G73Jh PCIe speeds

Thread to discuss PCIe thruput of HM55 in G73, etc.

Edit: posts have been merged in here from http://forum.notebookreview.com/asus-gaming-notebook-forum/531373-asus-updates-official.html which derailed.

 

I thought this whole debate on the PCI-E bus was done. It's a PCI-E 2.0 capable unit, limited by the HM/PM55 mobile chipset.

 

Yeah I am looking into that now, I did not check before on this,but if it is chipset related then it will stay as it is. I'm not going to worry about it though as the other issues will make things a lot better when they are fixed.

 

yea, that's not quite right.... HM55-YES / PM55 - NO (this is the limited part. both are 2.0 slots)
HM55 is fluctuating as well. i have been able to get 3.4 GB out of it.
PM55 is doing 4.7 GB

 

One one hand, I agree with you. On the other hand, these aren't new issues, so when you bought it the machine had been out for 8 months with no fixes ... so it isn't like you should be completely surprised.

also, does the card saturate the 1.1 bus, but only just? I mean, if it was 2.0, there isn't a performance gain to be had (in most situations)

 

If you're looking for the official spec, see page 51, 121 here: http://www.intel.com/Assets/PDF/datasheet/322169.pdf

How the actual motherboard is set up is another question.

 

Yes, Asus has already contacted Intel and were told this was a chipset limitation. No matter what anyone says, it's HM55/PM55 hardware limitation.

 

we all understand what intel says...but.
that still doesn't explain why the mx17x-r2/mx15x with pm55 chipsets show way better performance. other than pm55 is better than hm55.
i may give this a try to see if 4x/8x/16x make a difference.
PCI Express 2.0 Graphics Cards Tested : How Does PCI Express 2.0 Scale?

 

I haven't seen such. I'v seen above 3 GB/S only ONE way, which is useless. The G73jh has 3 GB/s both ways. Which shows overall, intel is not lying, it's around 2.5 gb/s limitation. Also MSI uses PM55 and their speeds match the G73jh.

Also it makes no difference for your benchmarking. I've tested benchmark with drivers where the speeds were only 700 mb/s each way and then tested with the 10.9-10.10 where I got 3.1 gb/s and the benchmarks were identical. No difference at all. My conclusion is unless you are using OpenCL or Stream, GPGPU applications, you won't notice any improvement.

And also, anything that is published by tomshardware I ignore. Just personal, I don't believe anything they say. But I have read similar article by TechPowerup who I do trust to say something legitimate. The only difference I see is difference between 4x, 8x and 16x, and far as I know, the G73jh is x16. I have seen where for whatever reason, the HD5870M was stuck at 8x and the performance was reduced. But for the actual PCi-e bandwidth speeds, I haven't seen any evidence where it has any impact on gaming or benchmarking using DirectX.

Just looking at benchmarks where the reviewer is using an X58 chipset with a HD5770 you can see that the PCIe 2.0 @ X16 speeds on G73jh are nearly identical. So in general, for those complaining, I haven't a clue why you would be so upset. Unless you are doing GPGPU and using a ton of OpenCL/Stream applications, you have nothing to worry about. Even then you're better off with a FirePro than a Radeon.

 

Unless you are doing GPGPU and using a ton of OpenCL/Stream applications, you have nothing to worry about. <----i do all this.
ps:
our hm55 is slower than any pm55. point blank.
and hm55 is useless in this laptop because we cant even use the on board video of a 620M, which is a big part of the hm series.

 

You said in another thread all you use the G73jh is for benchmarking and having a fast computer for web browsing. By that statement, the PCI-e speeds would still have no impact for you. Either way, I haven't seen any evidence of your claim PM55 being faster than HM55 except for a very few isolated examples, which means squat to me.

What needs to be said except for the few who are complaining, some who don't even use the laptop for other than benchmarking, for what the G73jh was made for, the PCI-e speeds have no impact on gaming or general graphic use.

 

thanks ALLurGroceries!
i was going to move it to the hm55 vs pm55 thread.
http://forum.notebookreview.com/hardware-components-aftermarket-upgrades/455301-pm55-vs-hm55.html

ruckus...i own 12 computers and an laptops. i would be a fool to use this laptop for that, when i have desktop computers far more capable.
also, i never said i was using the g73jh for that, now did i?
but i sure as heck tested it for that.
and as to evidence...you havent been looking hard enough my friend. you need to scout around a bit. plenty of evidence popping up all over the place. but it's all pointing to pm55 being better.(intel specs or not)

 

Can someone post some links to the PM55 vs HM55 numbers? I haven't found them.

 

second that, only thread I remember seeing the laptops compared was comparing the g73jh to a m15x with a 5850 that was massively overclocked. Where are the numbers that show we should even care? The only difference I've seen between our gpu's compared to regular mxm gpu's is that asus gpu is inferior when it comes to overclocking.

 

im just going off direct evidence found in the alienware and clevo threads. going by the better than average numbers..pm55 is pretty much killing hm55.
games/benching/over clocking/gpu applications.
the m15x with 5850 is even showing a link speed of 5 gt

this really only applies to people looking for more performance. the person not concerned..would never notice anything.
so as to the regular person..no, they should not care or notice one bit.( i agree) but to the person that does notice..(you wonder and try different means to improve that)

side note:
can anyone explain why his shows 5 gt?(forget about his over clock)


edit:
did a bit more reading and it seems that different drivers react differently to the gpu to cpu score.(how true that is..i have no clue)

 

Can't find the same evidence. Could you substantiate your claims with some sort of data?

Also, it's hard to take your complaints seriously ... I mean, if you weren't aware that different drivers give different performance numbers, I'm not sure that you're as dedicated to high performance computing as you claim to be. If you could provide a specific example of the benchmarking and how PCIe bus is limiting it, I'd be interested to hear it ...

 

not sure how you missed it.
the pictures pretty much show it all if you looked at them.
1: we do not show a 5 gt link speed. (mine shows 2.5, just like all the rest of the g73's)
2: i used a slower 5850, which is running on a pm55. and that pm55/5gt (which everyone seems to think is impossible)
3 and since im talking about over clocking and not stock. this is where you see it.(for those stuck on stock, you wont notice it)
4: when is th last time you seen a g73 with a vantage gpu over 10k you havent.
5: this hm55 doesn't allow for using the 620/640m on chip gpu, when it should. that was a main part of the hm55
6: all pm55 machines seem to show pcie 2.0 on both sides when viewing gpuz.
7: im hardly complaining. if that were the case i would be running around making conspiracy theory threads like i have see a few of you do.
8 i know where the problem lies, but that's up to asus to make that happen. if they do they do, they dont..oh well.

so, if your not benching marking or..and this is a big part. over clocking...then this makes no sense to you.
if your just running stock, then it should mean nothing to you as well..which i see is the case.

here you go..get your look though on.
look up mr moo's bench marking threads and gaming threads
look up vicioususmc gaming and bench marking threads.
http://forum.notebookreview.com/ali...-powerful-m15x-benchmark-thread-part-2-a.html
http://forum.notebookreview.com/alienware-m17x/455435-official-m17x-benchmark-thread-part-3-a.html
http://forum.notebookreview.com/asus-gaming-notebook-forum/458025-asus-g73-benchmark-thread.html

 

I'm also kind of disappointed that my g73's 5870 core clocks won't OC as high as those with the 5850 from mx15's...

 

clock speed is not an absolute indicator of performance.

 

yeah, me to.

 

That Mx15's 5850's higher clock speed has better performance than my 5870 with lower clock speeds. Makes me a bit sad inside =\

 

Yeah, you're right. the OC'd 5850 is faster than the stock 5870. My mistake... I think?

 

dont take this wrong but...
an OVER CLOCKED 5850M is faster than an OVER CLOCKED 5870M as well.
the only 5870M's keeping up or doing better are the 5870Ms running on a pm55 chipset.

 

an OCed 5870 from a g73. But it seems like some laptops like the msi 660 can reach core clocks of over 900 as well..

 

all laptops being compared also have a standard mxm graphics card, if the pcie speed is off who knows what they did to the graphics card.

 

Still means squat diddly. I know for example Ryzeki can run his at 900 core, but in game performance with my higher memory overclock, our in game performances nearly identical. And looking at M15X users posting in game numbers, I see the same thing. So again, all this, means squat diddly to me. Couldn't care less about complaint of not being able to benchmark as high. Lucky for me, I didn't spend my money on this laptop to get low benchmark scores, but to actually use it, and in that regard, it's fine.

815/1125 overclock is more than generous for me, all considering AMD rated my GPU for 700/1000 settings. I have nothing to complain about.

 

Yes, I am definitely wrong. You have one instance of an OC'd 5850 clocking higher than one instance of an OC'd 5870.

Even if there is a chip-wide inability of the 5870 to reach the same speeds as the 5850, that wouldn't mean anything. Perhaps a quick read on the process of binning would enlighten you? As for overclockability, I suppose that if your chip was marketed as overclockable, then you'd have some sort of point.

Otherwise, all I understand, is that you "think" that the pci bus is limiting the overclockability of the chip. Perhaps there is another reason for this overclock discrepancy? Like the quality of the reverse MXM boards, or the production methods of the GPUs (sourced at different foundaries?).

