HP T730 8GB RAM 32GB Storage Thin Client

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
I made a bootable EFI shell image for updating the BIOS without dicking about with ThinPro:
T730_BIOS_01.11_RevA_bootable_EFI.img.xz

Write to a USB stick, boot it in UEFI mode and the bios update is named L43_0111_uefiX64.efi in the root directory. I haven't tested this yet, my other machine won't be here for a few more days, but it should work fine. I used @fohdeesha's method for updating the T620 (thanks!)

Does anyone know if it's worth the time to update the TPM module firmware? I tried just popping the .bin into \hp\devfw\tpm\TPMfirmware.bin on my bootable EFI stick but I think the BIOS is looking at the EFI partition on the SATA drive and doesn't find the image on USB.

Here's HP's TPM firmware update advisory: HPSBHF03568 rev. 9 - Infineon TPM Security Update | HP® Customer Support
 
  • Like
Reactions: fohdeesha

fake-name

Active Member
Feb 28, 2017
150
113
43
70
Wow, the built-in NIC in these things is a trashfire:

Code:
[   13.150063] IPv6: ADDRCONF(NETDEV_CHANGE): enp1s0: link becomes ready
[  257.745077] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0003 address=0x0000000000004000 flags=0x0050]
[  270.880407] ------------[ cut here ]------------
[  270.880424] WARNING: CPU: 3 PID: 0 at /build/linux-3busU_/linux-4.4.0/net/sched/sch_generic.c:306 dev_watchdog+0x23d/0x250()
[  270.880429] NETDEV WATCHDOG: enp1s0 (r8169): transmit queue 0 timed out
[  270.880431] Modules linked in: binfmt_misc rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs lockd grace fscache hp_wmi sparse_keymap kvm irqbypass snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec_hdmi edac_mce_amd serio_raw fam15h_power edac_core snd_hda_intel k10temp snd_hda_codec i2c_piix4 joydev input_leds snd_hda_core snd_hwdep snd_pcm snd_timer snd shpchp soundcore mac_hid sunrpc ib_iser rdma_cm iw_cm ib_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi autofs4 btrfs raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c raid1 raid0 multipath linear pata_acpi hid_generic amdkfd crct10dif_pclmul amd_iommu_v2 crc32_pclmul ghash_clmulni_intel radeon aesni_intel aes_x86_64 lrw gf128mul glue_helper ablk_helper cryptd i2c_algo_bit
[  270.880524]  ttm psmouse drm_kms_helper syscopyarea sysfillrect usbhid pata_atiixp sysimgblt ahci fb_sys_fops hid libahci drm r8169 mii wmi fjes video
[  270.880549] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 4.4.0-133-generic #159-Ubuntu
[  270.880553] Hardware name: HP HP t730 Thin Client/8103, BIOS L43 v01.08 03/21/2017
[  270.880556]  0000000000000286 f0ee371fe96c68ed ffff88013ed83d98 ffffffff814032d3
[  270.880562]  ffff88013ed83de0 ffffffff81d751a0 ffff88013ed83dd0 ffffffff81084992
[  270.880567]  0000000000000000 ffff88008f9f5a80 0000000000000003 ffff880098304000
[  270.880572] Call Trace:
[  270.880576]  <IRQ>  [<ffffffff814032d3>] dump_stack+0x63/0x90
[  270.880592]  [<ffffffff81084992>] warn_slowpath_common+0x82/0xc0
[  270.880597]  [<ffffffff81084a2c>] warn_slowpath_fmt+0x5c/0x80
[  270.880604]  [<ffffffff8176dd8d>] dev_watchdog+0x23d/0x250
[  270.880609]  [<ffffffff8176db50>] ? dev_deactivate_queue.constprop.33+0x60/0x60
[  270.880615]  [<ffffffff810f1997>] call_timer_fn+0x37/0x130
[  270.880620]  [<ffffffff8176db50>] ? dev_deactivate_queue.constprop.33+0x60/0x60
[  270.880625]  [<ffffffff810f217d>] run_timer_softirq+0x23d/0x2f0
[  270.880630]  [<ffffffff81089839>] __do_softirq+0x109/0x2b0
[  270.880634]  [<ffffffff81089b55>] irq_exit+0xa5/0xb0
[  270.880642]  [<ffffffff81857840>] smp_apic_timer_interrupt+0x50/0x60
[  270.880647]  [<ffffffff818551ff>] apic_timer_interrupt+0xbf/0xd0
[  270.880649]  <EOI>  [<ffffffff816e432e>] ? cpuidle_enter_state+0x11e/0x2d0
[  270.880660]  [<ffffffff816e4517>] cpuidle_enter+0x17/0x20
[  270.880666]  [<ffffffff810c8c12>] call_cpuidle+0x32/0x60
[  270.880671]  [<ffffffff816e44f9>] ? cpuidle_select+0x19/0x20
[  270.880675]  [<ffffffff810c8ed6>] cpu_startup_entry+0x296/0x360
[  270.880681]  [<ffffffff81052fa7>] start_secondary+0x177/0x1b0
[  270.880685] ---[ end trace b368f53ef80520c9 ]---
[  270.899271] r8169 0000:01:00.0 enp1s0: link up
[  278.777192] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0003 address=0x0000000000004000 flags=0x0050]
[  300.908699] r8169 0000:01:00.0 enp1s0: link up
If you load it hard, it shits the bed every few minutes, and sometimes doesn't come back at all.

