Qualcomm Krait & MSM8960 @ AnandTech

The severe performance degradation with Extreme settings could be bandwidth-related, or it could be related to thermal throttling at these more demanding graphical settings.
 
Considering that according to Anand the HTC One Extreme scores are: 6190 / 5942 / 7246 and the Exynos5410 yields accroding to that site 5985 / 5538 / 8341 I don't think so. After all it's faster as a smartphone than Samsung's own current flagship tablet.

I'm with ams though, Anand should seriously look into his results again. It cannot be that the HTC One gets 11520 in default in the Physics demo and in extreme 7246 points unless the GPUs are playing a bigger role in the demo than the demo description claims itself Anandtech is quoting in that writeup.

Even worse a quad A15@1.6GHz in the 5410 is supposed to get similar performance to a dual A15 in the Nexus10? It's a physics demo running almost entirely on the CPU rendering 4 simultaneous worlds and in the latter case you have only 2 cores = 2 threads. Unless the 5410 for some dumb reason is running that demo on its A7 CPU cluster or is severely throttling frequencies while running on the A15 cluster the Anand results don't make sense.

So before anyone jumps to preliminary conclusions I would suggest that it would be wiser to wait until the dust settles, IHVs can optimize their drivers for current new benchmarks and websites get accustomed with the underlying sw, platforms and their behaviour before they rush out a bunch of meaningless and partially absurd results.
 
Turbotab said:
Another Exynos 5410 3D Mark result, with more feasible physics scores.

5410 - Physics
Default - 10809

Extreme - 10866

The HTC One results appear to be purloined directly from Ananadtech.

http://www.max-up.ru/news-apple/3dmark-sgs4-exynos-vs-htc-one.html
Click to expand...

Sounds more realistic so far since Ice Storm is IMHO give or take on GLB2.5 complexity as a benchmark. T-Rex/GLB2.7 is really pushing the envelope for graphics and there I truly don't expect the 544MP3 in the S4 to do particularly well.
 
Yes that makes more sense...we await the gs4with s600 numbers to make a proper comparison.

Certainly those a15s equiped with ample bandwidth put the hammer down on that physics test.
Batterylife is also gong to be very interesting...which approach is betterfor mobile?
 
For those that have not read anandtechs full review of the htc one..you should its very good (brian klugg)

In that article there is a very good breakdown of of krait 300 ipc improvements over krait 200...pretty outstanding if you ask me.

In some cases we see massive gains..in others moderate, typical qualcomm to be as realistic as possible with their performance forcasts as they state "15%" ipc increase...thats almost an understatement, nvidia really could learn alot from qualcomm about PR.

http://www.anandtech.com/show/6747/htc-one-review/11

http://www.anandtech.com/show/6747/htc-one-review/12

Edit..question. .does anyone know whether lpddr3 has higher latency than its predecessor?
 
Last edited by a moderator:
Helmore said:
Not sure, but I expect latency to be similar and yes, it's an excellent review.
Click to expand...

Well im not sure because in some benchmarks the galaxy s4 s600 version loses out on a few benchmarks where in reality, with its higher clock speed, lpddr3 and android 4.2.2 it should be easily the fastest.

The nexus 4 beats it in gl benchmark 2.5 & 2.7, the lower clocked and lower bandwidthed htc one beats it in others.

The only thing that I can think of is memory latency? Odd.
http://m.ibtimes.co.uk/samsung-galaxys4-benchmarks-fastest-nexus4-htc-one-454279.html
 
french toast said:
Well im not sure because in some benchmarks the galaxy s4 s600 version loses out on a few benchmarks where in reality, with its higher clock speed, lpddr3 and android 4.2.2 it should be easily the fastest.

The nexus 4 beats it in gl benchmark 2.5 & 2.7, the lower clocked and lower bandwidthed htc one beats it in others.

The only thing that I can think of is memory latency? Odd.
http://m.ibtimes.co.uk/samsung-galaxys4-benchmarks-fastest-nexus4-htc-one-454279.html
Click to expand...

Are you sure they aren't quoting onscreen vsynced results and to that at varying resolutions?

Nexus4:
GLB2.5 offscreen 1080p = 35.6 fps
GLB2.5 on screen 720p = 46.5 fps
GLB2.7 offscreen 1080p = 14.2 fps
GLB2.7 on screen 720p = 21.8 fps

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro27&D=Google+Nexus+4

GalaxyS4:
GLB2.5 offscreen 1080p = 41.9 fps

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Samsung+SHV-E300S&testgroup=overall
 
Ailuros said:
Are you sure they aren't quoting onscreen vsynced results and to that at varying resolutions?

Nexus4:
GLB2.5 offscreen 1080p = 35.6 fps
GLB2.5 on screen 720p = 46.5 fps
GLB2.7 offscreen 1080p = 14.2 fps
GLB2.7 on screen 720p = 21.8 fps

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro27&D=Google+Nexus+4

GalaxyS4:
GLB2.5 offscreen 1080p = 41.9 fps

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Samsung+SHV-E300S&testgroup=overall
Click to expand...

Perhaps this is the case.
However it still doesn't explain vellamo html5 and metal...quadrant is very close also when gs4 should have an outright lead in every area. .except browser which is more software optimisation.
 
For those whom didn't know, you can easily determine the SoC binning of a Snapdragon 600, by values found in the last_kmsg.

Depending on the PVS value, a different voltage table is loaded upon boot, the actual PVS value is determined, I imagine by blown Qfuses.

On the HTC One / Snapdragon 600, the PVS is listing numerically, not by name, as per the S4 Pro. All S4 Pro that I've seen listed are marked as SPEED BIN 0, whereas all Snapdragon 600s are listed (so far) as SPEED BIN 1.

