Read more.Promises that it will be available on boards by start of October.
Read more.Promises that it will be available on boards by start of October.
I thought they retired. Now they are back for one more time around to fix AMD cpu's.
1) 3 months? Subtract a few weeks for motherboard vendors to take the AGESA package and test it thoroughly on all of their motherboards, so call it 10 weeks. This going to be mostly a software problem (firmware on the chip/motherboard). 10 weeks to fix a subtle issue that's affected by a multitude of variables, on a handful of different chips with differring configurations. I think turning a problem like this around in 10 weeks is a damn good result, considering it's very unlikely the whole team was working flat out for that time.I suppose the real question is one of why it has taken AMD and its motherboard partners the best part of three months - July 7 to Oct 1 (AGESA 1003ABBA release date) - to ensure its processors meet their rated specification?
2) The issue is still blown way out of proportion. Meeting the rated specification means they operate at at least the base clock speed on demand. Boost speeds are opportunistic, so if you're getting close (within 100MHz) to the rated boost speeds consistently, you've got a good chip.
Wasn't boosting orginally designed so that multi-core CPUs could boost a single thread, or small number of threads for a short period to provide high responsiveness to user applications such as a web browser loading a page, or opening a word document? If that's the case, the amount it can boost and the length of time it can boost for depend on several factors (power delivery, cooling system, etc), which again means boosting is opportunistic. Why on earth are people complaining because their CPU isn't all-core boosting for long periods of time at the "up to" speed? Surely the rated base clock is what you should expect to see when gaming or running at 100% load?
Please correct me if I'm wrong, but I do think this has gotten out of hand.
Agreed, though I think AMD could have pitched this better. I did read someone's testing where a really simple integer thread would reliably get them up to maximum boost, but clearly people are using SSE/AVX workloads like Cinebench which exercise a *lot* of transistors and so draw a lot of power. Feels like complaining that may car wouldn't do it's rated top speed when fitted with a roof box.
Since early APU days AMD have released updated "facelift" parts where a lot of the performance increase comes from finessing the power management from what they have learnt since the original silicon was introduced. I have to wonder if AMD are giving us some of that now to make this issue go away.
[QUOTE=afiretruck;1214771]I agree. I believe some press affiliated with Intel had the directive to find something that is wrong with Zen 2 to lower its success. So they found this stupid thing, that isn't really an issue because they say that only one core should go to 4.6Ghz for 3900X and in certain conditions and turned into all cores should boost together to 4.6Ghz in all scenarios. That is not the case. Also, the single core boost is generally spot on, with 25-100Mhz delta to rated boost speed.Please correct me if I'm wrong, but I do think this has gotten out of hand.
Still, I think that people smart enough understood this shady move from the press/competition and saw the huge value of these new CPUs. Hence the out of stock situation after two months from launch. I recently built a system with 3700X and it was boosting to 4395Mhz. Performance and responsiveness of the system is amazing. Great value for a 320$ CPU.
Well Intel has to make it look like AMD are still terrible. They have to try to sell cpus... Thou why you'd pick intel over AMD now given all the spectre issues, and its even worse in server market. Rome is MURDERING Xeons for far less pricing AND no stupid restrictions/artifical segmentation. Oh and you get to use ALL the memory you put in your server too. You'd be idiotic to buy a full spec Xeon now.
I has been said time and time again that the Intel boost clocks are not guaranteed and all they guarantee is the base clock. Indeed, the default for Intel boost is that it lasts for 8 seconds or until temps get high or until it breaches a specific TDP which is higher than that specified on the box.
So frankly, with all that in mind, this is a load of tripe. It's 25-75MHz off the boost which is not guaranteed and the results were also from an uncontrolled set of tests.
So what has AMD really done here? Probably gone "this is nothing", taken their recent optimisations they were going to release anyway and instead said they're a "fix" for the "issue".
I expect that stuff like this will be refined with updates and drivers which is why I, and so many others, will wait until a platform is mature before leaping.
The boost clock advertised for each part is the minimum speed that a single thread workload should achieve on a given part assuming cooling and power delivery are within requirements. Boosting above that is the opportunistic performance bit.
So the issue lies in either something that's been preventing the boost hitting the rated minimum speeds, or those speeds were unrealistic in the first place and shouldn't have been advertised with the parts.
This just affects the 3000 series though, 1000 and 2000 series chips hit their rated boost speeds in the same conditions. As do all Intel chips advertised in the same manner.
Either way, they're still great performing chips with good value, it's just a perception issue (and I guess a legal liability) in not performing as advertised.
After getting used to one sentence in a Hexus article they went into a healthy conversation in many paragraphs. Is the start of something great ? Probably not.
I recently had a chance to buy a 3800X chip and a new motherboard. It was that, or put some new tires on my 4X4-Tacoma. (living on a fixed retirement income is no fun)
When people started complaining about the Boost-Issue, I decided for the sure thing. (new tires)
Hopefully, if this is fixed, I can start saving for that CPU/Motherboard combo and try out AMD's latest offerings.
Phoronix found some overall improvement:
https://www.phoronix.com/scan.php?pa...00x-abba&num=1
Doesn't look a deal maker/breaker, but improvements are always nice.
There are currently 1 users browsing this thread. (0 members and 1 guests)