****ing realtek.
 

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
Wow, the built-in NIC in these things is a trashfire:

Code:
[   13.150063] IPv6: ADDRCONF(NETDEV_CHANGE): enp1s0: link becomes ready
[  257.745077] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0003 address=0x0000000000004000 flags=0x0050]
[  270.880407] ------------[ cut here ]------------
[  270.880424] WARNING: CPU: 3 PID: 0 at /build/linux-3busU_/linux-4.4.0/net/sched/sch_generic.c:306 dev_watchdog+0x23d/0x250()
[  270.880429] NETDEV WATCHDOG: enp1s0 (r8169): transmit queue 0 timed out
[  270.880431] Modules linked in: binfmt_misc rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs lockd grace fscache hp_wmi sparse_keymap kvm irqbypass snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec_hdmi edac_mce_amd serio_raw fam15h_power edac_core snd_hda_intel k10temp snd_hda_codec i2c_piix4 joydev input_leds snd_hda_core snd_hwdep snd_pcm snd_timer snd shpchp soundcore mac_hid sunrpc ib_iser rdma_cm iw_cm ib_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi autofs4 btrfs raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c raid1 raid0 multipath linear pata_acpi hid_generic amdkfd crct10dif_pclmul amd_iommu_v2 crc32_pclmul ghash_clmulni_intel radeon aesni_intel aes_x86_64 lrw gf128mul glue_helper ablk_helper cryptd i2c_algo_bit
[  270.880524]  ttm psmouse drm_kms_helper syscopyarea sysfillrect usbhid pata_atiixp sysimgblt ahci fb_sys_fops hid libahci drm r8169 mii wmi fjes video
[  270.880549] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 4.4.0-133-generic #159-Ubuntu
[  270.880553] Hardware name: HP HP t730 Thin Client/8103, BIOS L43 v01.08 03/21/2017
[  270.880556]  0000000000000286 f0ee371fe96c68ed ffff88013ed83d98 ffffffff814032d3
[  270.880562]  ffff88013ed83de0 ffffffff81d751a0 ffff88013ed83dd0 ffffffff81084992
[  270.880567]  0000000000000000 ffff88008f9f5a80 0000000000000003 ffff880098304000
[  270.880572] Call Trace:
[  270.880576]  <IRQ>  [<ffffffff814032d3>] dump_stack+0x63/0x90
[  270.880592]  [<ffffffff81084992>] warn_slowpath_common+0x82/0xc0
[  270.880597]  [<ffffffff81084a2c>] warn_slowpath_fmt+0x5c/0x80
[  270.880604]  [<ffffffff8176dd8d>] dev_watchdog+0x23d/0x250
[  270.880609]  [<ffffffff8176db50>] ? dev_deactivate_queue.constprop.33+0x60/0x60
[  270.880615]  [<ffffffff810f1997>] call_timer_fn+0x37/0x130
[  270.880620]  [<ffffffff8176db50>] ? dev_deactivate_queue.constprop.33+0x60/0x60
[  270.880625]  [<ffffffff810f217d>] run_timer_softirq+0x23d/0x2f0
[  270.880630]  [<ffffffff81089839>] __do_softirq+0x109/0x2b0
[  270.880634]  [<ffffffff81089b55>] irq_exit+0xa5/0xb0
[  270.880642]  [<ffffffff81857840>] smp_apic_timer_interrupt+0x50/0x60
[  270.880647]  [<ffffffff818551ff>] apic_timer_interrupt+0xbf/0xd0
[  270.880649]  <EOI>  [<ffffffff816e432e>] ? cpuidle_enter_state+0x11e/0x2d0
[  270.880660]  [<ffffffff816e4517>] cpuidle_enter+0x17/0x20
[  270.880666]  [<ffffffff810c8c12>] call_cpuidle+0x32/0x60
[  270.880671]  [<ffffffff816e44f9>] ? cpuidle_select+0x19/0x20
[  270.880675]  [<ffffffff810c8ed6>] cpu_startup_entry+0x296/0x360
[  270.880681]  [<ffffffff81052fa7>] start_secondary+0x177/0x1b0
[  270.880685] ---[ end trace b368f53ef80520c9 ]---
[  270.899271] r8169 0000:01:00.0 enp1s0: link up
[  278.777192] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0003 address=0x0000000000004000 flags=0x0050]
[  300.908699] r8169 0000:01:00.0 enp1s0: link up
If you load it hard, it shits the bed every few minutes, and sometimes doesn't come back at all.