Thanks to recent HTC One Kernel Source release, I can can confirm that a Snapdragon 600 can have 1 of 7 possible PVS values: A higher PVS value is better, i.e the SoC requires less voltage, at a given frequency.

0 / 1 / 2 / 3 / 4 / 5 / 6

So far, we have only see PVS values of # 2 / 3 / 4 / 5 / 6 in shipping devices. For comparison a PVS #2 vs a #5

PVS 2

static struct acpu_level tbl_PVS2_1700MHz[] __initdata = {
{ 1, { 384000, PLL_8, 0, 0x00 }, L2(0), 925000 },
{ 1, { 486000, HFPLL, 2, 0x24 }, L2(5), 925000 },
{ 1, { 594000, HFPLL, 1, 0x16 }, L2(5), 925000 },
{ 1, { 702000, HFPLL, 1, 0x1A }, L2(5), 925000 },
{ 1, { 810000, HFPLL, 1, 0x1E }, L2(5), 937500 },
{ 1, { 918000, HFPLL, 1, 0x22 }, L2(5), 950000 },
{ 1, { 1026000, HFPLL, 1, 0x26 }, L2(5), 975000 },
{ 1, { 1134000, HFPLL, 1, 0x2A }, L2(14), 1000000 },
{ 1, { 1242000, HFPLL, 1, 0x2E }, L2(14), 1012500 },
{ 1, { 1350000, HFPLL, 1, 0x32 }, L2(14), 1037500 },
{ 1, { 1458000, HFPLL, 1, 0x36 }, L2(14), 1075000 },
{ 1, { 1566000, HFPLL, 1, 0x3A }, L2(14), 1100000 },
{ 1, { 1674000, HFPLL, 1, 0x3E }, L2(14), 1137500 },
{ 1, { 1728000, HFPLL, 1, 0x40 }, L2(14), 1162500 },
{ 0, { 0 } }

PVS 5
static struct acpu_level tbl_PVS5_1700MHz[] __initdata = {
{ 1, { 384000, PLL_8, 0, 0x00 }, L2(0), 875000 },
{ 1, { 486000, HFPLL, 2, 0x24 }, L2(5), 875000 },
{ 1, { 594000, HFPLL, 1, 0x16 }, L2(5), 875000 },
{ 1, { 702000, HFPLL, 1, 0x1A }, L2(5), 875000 },
{ 1, { 810000, HFPLL, 1, 0x1E }, L2(5), 887500 },
{ 1, { 918000, HFPLL, 1, 0x22 }, L2(5), 900000 },
{ 1, { 1026000, HFPLL, 1, 0x26 }, L2(5), 925000 },
{ 1, { 1134000, HFPLL, 1, 0x2A }, L2(14), 937500 },
{ 1, { 1242000, HFPLL, 1, 0x2E }, L2(14), 950000 },
{ 1, { 1350000, HFPLL, 1, 0x32 }, L2(14), 962500 },
{ 1, { 1458000, HFPLL, 1, 0x36 }, L2(14), 987500 },
{ 1, { 1566000, HFPLL, 1, 0x3A }, L2(14), 1012500 },
{ 1, { 1674000, HFPLL, 1, 0x3E }, L2(14), 1050000 },
{ 1, { 1728000, HFPLL, 1, 0x40 }, L2(14), 1075000 },
{ 0, { 0 } }
};
 
PVS5 uses the same voltage at 1.73GHz as a PVS2 at 1.46GHz.
It's a large difference.

I guess the Galaxy S4 will probably use the higher PVS numbers, since it'll clock up to 1.9GHz.

Can't wait to see how "lucky" I was when I get my HTC One Dev. Edition.
 
So you're saying you can have the same phone with different PVS value? I.E. your HTC One can have a different PVS value than your friend's HTC One. How much of a difference in battery life can this have? Will a phone with a PVS 5 setting have 5-10% more battery life than a phone with PVS 1, all else being the same?
 
Helmore said:
So you're saying you can have the same phone with different PVS value? I.E. your HTC One can have a different PVS value than your friend's HTC One. How much of a difference in battery life can this have? Will a phone with a PVS 5 setting have 5-10% more battery life than a phone with PVS 1, all else being the same?
Click to expand...
This is correct, you can get different battery life on the same phone model just based on the binning of the chip.

Samsung's binning is even more complicated; ASV and PVS are the same thing, just that they also change adaptive body bias voltages at the same time, I explained this once here.
 
Helmore said:
Runs cool? Isn't 43 degrees Celsius pretty warm? And as Nebuchadnezzar said, this isn't quite the right thread...
Click to expand...

Obviously I thought it was s600..im on mobile so I cant translate. .

Edit: Im pretty sure some on here would know this already if I had missed it being discussed previously. .but incredibly Canadian gs4 (s600) edges in front of ipad 4 in gl benchmark 2.7.
http://gfxbench.com/result.jsp

Bodes well for s800, we are spoilt as tech geeks :)
 
Last edited by a moderator:
You must log in or register to reply here.

Similar threads

W
New Qualcomm silicon for new Surface May 2024
4 5 6
Replies
102
Views
19K
MfA
M
Kaotik
3dfx Voodoo4 M4800 mobile
Replies
3
Views
363
Color me Dan
C
L
Dynamic register allocation in GPUs
2
Replies
24
Views
3K
dr_ribit
D
D
Performance scaling on big GPUs discussion ...
Replies
15
Views
2K
Rys
Rys
Q
Feasibility of a 1000W console
2
Replies
23
Views
1K
Cyan
Cyan
Back
Top