop part with the eDRAM available more understandable.coder543 - Monday, July 29, 2013 - link
System76 is located in Denver, Colorado, not Canada. Just FYI.IanCutress - Monday, July 29, 2013 - link
I'm not sure why my mind fixated them being based in Canada. Thanks for the heads up.Ian
Flunk - Monday, July 29, 2013 - link
That's probably because Denver Colorado looks a lot more like the American stereotype for Canada than most of Canada does.VivekGowri - Monday, July 29, 2013 - link
Perhaps you mixed them up with Eurocom?apriest - Monday, July 29, 2013 - link
Clevo, you are removing fingerprint readers from most of your laptops now. Please add them back! My customers don't want to lose a fingerprint reader to gain an IPS screen. It's the biggest complaint I've had this year.lkraav - Monday, July 29, 2013 - link
Smartcard reader is another "thing" in my book. Might have to stick with Lenovo and Dell :/Helang - Wednesday, July 31, 2013 - link
Uhh, how to purchase one like that in the U.S.gxtoast - Monday, July 29, 2013 - link
If windows doesn't utilize the L4 cache for much other than for processing graphics, is if a fair assumption that it is probably not going to be the preferred choice for hypervisor use, Ian?andy o - Monday, July 29, 2013 - link
Any idea of power consumption? The biggest promise for me about Iris Pro was lower power consumption due to being only one package. But an <a href="http://translate.google.com.ar/translate?sl=auto&a... review</a> puts it ahead of a Haswell/GT650m by only a handful of W when the GPU is loaded.Death666Angel - Monday, July 29, 2013 - link
Knowing CB, I wouldn't give too much stock on their results. Back when I used to read them (at a time when I didn't know about AT), they tested stuff "at 50% brightness" and compared everything in one chart. Oblivious that "50%" is not an adequate comparison point at all. Unless they have all their testing methodology updated and display on a separate side, I don't see them explaining what their settings are for that test. And just for reference, they use Crysis 3 in the power consumption test, which renders identically on the Iris Pro and the 750M. So in essence, the 750M system uses 10% more power for the same performance in that game. In general, the 750M is 25% faster, but that is done in other games. To get a better sense of the part (even in such a flawed setup with 2 totally different systems), they should have tested at least 2 more settings, one where the 750M is clearly better, one where the Iris Pro is better. As it stands, that test is near worthless in my opinion. One of the reason I quit reading CB. Along with numerous factual errors.andy o - Monday, July 29, 2013 - link
Thanks for the details, I find it tedious to read through a very long google translate. Much appreciated. Hopefully there are second opinions not far behind.Kevin G - Monday, July 29, 2013 - link
WinRAR seems to be an oddball result here. Was the memory speed on the 4770K at 4750HQ speeds also changed to match that platform?So far the L4 cache seems to be good for a few more % increase in IPC. I wonder if using a discrete GPU in addition to CrystalWell would further increase the IPC. The only test I've seen that showed a radical gain was in fluid dynamics. Having said that, I still want CrystalWell in a socket 1150 part with all the features enabled (overclocking, TSX, VT-d, Hyperthreading etc.).
IanCutress - Monday, July 29, 2013 - link
Both were run at 1600 C11; I think that the extra L3 cache of the 4770K helped this result, and by extension suggests that WinRAR adapts its code to the amount of L3 available, and doesn't detect the L4.Until we get a 4770R or similar in a full-on desktop model, we could only speculate at what effect the eDRAM has on anything gaming or compute. As far as I can tell, not many (if any) manufacturers are working on it - the 4x70R chips are ideally placed for AIOs, not desktops, but also would command a lot more expense.
Kevin G - Monday, July 29, 2013 - link
Agreed. It is mostly speculation but I feel that there is a clear interest in the enthusiast community for CrystalWell. However Intel or manufacturers haven't shown much interest in mass market systems using it to date. Case in point is this article where it seems that you had to jump through some hops to get a mere 90 minutes of testing in (which BTW thank you).With regards to the WinRAR test, the extra 2 MB of L3 cache would help the result but it seems like an abnormally large jump. It wouldn't have been my first guess for the performance delta.
rhx123 - Monday, July 29, 2013 - link
No word on any UK releases for this model from any of the resellsers. Rather sad.manwdaplan - Monday, July 29, 2013 - link
Put this in the next Surface Pro and I would be pretty excited. Or maybe the next gen Razer tabletmikk - Monday, July 29, 2013 - link
+2MB L3 doesn't make such a huge difference in Winrar. Maybe different turbo speeds?rootheday - Tuesday, July 30, 2013 - link
folks interested in Iris Pro 5200 gaming in a notebook form factor may want to look at a http://www.notebookcheck.com/Test-Schenker-S413-No...">review of the Schenker S413 at notebookcheck.comin general, they conclude that the Iris Pro 5200 is roughly on par with the NVidia 740M. See http://www.notebookcheck.net/Intel-Iris-Pro-Graphi...">here for more comparison data vs other mobile GPUs (including some desktop Trinity and Richland, not so much on AMD mobile skus - have to look at the pages for those specifically).
Helang - Tuesday, July 30, 2013 - link
Where to buy the "clevo" in North America?Visual - Tuesday, July 30, 2013 - link
I still want to see a comparison between Iris and Iris Pro, to decide if I should flame Intel or certain tablet manufacturers for not offering the latter. I'm pretty sold on flaming the ones that don't offer Iris at all and stick to GT2, but I am still not convinced about GT3 vs GT3e.Khenglish - Wednesday, July 31, 2013 - link
It looks like from the XTU screenshots that this laptop is overclockable?In the past clevo has only allowed overclocking on their 17" laptops, with 15" models supporting the 17" BIOS to get overclocking that way, and reference clock overclocking required a BIOS (ME FW) mod.
ddriver - Friday, August 9, 2013 - link
Nice one Intel, put 128 MB of extra memory on chip, result in no computational benefit... I at least expected to have some influence in C++ AMP, but it seems all that memory will be entirely dedicated to a GPU that is close to useless in performance scenarios...I wish intel used die space to put more CPU cores instead of their mediocre, GPUs - Haswell could easily pack FOUR more cores in that die area, I'd prefer DOUBLE CPU performance in the place of a crappy GPU I don't even use...
HardwareDufus - Wednesday, August 14, 2013 - link
Well... I use the 'crappy GPU', so I would prefer the extra on chip 128MB be made availabe for graphics...I no longer want to pay for discrete graphics boards in terms of dollars or real-estate.
So, I'm with the crowd who'd like to see a top of the line socketed I7-47XX part with GT3e for my mini-ITX build.