Yep, 2667v2 definitely has lower voltage ))) Here is performance in GFlops (and power consumption) under "high-AVX load" for reference.
Yeah, the OEM E5 2673 obviously uses a lower voltage than the 2667 accounting for the difference in TDP. But again, don't take my CPUz screenshots as a perfect representation. The voltage jumps around a good bit making comparison per clock impossible until we can go toe to toe on a testbench with same hardware. I will do a detailed study on voltage as well.
So I will ask again, what processors were you running in your Z820? Are you sure it was a z820 and not a budget wal mart economy gaming pc? Perhaps you are getting the two things confused. In terms of CPU cooler reliability, I've owned both my z820 rigs for about 6 months now and I've had zero problems with mine, and I will say they are actually very well built with very robust and reliable high end Nidec factory fans and include active cooling on all 16 RAM slots as well. So your response is a little puzzling since one look at the CPU cooler design should erase all doubts of any "cheap construction" going on here. In my estimation it is usually the electrical connection between the CPU cooler and mobo that you must be overlooking when you cry about reliability problems here. The alignment teeth can fail, sometimes get broken off and the electrical connection from CPU fan to motherboard can be compromised which will throw some errors during the POST process. This can be fixed with a 20 cent zip tie, fyi.
Sure for all out servers you are talking about I assume? This is a workstation, a rather old workstation, that has nearly 100GB effective bandwidth across the board, far higher bandwidth than even extremely high running state of the art and overclocked DDR4 higher performance desktops you would find on the market today. Thats the class I have it configured for, it's a gaming machine, desktop machine and some rendering work as well. Point being, for similar usage roles, this machine has roughly 50% MORE BANDWIDTH THAN A BRAND NEW HOT OF THE ASSEMBLY LINE HIGH PERFORMANCE DESKTOP / GAMING PC. So my original statement was correct. It still kicks ass in the memory bandwidth division based on the classification of using it as a gaming and general purpose high performance desktop PC.
Also, can you provide an example of a modern day single socket system that has considerably more bandwidth than 100GB R/W/C?
Have you ever heard about heavy AVX loads, son? Intel implemented different sets of multipliers in the 3rd version of xeon family for AVX load simply because v2's were unable to fit into the claimed TDP under AVX load and sustain the all core turbo (leading to turbo throttling).
Of course he knows. He knows everything about it.
So just a quick update on this controversial thread. I've now had years to test out the z820 E5 2673 v2 rig. This particular machine has stock liquid cooling and two 8 core processors that have a single core turbo of 4.0GHz and an all core turbo of 4.0GHz. The TDP comes in at 110Watts. With two processors that's 220W and 16 cores and 32 threads. Contrary to what you are saying, my 2673, despite the lower TDP always held 3.6GHz all core boost under any performance conditions, and never once dropped below that.
No problem, some people respond with slight anger if they are wrong because most people, especially technical people who work on computers think they are correct 100% of the time) any conflict in the thread but nothing of that has been from me, and im not perfect either. I'm simply correcting some issues I see, nothing personal and laying out the facts. I'm just basically stating specs and stats from general processors that have been double checked to be confirmed accurate. I have nothing against anyone here, you guys have been great. I hope we can all get through this sincerity from one and all because the last thing I want is conflict here. On anger / driving force; for other users to check CPU database sites to confirm or deny my claims are getting bad information on actually turbo speeds, nearly all online vendors get this wrong.
Here is where we are at right with memory bandwidth:
that's really cool, because 2650v2 (8 cores) can't even hold 2.9GHz all cores turbo under Linx, simply because of 95W TDP wall. First 10 seconds it is allowed to consume more, but once TDP limitation comes into play, frequency drops. And that's the same Ivy generation. But yeah, I believe that under light load everything is possible. Cooling has nothing to do about that, I've never seen CPU temperature above 62C
Correction: both processors have an all core turbo speed of 3.6GHz, I will respond to the other things later I just wanted to clarify that first. my bad, but I will be responding and expanding on my perspective shortly on the other controveries. lolthat's really cool, because 2650v2 (8 cores) can't even hold 2.9GHz all cores turbo under Linx, simply because of 95W TDP wall. First 10 seconds it is allowed to consume more, but once TDP limitation comes into play, frequency drops. And that's the same Ivy generation. But yeah, I believe that under light load everything is possible. Cooling has nothing to do about that, I've never seen CPU temperature above 62C
Yeah, you obviously cant tell much about "performance" until you benchmark said hardware. That's why they make benchmarks, lol. This is my z820 maxed out with best possible memory config for performance. Not bad considering Zen 2 was for a while in a very similar range in terms of memory latency numbers. This memory configuration with net you almost twice the memory bandwidth of a brand new state of the art dual channel DDR4 system running at approximately 4000MHZ. This is simply posted to demonstrate the competitiveness in 2021 of the aging DDR3 platform of the z820.
Hmm, that's strange.... my 2650v had no trouble holding 2.9GHz for even the longer duration benchmarks. Here take a look for yourself. Even monitoring real time I never saw the speed dip below 3GHz.that's really cool, because 2650v2 (8 cores) can't even hold 2.9GHz all cores turbo under Linx, simply because of 95W TDP wall. First 10 seconds it is allowed to consume more, but once TDP limitation comes into play, frequency drops. And that's the same Ivy generation. But yeah, I believe that under light load everything is possible. Cooling has nothing to do about that, I've never seen CPU temperature above 62C
LOL, that's because CPU-Z which you are using for benchmarking is unable to make more than 70W load on 2650v2. And I'm talking about Linx that pulls about 110W from 2650v2 (at 3GHz!) during first 10 seconds, then 'Power Limit 1' comes into play and CPU is not permitted to consume more than TDP hence dropping its frequency below 2.9GHz.