thank youJust telling you that the microcode that gave far too high of voltage and fried the chips is not being used any more.
I certainly have no idea if the fundamental silicon has changed or not. Only an Intel insider could verify that. There was a slight bump in frequencies. The specs are the basically the same otherwise.
I RMA's 3 Raptors. First one I ran stock. 2nd and 3rd at really conservative settings and they still started to degrade. By "started to degrade" I mean after a few months they wouldn't run the frequencies/voltages they were running in a stable manner. I remember I was giving Topaz tech support all kinds of hell because Photo AI was suddenly locking up my system, which I thought was due to one of their updates. They kept telling me to check my system stability and I was telling them I'm 100% certain I was stable. Then I backed my frequency down to 4GHz and all of the problems went away. I sheepishly went back and told them they were right. Topaz Video AI is a good application to test stability by the way. Once I figured this out I proceeded to get the RMA going on my third and last Raptor Lake CPU and seriously started to consider a Zen 5 rig...
I never ran them over 5.5GHz and never overclocked or ran over 85C. I could have increased voltage to maintain performance and maybe they would have "stabilized." Or maybe not. I didn't want to find out. After three strikes I retired the Raptor train and moved to a 9950X. Just running stock and no issues, no fiddling with the BIOS, load lines, and endless tweaking to hopefully find a configuration that will last under high stress rendering and other stuff I do a lot of. While I don't overclock I am "hard" on CPU's I think. I'll be rendering out a video in Vegas Pro while running PureRaw to process photos while at the same time processing them with Photoshop, running a tv stream in the background, and flipping through websites all at the same time. I expect my computer to be responsive and stable the entire time. I work in a very "bursty" manner. When I sit down, it's "go" time and I don't want to work around my computer. I want IT to work around me.
Perhaps I got bad samples or just wasn't good at set up or didn't have good enough cooling, could be on me. I'm about 5 months on the 9950X and so far so good, performance is better and I've spent literally zero time tweaking. I think I increased TDP, decreased TDP, and then settled on stock. AMD seems to have found the sweet spot so I'll stick with it.
It seems like all of the "fixes" for Raptor and the like are not silicon fixes but more tweaks to the BIOS. After my experience I am wary especially when there is such a well-performing and cost effective alternative available. I am a fan of CPU's. I don't care who makes them. It took me A LOT to give AMD a try after literally lifetime of Intel Only Inside for me. Part of me was supporting Intel because I don't want them to fail. I want them to succeed... and I want AMD to succeed as well. But eventually I was wasting too much time investigating, tweaking, and trying to figure things out. In the beginning it was fun but it became a chore.
Seems like Arrow Lake isn't having the same issues and performs very well in a lot of the apps I use though. I'll take a good look at both AMD and Intel when it's time to upgrade again. I am still under the impression that there is nothing inherently wrong with Raptor Lake silicon, Intel just got too aggressive with boosting frequencies and they are having a hard time walking it back due to the published specs, which they obviously don't want to revise so they are dancing all around trying to maintain the specs in theory but not in practice. That is my theory anyway.
The NOC Fabric connecting the dies and the slow L3 is the issue also nothing will beat a monolithic die in terms of latency but different kind of packing affect latency differently.Then how ocme Raptor Lake with similar deisgn is so much better with latency and thus gaming?
Is it due to topology difference where they have e-cores between P cores on Arrow Lake rather than all 8 P cores in a row then the e-core clusters on ADL and RPL.
An also due to the fact maybe even ore so that the IMC is not on the ring but a separate tile unlike Raptor and Alder?
The NOC Fabric connecting the dies and the slow L3 is the issue also nothing will beat a monolithic die in terms of latency but different kind of packing affect latency differently.
AMD's current cheap packing affects the latency to a large extent but foveros/Cowos type of packing doesn't to that extent.
No AMDs current packing is cheap packing that any OSAT can do and CowoS is TSMCs IP while Foveros is Intel's and yes there is no type advanced packing can catch up to Monolithic Die Latency and bandwidth at least till now.Are you saying AMD using foreros.Cowos type is not as bad for latency as what Intel used for Arrow Lake? But its still much worse than Intel's monolithic dies like Raptor and Alder Lake for latency.
Yeah the stacked cache is pretty good it uses Hybrid Bonding a form advanced packing as well.AMD can just mask it with the staked cache which caused the X3D CPUs to be released for gaming?
No AMDs current packing is cheap packing that any OSAT can do and CowoS is TSMCs IP while Foveros is Intel's and yes there is no type advanced packing can catch up to Monolithic Die Latency and bandwidth at least till now.
