4820T vs 4820TG

[ngaw]

Thanks for the response. I was using cpu-z when I was running the synthetic benchmarks, but when I'm running full screen it's tough to check its readout.If anyone is interested, it seems Core Temp will write a log with temps and speed. I'll let everyone know what I come up with.

[ngaw]

Well, sadly I am seeing it throttle even just in normal gameplay - though it does seem to be rather rare. I found Everest to make the best logs, you can track CPU & GPU Temp, CPU/GPU Mhz and a lot more. I threw them into excel to make a few charts.

First, a very clear view of it happening under a synthetic stress test:

(I stopped the stress at ~4:05)


During gaming:

Both CPU & GPU throttle at 8:51 for about 25 seconds - the chart event coincided with a frame rate drop and a muddiness in my BT headphones. There also appears to be a GPU throttle event at 8:44.


I'll follow up on this after I get a few more FPS's installed. I'd be very interested if anyone is able to duplicate or not duplicate these results on any of the TG series. To get the logs, in everest open file->preferences and select logging under hardware monitoring. Even if you're nowhere near thermal limits the charts provide a very detailed look into how the laptop is performing heat wise - it would be neat to compare the different models and CPUs/GPUs.

Cheers,

[Acer.3820TG]

correct me if im wrong but isn't that throttling normal for the i3? it has 90C threshold.. so throttling at 85-87C looks right

[Bronsky]

yea ... looks right to me. You probably won't see this kind of throttle from an I5-450M. Guaranteed you won't see it from an I7-620M. The I7s run strong at 89C-90C with no hint of throttle. The I3 just runs out of head space. The I3s might need a cooler underneath to keep temps down.

I like the graphs. Can you just feed the data into excell automatically from the log or do you have to key it in?

Bronksy

[ngaw]

Dismissing this as obvious and not worth discussing seems a little premature at this point - wasn't it just a week or two ago that I was posting in this thread as the first one to document this behavior and explain it? I'm just trying to follow up on that and post info I wish was available when I was making a buying decision.

In those posts I said I didn't think it would happen in normal usage, and a few other people agreed. So the fact that it happens in normal every day game play is new. So is (As far as I know) that the gpu is throttling down from 550mhz to 100mhz under heavy load. It's not really a very good gaming machine at all if during the most intense gaming you suddenly find yourself with 40% of the cpu power and 20% of the gpu power.

The fact that m4cr0s reported he has gaming slowdowns with his i5 3820 without a known cause makes me wonder if it's truly limited to only an i3 and the 4820. As the GPU seems to be throttling independently, it might even be more of a problem on an i5 due to the extra heat it will be pushing out.

It seems to me like there is possibly a very real design issue here. All the CPUs share a 35W TDP yet the i3 will obviously be producing the least waste heat. My logs show I'm using half that power (16W-18W max with the IGP turned off) when the throttle kicks in. If I knew the i5 wouldn't suffer any of the same problems I might be tempted to return this for one - the irony is I specifically picked the slower CPU because I was worried about heat. But without any tests that's far from assured.

@bronsky - yup everest puts out the data in a CSV file that's perfect for importing into excel. I stumbled a bit before I could get a pretty chart out of it, but it's very easy once you learn how. I'd be happy to post an example xls if anyone wants.

[Bronsky]

Dismissing this as obvious and not worth discussing seems a little premature at this point - wasn't it just a week or two ago that I was posting in this thread as the first one to document this behavior and explain it? I'm just trying to follow up on that and post info I wish was available when I was making a buying decision.

In those posts I said I didn't think it would happen in normal usage, and a few other people agreed. So the fact that it happens in normal every day game play is new. So is (As far as I know) that the gpu is throttling down from 550mhz to 100mhz under heavy load. It's not really a very good gaming machine at all if during the most intense gaming you suddenly find yourself with 40% of the cpu power and 20% of the gpu power.

The fact that m4cr0s reported he has gaming slowdowns with his i5 3820 without a known cause makes me wonder if it's truly limited to only an i3 and the 4820. As the GPU seems to be throttling independently, it might even be more of a problem on an i5 due to the extra heat it will be pushing out.

It seems to me like there is possibly a very real design issue here. All the CPUs share a 35W TDP yet the i3 will obviously be producing the least waste heat. My logs show I'm using half that power (16W-18W max with the IGP turned off) when the throttle kicks in. If I knew the i5 wouldn't suffer any of the same problems I might be tempted to return this for one - the irony is I specifically picked the slower CPU because I was worried about heat. But without any tests that's far from assured.

@bronsky - yup everest puts out the data in a CSV file that's perfect for importing into excel. I stumbled a bit before I could get a pretty chart out of it, but it's very easy once you learn how. I'd be happy to post an example xls if anyone wants.