Anyways, theories are just that - unproven ideas. Welcome to the club.

 

just like he said earlier..clocks don't mean anything...he was right..with out the bandwidth to support those clocks...you can't get any better numbers out of your gpu...
example:
those running the 920/940m. with out throttle stop, they would be useless as asus intended. why? because you can set all 4 cores to 25, but when those cores are maxed..you would drop down to 4 cores at 15. this is the same theory going on with "our" gpu. ramp the clocks up and you start losing performance or artifacting and glitching.

that's good for you. im happy for you.

you guys really need to get off this "per spec thing" that only applies if you "do not alter you laptop" once you alter it..those specs become guidelines

bottom line, no matter what i "think" a g73jh is slower at any gpu anything. and to speed it up..it needs a serious bios adjustment, not just a vbios adjustment. (my opinion only)
i already ran 10 different vbios's for testing and all that got was a more "stable to use gpu with no gains"
but hey, if your happy your happy. no harm no foul. im happy with every day use out of it. but gaming...yeah, i dont think so. got other machines for that. (but that's just me )

 

Damn johnksss you just make my day

 

sorry...
it's a great machine and all, but...
if it wasn't for throttle stop. i would have sold this already.
pcie is really not much of a concern anymore..since im now done benching this 5870M


side note:
i think a 5850M would choke in here as well. (personal opinion only)

 

so everyone who owns a g73jh that doesnt show 2.0 pci-e suck? becuase im part of that group

 

Complete opposite. For actual use and gaming and not just benchmarking low scores, the G73jh performance is identical to any other HD5870M laptop. Don't worry about it.

 

Everyone that has the issue of non "PCI-E 16@x16" listen to THIS:

I bought my G73Jh-A1 from Hong Kong on about 25 June 2010(820QM).
When I finally got myself in the forums I checked GPU-Z and had
PCI-E 16@x4 with PCIeSpeedTest_v0.2 of 850MB/sec both ways.
After updating BIOS 211/ vBIOS / CCC had the same results.
I was very pissed seeing other people having 16@16 with 2.5 GB/sec.
Couple of days ago I decided to try something (mentioned by Chastity in
an other topic)... so here it is::::::

-I removed the HDD/mem back plastic cover
-Unscrewed the only 2 visible screws holding the GPU on the M/B
-Swing the GPU card (slowly in out just a few millimetres)
-Blow air inside near the GPU slot (as much as possible) with a bottle
of Compressed Gas Duster (DYNEX from best buy).
-Put the screws/cover back and boot up.

The results are as you imagine (or not):
-GPU-Z ... PCI-E 16@x16 (still v1.1 - if that changed I would be GOD)
-PCIeSpeedTest_v0.2 of 3.0 GB/sec both ways.

 

shoot! i wish you knew which pins were not contacting...was thinking about dropping my card down to 8x. since we "should" be able to control the pci express link power state. another fail. (personal opinion only! - and this is in the bios - the parts you can't see.)

 

I have little problem with PCIeSpeedTest v0.2. At 536870912 bytes it stop responding and driver for GPU stop working. Where is the problem? Sometimes it appears when I play games too. I have 10.10e driver.

 

I have deleted posts & issued warnings about bickering, flaming, and trolling. This thread will be closed in short order if it cannot stay on topic with a technical discussion. Posting personal attacks and flamebait will get you banned.

 

and look what we have here...another 5.0 gb rated board.
http://img28.imageshack.us/img28/80/memcapture.png

 

Nice alienware gets 5gb/s sweet. I see it's running the pm55 chipset.

 

Doesn't mean anything actually unless the person actually benches it and I guarantee that the Alienware doesn't get 5gb/s both ways, probably not even close to it. I know HWINFO32 is constantly updating their info. I've asked two people with MSI with HD5870M and with PM55, and their results are exactly identical to G73jh. We've also had people with G73JW with 460M run the test and result identical also.

The 2 things that are important.

1). There is nothing Asus can do. This is a hardware chipset limitation on ALL mobile chipsets made by Intel.

2). This has NO effect, it has ZERO impact on your gaming and use of HD5870M as a gaming/consumer GPU. This only has an effect for those who buy FirePro or the HD5870 desktop/HD5970 specifically for GPGPU applications.

As is for most, this issue is absolutely ridiculous to get arms into air about. It's asinine. Adobe Premiere, CS5 etc all still work with your HD5870M. For encoding doesn't matter since AVIVO sucks and x264.exe still runs off your CPU, not GPU. And for gaming it has no impact. So overall this issue is only being whined about for no reason at all. The G73jh was not made for research purposes as that's where the bulk of the GPGPU applications are being used. As a gaming machine, this is not an issue at all.

