New client! v2.8012.466

[crYnOid]

At the client page of course I have just downloaded it, and will give it a try.

Anyone have info on what it speeds up?

 

[micron]

I've been running it for quit a while.

 

[crYnOid]

ahh, i just don't remember it being mentioned (stupid brain). But what does it speed up?

 

[ViRGE]

Faster OGR core, some display changes.

 

[Nohr]

Here's the changes.txt file:

new: x86: most unix-ish clients now support the RC5/SMC core.
imp: some: even faster OGR core.
chg: all: reverted default crunch-meter for ogr to 'absolute' type.
imp: win32 cui: tweaked for considerably improved display speed.
fix: win32 cui: redirection to file works as it did prior to .465.
chg: all: the list of available cores is a superset of all cores
available for a particular architecture, some of which may
not be available for a particular client/OS/hardware flavour.
fix: riscos: time and timezone oddities.
fix: win32: workaround for time bug on some win2k HALs.
fix: all: suffices weren't lowercase for remote buffer filenames.
fix: win32: client -install'ed to run hidden appeared on desktop

 

[crYnOid]

it doesn't appear to make the P6 chips any faster at ogr. Next up is K6 chip testing

 

[Nohr]

My K6-III+ 550 didn't get a faster OGR benchmark.

 

[Viztech]

I'm still using the 463b client on my K6s, as there was a slight loss with the 464g where the P6 OGR improvements were implemented.

What kind of OGR benches are you getting with the 466 anyway?

viz

 

[crYnOid]

i just benchmarked my K6-3 450 and it report about the same speed as 465 (which is slower than 463). So no obvious changes in performance yet

 

[Nohr]

I'm getting about 3.286Mn/s on the K63+ 550, which is pretty much the exact same thing I got with 465.

(edit: M*n*/s, not Mk/s... I live in a RC5 world)

 

[crYnOid]

With 463 i am getting 3,005,000nodes/sec for a K6-3 450. With the 466 client on the same cpu i was getting 2,650,000nodes/sec

 

[Nohr]

With 463b I'm getting 3.616Mn/s. I think I'll be sticking to 463b for now..

 

[Nohr]

On my K62 350@330 (crappy motherboard), I get around 1.985Mn/s with 466. It does 1.941Mn/s with 463b.

 

[Fardringle]

My K6-3+ 600Mhz does 4.0 Mnodes/sec with 463 and about 3.5 with 466...

(Just to add my 'contribution' to this discussion...)

 

[Fingers]

anyone got a P4 that they want to try out the new client on, maybe it helps take advantage of the P4 optimizations.

 

[Moose]

The 466 client does not have P4 optimizations in it (Yet.) We do have a P4 asm core in the works. It now passes all 32 tests but only performs slightly better (50kkeys/sec more on a P4 1.5GHz) than the previous cores. We prob won't release it until it is optimized more.

moose

 

[ViRGE]

Does anyone have any Athlon benches? I have the CPU, but no previous numbers to compare it to.

PS Thanks for working on a P4 core Moose

 

[BGod]

ViRGE, AMD Thunderbird 1.2GHz @ 1.32GHz on an MSI K7T Pro2-a running the Beta BIOS.
dnetc v2.8012-465-GTR-01010823 for Win32 (Windows 4.10).
[Feb 13 05:38:24 UTC] RC5: Summary: 263 packets (1536.00 stats units)
1.00:43:36.37 - [4,628,080 keys/s]

I'll post again after I've let the new version crunch for a while.

dnetc v2.8012-466-GTR-01020321 for Win32 (Windows 4.10).
[Feb 13 21:15:35 UTC] RC5: Summary: 163 packets (929.00 stats units)
0.14:58:56.22 - [4,610,868 keys/s]

 

[TwoFace]

It had the same speed as 465 and 463b on my amd setups, BUT I found a flaw (not sure if it's my system or not) after a while of crunching the 466 client loses about 12% of the cpu cycles, and I assume it does so in regular intervals since I didn't catch this until after work yesterday, my pproxy then hadn't seen that cow for 8 hours I fired it up again and let it run for a while, and that's when I noticed the drop in processor usage... Never seen this before, and the 465 client doesn't behave that way at the same system, altho' it has that crunching rate problem :frown: I don't have the 465a (or what it was that fixed the rate prob) but I don't really look much at the display so it doesn't matter much to me...

Got any thoughts on what the problem I have with the 466 might be moose or anyone? I would post a bug report at the DNet pages if I only knew how, never been able to figure that one out

With love and respect your fellow TA member

Two-Face
My stats:
RC5
OGR-25
Seti
Gamma Flux

 

[Eug]

NE1 have the same glitch I did?

I've just done a fresh Windows 2000 install so it could be something else, but I suspect not. Today, my computer slowed to an absolute crawl, and it seemed that the dnet 466 client was hogging everything. My screen and all my tray icons were messed up, except the dnet client. Had to hard reset and lose the current OGR stub.

Anyways in benches it's about 13% faster in OGR compared to 463b, but it is less than 0.5% faster than the 464 client. So I'm sticking with 464 for the time being.

Besides the browser (MS IE 5.5), the only other software I have running in the background is the Dell Digital Audio Receiver MP3 server software, which is also new, but it wasn't actually serving anything at the time (and it was running fine yesterday with the 464 client. In fact, I have no other software installed besides WinZIP and WinRAR - like I said it's a fresh Windows 2000 install (SP1, with most recent critical updates).

This machine is 100% stable. Heavily tested in various ways, and it had run dnet fine for weeks at a time without a reboot. Celeron 533A@880 - runs stable over 920 MHz by the way.

 

[Polo]

On a fresh W95 setup (I've build a P133, 32Mb for a friend), V 466 is unable to fetch/flush....
Back to v463 and no probs.

 

[crYnOid]

I have found that 466 is slightly unstable (crashing while running as a service). I have gone back to 465.

 

[Dutchman2000]

in case anybody missed it:

Moose's post on the client

[edit] spelling [/edit]

 

[Eug]

Thx. Will do immediately.