Joined: 07 Nov 2008 Posts: 142 Location: Spring Hill, Fl
Posted: Sat Oct 22, 2022 2:21 Post subject: nand for r7800
hiya peeps, anyone got any idea where I could order the flash for an r7800
micron MT29F1G08ABBEAH4:E
the router bit the dust and I got alot of free time so I was gonna try and salvage it.
thanx in advance _________________ modem: arris tm1602
router: r7800 voxel 1.0.2.77sf w/ kamoj 5.3b12 addon
ap: wrt1900ac v1 bs 43136,dir-825 b1 bs41117
NAS: Iomega ix2-200 CE 4tb, Seagate 1tb usb drive
Wash your hands and DONT TOUCH YOUR FACE!!!
Joined: 05 Oct 2008 Posts: 666 Location: Helsinki, Finland / nr. Alkmaar, Netherlands
Posted: Sat Oct 22, 2022 8:52 Post subject:
How many f/w flashing is it expected to last? Any educated guesses?
I routinely flash every new build, but sometimes I think that I maybe shouldn't.
If my R7800's flash memory wears out, I would have a problem, since they're not being made any longer, even though it's the best router I have ever owned (I have one at two different locations).
My first D-link blew itself up ... Oh, well my Buffalo WZR-HP-G300NH was a winner, too.
I would buy some used hardware when good enough deals appear, otherwise brand new XR500 are still being sold.
XR500 has history of bad blocks out of the factory, so the chicken and the egg, need to flash to discover then RMA.
Maybe that issue is overblown. I have flashed EA8500 hundreds of times. When Wi-Fi 7 is available migrate to x86.
Unless by some miracle a new Qualcomm router becomes available with ability to be supported, kernel 5.4 or later.
Joined: 05 Oct 2008 Posts: 666 Location: Helsinki, Finland / nr. Alkmaar, Netherlands
Posted: Sat Oct 22, 2022 11:20 Post subject:
Apart from having bad chips the XR500 is also awfully ugly, and overspecified for the speeds I get or need. Plus very expensive (€312 in Finland).
I cannot place an XR500 in plain sight where my wife is living, too. Even the R7800 required convincing to be accepted. There's enough industrial designing in an unobtrusive plain white or black rectangular box with 4 antenna's ....
Joined: 31 Jul 2021 Posts: 2146 Location: All over YOUR webs
Posted: Sat Oct 22, 2022 11:31 Post subject:
I know that a particular developer device has been flashed over 2000 times and still chugging along like a champ.
Ive flashed mine hundreds of times even sometimes up to 4 times a day when there was that much development going on I needed to test my bits. Currently the rate of flashing is never less than once a week and at the most 6 times a week.
Some vendors however may still carry parts because units that go for RMA are sent to authorized repair places who must have parts for most common issues, so hard to find some parts, sure YMMV but impossible I wouldn't stake any claims on that.
The Chinese make plenty of crap for spares, probably easiest place to get them for reputable retailers as many sell clones which wouldn't be so bad except the quality mostly is so terrible (probably factory reject into mix) that simply dont work.
...XR500 is also awfully ugly, and overspecified for the speeds I get or need. Plus very expensive...
Well you do realize the R7800 has about the same specs as XR500 IPQ8065 2x QCA9984?
Seems as if "buy some used hardware when good enough deals appear" was overlooked,
instead to focus on the ugly XR500. Overspecced? Please see a used R7500v2 or EA8500.
Can't help with wife, aside from the solutions Ethernet inside walls + ceiling access points.
Because that's a route I would take, x86 + dedicated access points integrated into rooms.
Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Sat Oct 22, 2022 16:00 Post subject:
xr500 isnt overspecd, its under specd, just like the r7800. and r7500v2/ea8500 is even more underspecd, for cpu speed. on kong's builds they were all mostly fine, but since the newer (at the time anyway, we falling behind on kernels again) bs build kernel ruined krait (ipq806x) throughput and latency, even local performance. the qca9984 and even 988x have more potential in them when not limited by an ARM cpu, and especially broken or misconfigured kernels. when i made the switch to x86, mostly for the reason stated above, i got about another 70mbps out of the qca9984.. with better latency.
running the grc shields up dns spoof test on r7800 with a unbound track/ad block list of around 550k lines made unbound max one core at 100%.. with x86 at 1.74m lines now its 0.3%. _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers
Joined: 05 Oct 2008 Posts: 666 Location: Helsinki, Finland / nr. Alkmaar, Netherlands
Posted: Sat Oct 22, 2022 19:25 Post subject:
Eventually my R7800's may fail. I'd like to be prepared strategy wise for such an event, so as to not end up in a fix and then possibly a victim to marketing hype. SMD soldering is not for everybody.
In my experience the R7800 is very well supported by DD-WRT and I'd like to stay on that track.
In hindsight, I do recall reading about the XR500 being a rebranded R7800. With an additional random flaw of bad memory.
I looked up its price which is much higher than the R7800 was, which made me assume the XR500 must therefore somehow be initially a better engineered product until the memory flaw was added or crept in. Marketing hype 'for gamers' serves to make me suspicious. Marketing is usually lying or very close to it.
What use is the passing of time (years), if it is not used to make a good product like the R7800 into a *better* product ...
Well never mind, let's forget about the XR500 and other space ships or 'transformers' ...
In a very useful comment Tatsuya advocated switching to an x86 processor so it can keep up with the communication chips better than an R7800, and others confirmed the lack of CPU power in the R7800.
So then, what current router has an x86 (and communication chips like the R7800 has) or other CPU that can keep up better, therefore earning consideration as a replacement for the R7800?
It would have to include a reasonable level of DD-WRT support.