If you don't believe me, you can do this.

Install Catalyst 10.8. Believe your PCI-e speeds will be around 700 gb/s. Bench a game like Crysis.

Then install Catalyst 10.11. Your PCI-e speeds should be around 3 gb/s. Bench the same game. I guarantee your results will be the same. This issue is just idiotic unless you have a professional grade FirePro or Quadro....

I've done the benchmarking now to show how stupid this issue is in terms of what the G73jh was built for and for why 99% of people buy this machine. Anyone who wants to do GPGPU work will buy a FirePro or more likely a desktop Quadro since OpenCL/AMD Stream is still in infancy and CUDA is the leading GPGPU tech atm. So even IF you did GPGPU work, you would NOT be using a G73jh.

Catalyst 10.8 649mb/s


Catalyst 10.11 3gb.s

 

Amen Brother.

 

fair enough.. as long as the drivers increase the speed as 10.11 did , idk... i do some professional work so it would help..

 

My M17x-R2 gets terrible numbers with Crossfire enabled but around 3 GB/s both ways when it's disabled using Catalyst 10.11:

 

pssss, don't see you testing that against a pm55 chipset laptop you own. . because that's the only way your test would be considered valid.

looked interesting and all...
but still it's a hm55 chipset. and pm55 is better.

yeah, that's what my desktop 5870 was doing as well..with the low cpu side of things...other than it being a driver issue, im pretty sure.

 

Here is some more direct link information from intel about the PM55 and HM55 chipsets. The PM55 has no Arendale support and has 8 PCIE 2.0 whereas the HM55 has Arendale support and 6 PCIE 2.0. Those are the only two known differences.

PCI-E speeds may be affected by the CPU you have. The i7 720QM and up have integrated the Northbridge. Traditionally the Chipset was split into two, the Northbridge and Southbridge. The i7 720QM has support for PCI-E integrated with the Northbridge integrated. This still has to communicate with the Southbridge, which is the HM55/PM55. My guess, there may be slight differentiation as it is scene the clock speed, multiplier etc on an i7 affects everything from core speed to memory controller etc.

Mobile Intel PM55 Express Chipset - Overview

Limitation on the PM55 for PCI-E is identical to HM55. In terms of performance for CPU and GPU, the HM55 and PM55 are IDENTICAL. There is no evidence from intel to say contrary.

http://www.intel.com/Assets/PDF/datasheet/322169.pdf The section on Platform Controller Hub will show that the PCI-E speeds are the same for ALL Mobile 5 Series chipsets!

The limitations of the HM55 affect the limitations of the PM55 also. The SSD issue is clear example!
http://forum.notebookreview.com/har...et-can-not-take-full-advantage-fast-ssds.html

Lastly in many respects, the HM55 is the better chipset than the PM55. PM55 has a few extra ports, that's it. PM55 is described as being the base chipset with the least feature for mobile 5 series. HM55 being the middle and HM57 being more feature rich. The Q series are for business class, corporate, workstation laptops.

The PM55 is based on the P55 Chipset. And magically behold! The P55 is the cheap economical chipset for desktops also! And it also limited at 2.5 gb/s. P55 can be described as being, unimpressive.
http://www.intel.com/products/desktop/chipsets/p55/p55-overview.htm



There is so far only one freak PCI-E benchmark I've seen from miahsoul. The PCI-E 2.0 stated by intel is 2.5 GB/S one way or 5 GB/s both ways. You can see what happened with miahsoul. If you add up his bandwidth is 5 GB/S. OMG OMG, SURPRISE!



djtamaki with Asus N61 using HM55 Chipset


iemie with G73jh, having very similar performance as 5150Joker on Alienware M17x, just switch it around


5150Joker Alienware M17x with CCC AI off, single HM5870M with PM55.



Ryzeki MSI GX660 with PM55 HD5870M



IKAS IV using MSI GX660 with PM55 Chipset, having identical issue with drivers as G73jh


LOUSYGREATWALLGM PCI-E running at v1.1 speeds. Alienware M15x with HD5870M


EGM92 also confirming v1.1 speeds on CLEVO with HD5870M
http://forum.notebookreview.com/6699047-post2037.html

 

Ruckus: I have to admit, you've been inspired about this topic.

 

I'd like to point out that the desktop 5870 is completely different from the 5870m. Keep that in mind.

 

to say the least!

my favorite quote..
Quit jumping on examples as if they are topics! (a favorite forum mistake)

this one seems to be the most interesting one.. he was using a clevo card in an alienware machine.

and mine:
oh wait! look! 6.4 gb!! yeah...i dont think so!