From the first chart, A53 relative score ~0.8, A57 1.45. Both taken from overall score. From the 2nd chart, A72 crypto relative score 1.38.
Back of the envelope estimate, from A53 to A72 = 1.45/0.8 * 1.38 = 2.5. If A53 = 1, A72 = 2.5 times performance on crypto at the same clock frequency.
2.5 times the performance is hard to ignore if software and test done correctly.
You wrote something else above and I replied to that
From A72 to A73, about 30% overall performance boost.
I only wrote that this is unlikely to be the case - but you can still dream
You digressed. Hence, I brought you back to the core discussion A72 vs A53.
Without the extra boost from A73 or regardless marketeers calling it 'efficient' or 'performance' core, doesn't matter because they don't swing the needle a bit.
Hey there, a question:
Excluding the WIFI 7 NIC board does BPI R4 have no wireless at all?
I read on the wiki that there is MediaTek MT7975N 2.4GHz + MediaTek MT7977B 5GHz + MediaTek MT7977A 6GHz wifi on board but on Daniel's commit it only talks about optional card:
WiFi: optional MediaTek MT7996 Wi-Fi 7 module (using 2x PCIe gen3 x2 on the mPCIe slots and 12V power)
The info on OpenWrt's wiki is a bit outdated & stale. Try the vendor wiki by google "bpi r4 spec". In a nutshell, no WiFi included in R4. WiFi is available through official WiFi daughter board (optional purchase) or your own 3rd-party WiFi cards.
The official case is available in store. Yay.
Link to official store item: https://it.aliexpress.com/item/1005006860207712.html
The listed price is slightly above half that of the 3rd-party pre-order. So another win if you hold up until now.
The wifi 7 card is available already? Otherwise, could you recommend a wifi 6e card that will work with hw offloading?
I think 6-antenna daughter board is being finalised for sale. I've seen sample boards on BPI forum. Any 3rd-party with mini-PCIe socket and linux driver support should work. My plan is to use R4 as a wired router + low-power Internet server. I haven't paid much attention on WiFi. Btw, all these info are available on vendor wiki you should really look it up and read it through.
--
A general advice to new comers on R4. Read the first post on this thread. Lots of good information there. Read the vendor's Wiki (I think linked in the first post) that has most complete information about R4. Use google search to mine this thread. Use google search to mine BPI forum. This perhaps could get your answers lots faster, and also avoid low-effort questions that have been asked and answered before.
What's the guess, will passive cooling work with fiber sfp+ and 4 gbit ethernet but without wifi in the official case? Would power it via USB C
Am I correct that the asiarf AW7916-AED Wi-Fi 6E AX3000 M.2 AE Key Module is not plug and play compatible with the BPI-R4 (because key IDs A and E on the asiarf device do not match up with the key IDs on the BPI-R4 which are KEY-B and KEY-M?
I believe it'll be fine. If you start from a clean sheet state like me, you may want to get a heatsink /w fan. Give yourself more headroom and flexibility.
I'm looking a CPU heatsink /w 40x40x10 standard fan. Since the space above the SoC is pretty much dead space, I'm looking one that's larger than 40x40, more aluminium mass/fins meaning better heat dissipation. Too bad the official R3 heatsink won't fit without modding.
The two slots for WiFI cards are *mini-PCIe sockets. Any version of m.2 will require an adaptor. Pay attention to the clearing space from the case bottom if you go this route.
As an aside, the advantages of R4 is it comes with many expansion sockets. As long as you know bandwidth and power requirements for your add-in card and the spec of physical connector, you can plug anything into R4 just like a general PC. As long as there is a linux driver for your add-in card. Let's hope more and more router boards will go this direction like R4.
I am starting with a clean slate but the idea of a fan in a home router is frankly revolting...
Guess I will order board and case and see...
Why? If the fan is PWM controlled, it will only wake up in case of sensor showing hot hardware - and hot hardware means danger of wearing out. Therefore having a fan is a pretty good thing in my eyes.
I totally agree that a fan that is running 24/7 is nothing I would want to operate in m
For my first router with fan - a WRT1900AC - until now (in ~8 years of operation!) I never noticed it waking up during normal operation (the fan does wake up for a couple of seconds everytime I restart the device).
The better microchips are cooled, the longer they live.
Most of those tiny fans fail way before the silicon ever does... And more often than not, they spin constantly, if perhaps more slowly when there is little heat.
I prefer there exists heat dissipation through metal casing (like those NanoPis), however the current metal case isn't.
My review is finished, enjoy!
Sadly I couldn't set Jumbo frames for the R4. I received this patch from Sinovoip. Maybe someone wants to integrate it into OpenWrt. This is a patch for their image, it might need adjustment for official OpenWrt.
diff --git a/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 37b6c68..c4751ad 100755
--- a/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -662,6 +662,8 @@
mcr &= ~(XMAC_MCR_TRX_DISABLE);
mtk_w32(mac->hw, mcr, MTK_XMAC_MCR(mac->id));
+
+ mtk_w32(mac->hw, XMAC_JUMBO_FRAME_9K, MTK_XMAC_RX_CFG2(mac->id));
}
}
@@ -3971,6 +3973,7 @@
eth->netdev[id]->irq = eth->irq[0];
eth->netdev[id]->dev.of_node = np;
+ eth->netdev[id]->max_mtu = MTK_MAX_RX_LENGTH - MTK_RX_ETH_HLEN;
return 0;
diff --git a/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index 39543c7..11ad689 100755
--- a/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -17,7 +17,7 @@
#include <linux/phylink.h>
#define MTK_QDMA_PAGE_SIZE 2048
-#define MTK_MAX_RX_LENGTH 1536
+#define MTK_MAX_RX_LENGTH 9252
#define MTK_DMA_SIZE 2048
#define MTK_NAPI_WEIGHT 256
@@ -658,6 +658,10 @@
#define XMAC_MCR_FORCE_TX_FC BIT(5)
#define XMAC_MCR_FORCE_RX_FC BIT(4)
+/* XFI Mac Rx configuration registers */
+#define MTK_XMAC_RX_CFG2(x) (MTK_XMAC_MCR(x) + 0xd0)
+#define XMAC_JUMBO_FRAME_9K 9216
+
/* Mac status registers */
#define MTK_MAC_MSR(x) (0x10108 + (x * 0x100))
#define MAC_MSR_EEE1G BIT(7)
,
I still believe there is something very odd going on with those benchmarks. The a73 cores should be much faster than the age old a53 in the R3. Thermal issues or CPU governor maybe?
Are the N100 N305 boards still the best x86 alternative to this? I'm not aware of any N100/N305 flavor that has m.2 unrestricted for a 5G card such as RM520N or RM521F.
We can switch to larger ram by following this:
Just put the latest snapshot w/6.1.86 kernel, I also stick a small piece of heatsink on top of the chip, running fairly stable with 1GbE line (I have faster one but don't want to affect the whole network so I picked the easiest one first).
With USB power (I run mine with Anker USB-C PD power), a very simple Ookla speed test showing 912Mbps download + 939Mbps upload, system load never go over 0.2, and power consumption is between 2-3W, pretty impressive.