Joined: 31 Jul 2021 Posts: 2146 Location: All over YOUR webs
Posted: Sat Oct 22, 2022 19:29 Post subject:
x86 or x64 have all the features dd-wrt supports, activation (paid) maybe needed in some cases for full support, but nothing is guaranteed bug free so ymmv.
Test out the free images (free or not paid) on a vm before you make a final plunge.
Joined: 05 Oct 2008 Posts: 666 Location: Helsinki, Finland / nr. Alkmaar, Netherlands
Posted: Sat Oct 22, 2022 20:55 Post subject:
the-joker wrote:
x86 or x64 have all the features dd-wrt supports, activation (paid) maybe needed in some cases for full support, but nothing is guaranteed bug free so ymmv.
Test out the free images (free or not paid) on a vm before you make a final plunge.
Search engines on how to are your friend.
Do you really mean a 'PC' is used as router .... ?
Lately I have been thinking about eventually replacing my HP Probook laptop with a NUC(?), because my laptop's keyboard is failing, and I don't take my laptop out on trips anyway, but this is a totally new concept, for me at least. PC's come with one Ethernet port and it looks like an NUC has no space for extra network cards.
It would be a very expensive router ...
Let's see what searching the internet turns up ...
xr500 isnt overspecd, its under specd, just like the r7800. and r7500v2/ea8500 is even more underspecd, for cpu speed.
ho1Aetoo wrote:
My talk, the small processor can not utilize the radios.
Disagree, sorry. Hardware is not to blame just see results of older DD-WRT builds, or OpenWrtno software offloading.
tatsuya46 wrote:
on kong's builds they were all mostly fine, but since the newer (at the time anyway, we falling behind on kernels again) bs build kernel ruined krait (ipq806x) throughput and latency, even local performance. the qca9984 and even 988x have more potential in them when not limited by an ARM cpu, and especially broken or misconfigured kernels.
ho1Aetoo wrote:
With 4 spatial streams the CPU capitulates and the latency is high.
Agree, IPQ806x poor performance under heavy load ignores previous achievements. Also ignoringdormanthardware.
ArjenR49 wrote:
With an additional random flaw of bad memory.
It is luck of the draw, e.g. R7800 new in box bad blocks. Alreadymentionedx86 twice build own or a custom mini PC.
ArjenR49 wrote:
Let's see what searching the internet turns up ...
"Unless by some miracle a new Qualcomm router becomes available with ability to be supported, kernel 5.4 or later."
Full circle to the future if no new router hardware is to be supported then we will be forced to x86+access points etc.
Disagree, sorry. Hardware is not to blame just see results of older DD-WRT builds, or OpenWrt no software offloading.
last post from your link of 2019
Quote:
I just updated my EA8500 to OpenWrt 18.06.4, and I can confirm that the ondemand tuning above still helps. My iperf3 results LAN-to-WAN would bounce between 680Mbps and 400Mbps with the default settings, but are consistently 680Mbps as changed above.
...
top reports about 77% sirq and 14% idle while testing NAT throughput.
insane, just open the WebIF then it is 100% load
I have already tested all possible builds.
first of all, older dd-wrt or openwrt do not have better wifi performance (even the old dd-wrt kong builds have crappy wifi throughput).
They reach a maximum of 600Mbit - miles away from 1 Gbit.
The best BS builds reach 850-900Mbit (TX) but only with large buffers (netdev_max_backlog=2000) and have 100% CPU load and very high latency.
Only the stock firmware and some OpenWRT builds with NSS acceleration manage to use the radios properly (but the CPU load was still very high - still 70%).
secondly, WLAN is not the only thing that generates 100% CPU load.
If I use QoS with Cake and start 32 downloads and 32 uploads the CPU also has 100% load and that with ~150Mbit throughput.
tatsuya had written that even with a long blocklist and unbound he had very high CPU load etc.
my x86 has maximum 10% load with 930Mbit WLAN throughput
It is also pointless to discuss about old builds with old kernel versions.
Even if the performance would have been better, several years old builds with 100 vulnerabilities are no alternative and we won't get the performance back.
It is also pointless to discuss about NSS support, we don't have it with dd-wrt and we won't get it with the kernel version we use.
And the accelerators have not only advantages but also disadvantages like all other technologies like SFE/CTF/FA etc.
It's a nice gimmick - but can't replace real CPU power.
That quote I did not highlight its from another post see above. Not everyone knows how to properly configure or test
WAN to LAN with two PCs to avoid loading down the router with iperf3. EA8500 800/200 50% cpu load maybe lucky?
Too good to believe? See second link 870-900 Mbits/sec 50-65% sirq not bad for a poor underspec IPQ8064 1.4 GHz.
I am aware you have tested many builds, but also tatsuya46! Seems focus is primarily wifi performance which is fine.
Of course we understand many things compete for resources. Certainly x86 always wins, with reasonable CPU choice.
2019 OpenWrt kernel 4? Old builds pointless? Performance vanishes, just happens, nothing to see here, move along!
Certainly security reasons prevent us from using these builds today, but to ignore this performance as forever lost...
I believe NSS driver has reached point of stability, but needs more testing and so worth following along the progress.
Yes acceleration weighs pros and cons, but NSS cores are not only capable to accelerate NAT see various other uses.
But one thing for certain it is no gimmick to free up the CPU, really nonsense considering progress made and results.
Getting sidetracked, we are talking about failing hardware. Our current options are to buyusedrouters, or gotox86.
New router hardware over next year will be kernel 5.4, so sooner than may be comfortable we are facing this reality.
Also, would be nice to be running an x86 kernel that natively supports new intel NICs i225, i226 among other things.
Last edited by blkt on Sun Oct 23, 2022 10:39; edited 1 time in total