****ing realtek.
That's pretty much exactly what the dual Realtek interface on my J1900 does. It was fine when my ISP link was 60/5 but now that I've got symmetric gigabit it shits itself pretty regularly.
 

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
You might need to hunt down the dkms drivers from Realtek and install that instead. The bundled drivers are a known headache for the past few years.
It's also worth hunting down the Realtek BSD drivers and compiling them if you're running pfsense. I've followed these instructions here:
compile realtek network driver for pfsense 2.4.x

Switch out the BSD version for whichever your pfsense version is based on. I've built on FreeBSD 10.3 and 11 successfully.

@fake-name What OS were you running where the adapter crapped itself? I've already built the Realtek dkms driver and I'm running a 15 minute iperf3 across it now, I'll reboot and load the stock driver and test that one when this is done.
 
Last edited:
  • Like
Reactions: WANg

WANg

Well-Known Member
Jun 10, 2018
871
501
93
Reckon these would be any good running parsec gaming client with a predator x34 (they seem to support 21:9 displays)?
Probably? AMD designed the SoC to act as arcade and gaming (video slot/video poker) boards targeting integrators like EGT, IGT, Konami and Sega. Hell, if I get a second one at a reasonable price I'll test it as a standalone MAME machine - I bet it'll do quite well provided there are decent video drivers.
 
  • Like
Reactions: Cadal

WANg

Well-Known Member
Jun 10, 2018
871
501
93
It's also worth hunting down the Realtek BSD drivers and compiling them if you're running pfsense. I've followed these instructions here:
compile realtek network driver for pfsense 2.4.x

Switch out the BSD version for whichever your pfsense version is based on. I've built on FreeBSD 10.3 and 11 successfully.

@fake-name What OS were you running where the adapter crapped itself? I've already built the Realtek dkms driver and I'm running a 15 minute iperf3 across it now, I'll reboot and load the stock driver and test that one when this is done.
Heh, is it still crapulent with their drivers, or do you see a substantial improvement? I can tell you from experience that even on a decent machine with Win7, Gigabit transfers will still peg a machine with 70% CPU utilization. Man, I miss the days where notebooks are bundled with Broadcom Tigons and 3Com NICs. Realteks and Marvells are not stuff you find on network gear for a good reason.
 

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
Heh, is it still crapulent with their drivers, or do you see a substantial improvement? I can tell you from experience that even on a decent machine with Win7, Gigabit transfers will still peg a machine with 70% CPU utilization. Man, I miss the days where notebooks are bundled with Broadcom Tigons and 3Com NICs. Realteks and Marvells are not stuff you find on network gear for a good reason.
The DKMS driver was rock solid while blasting full line rate through it for 15 minutes straight. The normal Debian package won't install on Proxmox 5.2 though, there was a breaking change in kernel 4.15, so you need the package from Buster or Sid.