Mac sees throttling only in certain older games, not in the more demanding ones. I would think an across the board problem would appear in them first. Also, the I7's (merely a renamed I5) do not throttle until they get well in the high 90C range. This has been observed a number of times for the few I7s that were tested early on. I think Mac's problem is something different but it is merely a guess. The I3 is easier to understand when you look at its thermal properties. It's throttling is consistent.

When I get the chance, I have an I5-450M in my 3820TG and could run the test. I want to run my unit through a series of stress testing anyway to make sure everything is running well at the extremes, since I am putting an I7-620M in my unit soon.

I would love to see how to feed the data into excell. I think I have a copy somewhere around here.

Bronsky

[Bronsky]

Dismissing this as obvious and not worth discussing seems a little premature at this point - wasn't it just a week or two ago that I was posting in this thread as the first one to document this behavior and explain it? I'm just trying to follow up on that and post info I wish was available when I was making a buying decision.

In those posts I said I didn't think it would happen in normal usage, and a few other people agreed. So the fact that it happens in normal every day game play is new. So is (As far as I know) that the gpu is throttling down from 550mhz to 100mhz under heavy load. It's not really a very good gaming machine at all if during the most intense gaming you suddenly find yourself with 40% of the cpu power and 20% of the gpu power.

The fact that m4cr0s reported he has gaming slowdowns with his i5 3820 without a known cause makes me wonder if it's truly limited to only an i3 and the 4820. As the GPU seems to be throttling independently, it might even be more of a problem on an i5 due to the extra heat it will be pushing out.

It seems to me like there is possibly a very real design issue here. All the CPUs share a 35W TDP yet the i3 will obviously be producing the least waste heat. My logs show I'm using half that power (16W-18W max with the IGP turned off) when the throttle kicks in. If I knew the i5 wouldn't suffer any of the same problems I might be tempted to return this for one - the irony is I specifically picked the slower CPU because I was worried about heat. But without any tests that's far from assured.

@bronsky - yup everest puts out the data in a CSV file that's perfect for importing into excel. I stumbled a bit before I could get a pretty chart out of it, but it's very easy once you learn how. I'd be happy to post an example xls if anyone wants.

I must admit that I am surprised to see the I3 throttling in normal game play. With the temps that we were seeing initially, I doubted that the cores would reach the necessary 86C temps without serious streess testing. That, apparently, is not the case.

Mac sees throttling only in certain older games, not in the more demanding ones. I would think an across the board problem would appear in them first. Also, the I7's (merely a renamed I5) do not throttle until they get well in the high 90C range. This has been observed a number of times for the few I7s that were tested early on. I think Mac's problem is something different but it is merely a guess. The I3 is easier to understand when you look at its thermal properties. It's throttling is consistent.

When I get the chance, I have an I5-450M in my 3820TG and could run the test. I want to run my unit through a series of stress testing anyway to make sure everything is running well at the extremes, since I am putting an I7-620M in my unit soon.

I would love to see how to feed the data into excell. I think I have a copy somewhere around here.

Bronsky

[aylafan]

I must admit that I am surprised to see the I3 throttling in normal game play. With the temps that we were seeing initially, I doubted that the cores would reach the necessary 86C temps without serious streess testing. That, apparently, is not the case.

Mac sees throttling only in certain older games, not in the more demanding ones. I would think an across the board problem would appear in them first. Also, the I7's (merely a renamed I5) do not throttle until they get well in the high 90C range. This has been observed a number of times for the few I7s that were tested early on. I think Mac's problem is something different but it is merely a guess. The I3 is easier to understand when you look at its thermal properties. It's throttling is consistent.

When I get the chance, I have an I5-450M in my 3820TG and could run the test. I want to run my unit through a series of stress testing anyway to make sure everything is running well at the extremes, since I am putting an I7-620M in my unit soon.

I would love to see how to feed the data into excell. I think I have a copy somewhere around here.

Bronsky

Nice double post, Bronsky.

[ngaw]

I guess all I'm saying is no one expected this machine to act like this until I showed that it was. That i5 on youtube was up to 95c. As far as the old games thing, it certainly could fit - a game with lower GPU needs simply runs faster (higher FPS) - so the load on the GPU stays similar, but it actually takes more work from the CPU to keep up. So a less complex game probably does generate more heat.

I attached the excel files in zips that made those two charts. I'd be very happy to hear if it wasn't possible to duplicate my results.

EDIT: Apparently it is possible to undervolt a core i processor after all. While the CPU does do its own voltage and thermal management, apparently if you start with a lower voltage, it doesn't ramp up as high under load. I'm going to see if I can find a tool that will do that without support in the bios. If you can, in theory that'd also extend battery life somewhat. ( link 1 link 2 )

[ngaw]

Even with the previous results, I was surprised to see how common it was in left 4 dead 2: