• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

New High Performance, x86 Compatible Microprocessors from Centaur / VIA

Cool. I had a Centaur WinChip 200 back in the day. Wasn't anything special except for being very cheap.
 
The fact that it's supposed to be fast, cheap, and energy efficient at the same time, sounds so good, that I doubt that it's actually "a new challenger" but rather something that's going to be a niche product for people who just need lots of x86 cpu...
 
The Centaur guys are good at what they're doing, but unfortunately the funding from VIA is what's prevented them from bringing out some real competition.
That said, the Zhaoxin chips so far has been based on old Centaur tech, maybe these new ones will really manage to catch up.
I know a few guys working at Centaur and I know they've been working on this for quite some time.
And yes, these chips will be socket compatible with Intel, but as such, they won't be sold outside of China.
 
Heh so that's the link between those chips huh. Interesting.
 
The last VIA CPU I saw was soldered to the motherboard, but it did manage to run XP, within reason.
 
i miss cyrix.
I don't. The 6x86 CPUs were pure garbage. I remember one game crashing upon launching because of a bug in the CPU.
I don't miss VIA either. I believe it was KT133 chipset that was bugged to oblivion as well.
 
Last edited:
I don't. The 6x86 CPUs were pure garbage. I remember one game crashing upon launching because of a bug in the CPU.
I don't miss VIA either. I believe it was KT133 chipset that was bugged to oblivion as well.
I believe that was Quake, but it didn't crash; it did, however, run slow as all hell because it effectively had half the floating-point compute capacity when compared to the then Pentium chip.
 
Didn't play Quake, but I'm not surprised it wasn't just one game. The CPUs were crap :D
 
I remember trying to play Microsoft Flight Sim of some generation on a Cyrix thing...

It was really bad.
 
The issue with the 686 is it's got alu performance that's better than p2 clock for clock but it's floating point module was the used in the 486 era, a non pipelined 32bit fp.
 
I don't. The 6x86 CPUs were pure garbage. I remember one game crashing upon launching because of a bug in the CPU.
I don't miss VIA either. I believe it was KT133 chipset that was bugged to oblivion as well.

To be fair they were pretty darn smart ppl given that they created an x86 cpu from scratch and reverse engineering. That didn't stop Intel from litigating them to death lol.
 
I don't remember any of this, probably because Cyrix was just before my time :p
 
Maybe my bogus R15 score will become a reality one day...
134885
 
CentaurHauls Family 6 Model 71 Stepping 1
ZHAOXIN KX-7000 ZX-F OctaCore 2000MHz @ 2.00 GHz
1 processor, 8 cores



L1 Instruction Cache: 32 KB x 8
L1 Data Cache: 32 KB x 8
L2 Cache: 256 KB x 8
L3 Cache: 16 MB x 1


Single-Core Performance
Single-Core Score: 469
Crypto Score: 710
Integer Score: 446
Floating Point Score: 478

Multi-Core Performance
Multi-Core Score: 3 264
Crypto Score: 4 553
Integer Score: 3 133
Floating Point Score: 3 334


source: https://browser.geekbench.com/v5/cpu/526995
source: https://browser.geekbench.com/v5/cpu/526995.gb5
 
Last edited:
Centaur Technology's Deep-Learning Coprocessor Technology
source: http://www.ece.utexas.edu/events/ce...5DL_d1sa11HCDEU5OpZZ0UGeR0D3Aoa_sdB9FXdBusQog

Centaur Technology Reference Design v1.0 (8-core Centaur Integrated x86 CPUs @ 2.5GHz with Centaur Integrated AI Coprocessor)
source: https://www.cdrinfo.com/d7/content/mlperf-releases-results-machine-learning-inference-benchmark

Code:
{
    "division": "closed",
    "status": "preview",
    "submitter": "CentaurTechnology",
    "system_name": "Centaur Technology Reference Design v1.0",
    "number_of_nodes": 1,
    "host_memory_capacity": "32GB",
    "host_processor_core_count": 8,
    "host_processor_frequency": "2.5GHz",
    "host_processor_model_name": "Centaur Integrated x86 CPUs",
    "host_processors_per_node": 1,
    "host_storage_capacity": "120GB",
    "host_storage_type": "SSD",
    "accelerator_frequency": "-",
    "accelerator_host_interconnect": "-",
    "accelerator_interconnect": "-",
    "accelerator_interconnect_topology": "-",
    "accelerator_memory_capacity": "4GB",
    "accelerator_memory_configuration": "none",
    "accelerator_model_name": "Centaur Integrated AI Coprocessor",
    "accelerator_on-chip_memories": "-",
    "accelerators_per_node": 1,
    "framework": "TensorFlow commit f5ce1c00d4397875ff3d706881bd46430f4a9667 and f6fbfe013898dc16bd35ba380387ff02d0275ac3 + custom patches",
    "operating_system": "ubuntu-18.04",
    "other_software_stack": "Centaur ML Library Build 1223 and 1220",
    "sw_notes": "The same hardware system is used for all mobilenet, resnet, ssd-small, and gnmt sumbissions, but the software is slightly different for gnmt.
Submissions for mobilenet, resnet, and ssd-small use TF commit f5ce1c00d4397875ff3d706881bd46430f4a9667 + custom patches to TF-Lite,
Centaur ML Library Build 1223, and host_processor_frequency set to 2.5GHz. Submissions for gnmt use TF commit f6fbfe013898dc16bd35ba380387ff02d0275ac3
+ custom patches for compiler/graph matching, Centaur ML Library Build 1220, and host_processor_frequency set at runtime via software to 2.3GHz."
}
source: https://github.com/mlperf/inference_results_v0.5/blob/master/closed/CentaurTechnology/systems/0.json
 
Last edited by a moderator:
The Geekbench already has a 2GHz sample of the Chinese x86 Zhaoxin KX-7000 processor. 80% IPC Zen 2

source:
https://translate.google.com/translate?hl=sk&sl=cs&tl=en&u=https://www.cnews.cz/procesor-via-zhaoxin-kx-7000-unik-geekbench-2ghz/


AMD FX-8300 3300MHz @ 2.00GHz
Single-Core Score: 306 (65,25%)
Multi-Core Score: 1 705 (52,24%)


vs


KX-7000 ZX-F OctaCore 2000MHz @ 2.00GHz
Single-Core Score: 469 (+53,27%)
Multi-Core Score: 3 264 (+91,44%)


source:
https://browser.geekbench.com/v5/cpu/compare/526995?baseline=562975


___________________
srR1pdG.jpg

source: https://browser.geekbench.com/v5/cpu/562975.gb5
source: https://browser.geekbench.com/v5/cpu/562975


P.S. If you have Geekbench 5 results your modern CPU @ 2.00GHz so pls write here! Thank you.
 
Last edited:
MLPerf Inference v0.5 Results
source: https://mlperf.org/inference-results


COCO object detection on SSD-MobileNet v1 (images/sec)
Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics)217,9333,43 %
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor)651,89299,13 %
ImageNet image classification on MobileNet v1 (images/sec)images/sec
Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics)507,718,40 %
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor)6 042,341190,12 %
ImageNet image classification on ResNet-50 v1.5, (images/sec)
Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics)100,938,28 %
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor)1 218,481207,25 %


source: source: source: https://www.intel.ai/mlperf-nov2019/#gs.h01eu8
source:
 
Back
Top