Code:
root@bill:~# dmesg | grep r8168
[    3.845568] r8168: loading out-of-tree module taints kernel.
[    3.847272] r8168 Gigabit Ethernet driver 8.046.00-NAPI loaded
[    3.863277] r8168: This product is covered by one or more of the following patents: US6,570,884, US6,115,776, and US6,327,625.
[    3.863288] r8168  Copyright (C) 2018  Realtek NIC software team <nicfae@realtek.com> 
[    3.918932] r8168 0000:02:00.0 enp2s0: renamed from eth0
[   16.138544] r8168: enp2s0: link up
...
root@bill:~# iperf3 -c meta -t 900
...
[  4] 889.00-890.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 890.00-891.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 891.00-892.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 892.00-893.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 893.00-894.00 sec   112 MBytes   942 Mbits/sec    0   3.01 MBytes       
[  4] 894.00-895.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 895.00-896.00 sec   112 MBytes   942 Mbits/sec    0   3.01 MBytes       
[  4] 896.00-897.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 897.00-898.00 sec   112 MBytes   942 Mbits/sec    0   3.01 MBytes       
[  4] 898.00-899.00 sec   112 MBytes   941 Mbits/sec    0   3.01 MBytes       
[  4] 899.00-900.00 sec   112 MBytes   936 Mbits/sec    0   3.01 MBytes       
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-900.00 sec  98.6 GBytes   941 Mbits/sec    0             sender
[  4]   0.00-900.00 sec  98.6 GBytes   941 Mbits/sec                  receiver
iperf Done.
I'm actually about to start testing the stock kernel driver right now, I'll update in a bit.
 

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
@arglebargle - Ubuntu Server 16.04 LTS.

If there are non-shit drivers, why the hell haven't they been merged into the kernel mainline? What the hell?
I ask that every time I install Ubuntu and have to plug a lan cable into the machine because the wifi drivers are shit.

Don't even get me started on the realtek drivers pfsense ships with, holy ****.
 

WANg

Well-Known Member
Jun 10, 2018
871
501
93
@arglebargle - Ubuntu Server 16.04 LTS.

If there are non-shit drivers, why the hell haven't they been merged into the kernel mainline? What the hell?
Licensing, mostly. A significant chunk of Linux drivers out there are either reverse engineered (usually kinda buggy or slow unless it matured over the years), or is just a bunch of code shim talking to a binary blob interfacing with the hardware (faster but much less flexible, and dependent on the whims of driver developers hacking shims to get it to work) - it's getting increasingly rare for Linux driver devs to get register level specs on hardware.
 
Last edited:

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
Code:
...
[  4] 892.00-893.00 sec   112 MBytes   941 Mbits/sec    0   1.99 MBytes      
[  4] 893.00-894.00 sec   112 MBytes   936 Mbits/sec    0   1.99 MBytes      
[  4] 894.00-895.00 sec   112 MBytes   941 Mbits/sec    0   1.99 MBytes      
[  4] 895.00-896.00 sec   112 MBytes   941 Mbits/sec    0   1.99 MBytes      
[  4] 896.00-897.00 sec   111 MBytes   934 Mbits/sec    0   1.99 MBytes      
[  4] 897.00-898.00 sec   112 MBytes   938 Mbits/sec    0   1.99 MBytes      
[  4] 898.00-899.00 sec   112 MBytes   941 Mbits/sec    0   1.99 MBytes      
[  4] 899.00-900.00 sec   112 MBytes   941 Mbits/sec    0   1.99 MBytes      
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-900.00 sec  98.6 GBytes   941 Mbits/sec    0             sender
[  4]   0.00-900.00 sec  98.6 GBytes   941 Mbits/sec                  receiver
iperf Done.
Well, I'm not sure what's up with the Ubuntu 16.04 drivers. I just ran 15 minutes of iperf through the stock r816* driver that ships with proxmox and it functioned flawlessly.

Code:
root@bill:~# uname -a
Linux bill 4.15.18-2-pve #1 SMP PVE 4.15.18-20 (Thu, 16 Aug 2018 11:06:35 +0200) x86_64 GNU/Linux
root@bill:~# lsmod | grep ^r
radeon               1474560  1
rdma_cm                61440  1 ib_iser
raid6_pq              114688  1 btrfs
r8169                  86016  0
root@bill:~# dmesg | grep r816
[    3.785421] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[    3.797502] r8169 0000:02:00.0 eth0: RTL8168h/8111h at 0x        (ptrval), -, XID 14100800 IRQ 43
[    3.797508] r8169 0000:02:00.0 eth0: jumbo features [frames: 9200 bytes, tx checksumming: ko]
[    3.845545] r8169 0000:02:00.0 enp2s0: renamed from eth0
[   13.083028] r8169 0000:02:00.0 enp2s0: link down
[   13.083061] r8169 0000:02:00.0 enp2s0: link down
[   16.179296] r8169 0000:02:00.0 enp2s0: link up
root@bill:~#
This is the module that ships with proxmox, no fancy DKMS package ¯\_(ツ)_/¯
 

WANg

Well-Known Member
Jun 10, 2018
871
501
93
iperf it up first for like 10 minutes and see how it performs. Either the default drivers improved significantly in the past 2-3 years, or someone merged the vendor driver upstream. Either ways, I still prefer anything other than Marvell/Realtek if I can help it.
 
Last edited:

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
iperf it up first for like 10 minutes and see how it performs. Either the default drivers improved significantly in the past 2-3 years, or someone merged the vendor driver upstream. Either ways, I still prefer anything other than Marvell/Realtek if I can help it.
Yeah, that's the tail end of 15 minutes of 1Gbps iperf traffic, note the timestamps before the summary, it was totally fine the entire time.

I only planned to use the onboard nic as a fallback management interface in case a kernel update broke the Mellanox DKMS drivers but it seems like the Realtek is pretty alright.
 
  • Like
Reactions: WANg

WeekendWarrior

Active Member
Apr 2, 2015
299
96
28
53
Not to hijack the thread, but some Lenovo thin clients may be better than the T730 for general computing at the same price point.

For example: Lenovo ThinkCentre M93P Tiny Core i5-4570T 3.30GHz 4GB 500GB Desktop Computer PC | eBay. I just picked up a couple m93s with 8GB and a SSD and Win7 pro for $165 each.

Although some ThinkCentre "Tiny" (i.e., thin client) models have low-end components, the m93p has an Intel 82579 based NIC. An online component list is available. See page 186 at http://content.etilize.com/Manufacturer-Brochure/1026992540.pdf.

The idle power of the m93p will be higher than the HP/AMD units, which may factor into pfSense use, but otherwise m93p looks good.
 

WANg

Well-Known Member
Jun 10, 2018
871
501
93
Not to hijack the thread, but some Lenovo thin clients may be better than the T730 for general computing at the same price point.

For example: Lenovo ThinkCentre M93P Tiny Core i5-4570T 3.30GHz 4GB 500GB Desktop Computer PC | eBay. I just picked up a couple m93s with 8GB and a SSD and Win7 pro for $165 each.

Although some ThinkCentre "Tiny" (i.e., thin client) models have low-end components, the m93p has an Intel 82579 based NIC. An online component list is available. See page 186 at http://content.etilize.com/Manufacturer-Brochure/1026992540.pdf.

The idle power of the m93p will be higher than the HP/AMD units, which may factor into pfSense use, but otherwise m93p looks good.
The M93p doesn't have a PCIe x16 slot, which is the big draw of the t620Plus/t730 thin clients - since you'll be able to drop quadport 1GbE and dualport 10GbE NICs in there and then segregate it to various VMs via PCIe VFs. You can't do that on NUC/NUC-like chassis in general. Also, from a "fun" perspective, the t730 supports quad displays via its 4 Displayports. I doubt that many of those mini PCs can do that as well.

If it's general computing on a NUC-like chassis, there are alternatives similar to the m93 like the HP EliteDesk 800G2 Minis or the Dell Optiplex 7050 Micro and the like.
 
  • Like
Reactions: Quantum_Cake

arglebargle

H̸̖̅ȩ̸̐l̷̦͋l̴̰̈ỏ̶̱ ̸̢͋W̵͖̌ò̴͚r̴͇̀l̵̼͗d̷͕̈
Jul 15, 2018
656
233
43
The pcie slot is the key to these machines. With a 10GbE nic you have the bandwidth to exploit the SSD for clustered storage in addition to the pfsense role.
 

WANg

Well-Known Member
Jun 10, 2018
871
501
93
Okay, so I had some additional time over the weekend to do some testing - although not the testing that I wanted to do. I was able to load a Windows 10 1803 image on a 256GB SSD and got the Thin client to boot off it. I then loaded the standard t730 driver pack for Windows 10 and proceeded upon some hardware performance testing - mostly to see how well the Radeon R7 on the RX427BB performs. Unfortunately the Windows 10 install doesn't seem very stable as I constantly observe system crashes (no such issue is seen in Linux or ESXi, the machine had been rock solid to me). That being said, I was able to get the following metrics:

3DMarksVantage
P6588 (6257 GPU/7831 CPU)
P6623 (6259 GPU/8023 CPU)
P6603 (6258 GPU/7956 CPU)

3DMarks05
11112

Note that I didn't have a chance to install 3DMarks11 - but I would say that normal gaming/performance will be similar to the performance of an A8Pro-9600 (which is the same as the HP mt43 mobile thin client - I wouldn't recommend buying one as the mt44 client is Ryzen based and will drop in price quickly once the mt45 comes out)