Yeah the stacked cache is pretty good it uses Hybrid Bonding a form advanced packing as well.
So are you saying Intel's packing of Arrow Lake CPUs is better than AMD's Ryzen 5000-9000 CPUs??
Its just that Intel screwed up with their implementation of Arrow Lake with tiles and IMC being separate tile rather than on ring and their L3 cache implementation making L3 cache so slow which is why its performance is so much worse in latency sensitive tasks like gaming compared to Intel's own prior Gen RPL??
And it seems AVX512 is still have to be disabled.
Only if they have trouble getting good yields for all 12 P-cores on the die. Gamers and enthusiasts that dislike both Arrow Lake and Ryzen for any reason will go crazy for this SKU.And for some reason I expect they'll be priced professionally.
Only if they have trouble getting good yields for all 12 P-cores on the die. Gamers and enthusiasts that dislike both Arrow Lake and Ryzen for any reason will go crazy for this SKU.
Think it will help get heat and power down vs raptor?Just to be disappointed that is nearly zero gain.
Think it will help get heat and power down vs raptor?
Think it will help get heat and power down vs raptor?
12 strong cores might do well in CPU heavy games that like a lot of threads. Looking forward to results for games like Space Marine 2, MS Flight Sim, Spiderman Miles Morales and 2 with max RT. No big outlet tests stuff like Planet Coaster but sims like that seem to love a lot of strong cores.
Yield is not an issue with Intel 7/Intel 3 this point.Only if they have trouble getting good yields for all 12 P-cores on the die. Gamers and enthusiasts that dislike both Arrow Lake and Ryzen for any reason will go crazy for this SKU.
I doubt so cause the issue of the heat is the Raptor Cove P cores cause they require a good chunk of power for the boost performance.Think it will help get heat and power down vs raptor?
12 strong cores might do well in CPU heavy games that like a lot of threads. Looking forward to results for games like Space Marine 2, MS Flight Sim, Spiderman Miles Morales and 2 with max RT. No big outlet tests stuff like Planet Coaster but sims like that seem to love a lot of strong cores.
Only if they have trouble getting good yields for all 12 P-cores on the die. Gamers and enthusiasts that dislike both Arrow Lake and Ryzen for any reason will go crazy for this SKU.
I seriously doubt 12P cores on Intel's last gen architecture and node is going to beat 9800x3D. And if you need more cores, there is 9950x3D. I know it is not on one CCD, but I think AMDs more efficient and modern core design, along with the vcache will more than cancel out the latency penalty of dual CCDs. I'm not even sure it will beat 14900K. It all depends on what sort of frequencies they are able to get at a manageable power usage. If they have to sacrifice clockspeed to get power and temps under control, I dont hold out much hope for significant improvement over 14900k.9800Well Arrow Lake already disliked by most gamers and for good reasons as it regressed in gaming badly vs Raptor Lake. The updates to firmware only marginally made it better.
The ones who will go crazy for this SKU are those that hate nor trust Intel hybrid Big.Little arch and dual CCD where there is a maximum of 8 cores if you go single CCD and oh most important want more than 8 cores with modern IPC and feature set.
If this is released and unlocked and on consumer mainstream market this would be the first CPU with more than 8 P-cores cores on one node/die from either AMD or Intel since the 10 core Intel 10th Gen Comet Lake parts.
I may go crazy for it, but am so sick of part swapping. Have done too much already and would have been done a while ago if they had this CPU just 2 years ago and also would have just stuck to 9800X3D during my TRTX 5090 venture which led to platform issues but because of bad NVIDIA drivers (when NVIDIA drivers are usually very good all the time) that made me think AMD bad platform issues.
Not so with better NVIDIA drivers now.
But I still kind of itch for more than 8 homogenous modern IPC featured P cores on one node lol.
But so tired of parting in and out and returning and selling things that I may just stick with 9800X3D no matter what for a whole.
Its just too much work.
It's supposedly monolithic entirely basically Gracemont cluster is replaced with P cores In a 13900K.Sorry if this has already been asked, but what's the core layout/topology for 12P Barlett Lake?
Well the efficiency upgrade would have been very good in a monolithic design.I think Bartlet Lake would have been more interesting on Intel3 with Redwood cove cores instead of raptor cove on 7. While RWC is by no means a huge upgrade, it has already been implemented on Intel3 (and would have been far enough along during Bartlett lake development to have been used). As we see on Arrow Lake U, it clocks high enough for good ST performance, but would have shined with all 12 cores fully loaded and the process advantage could show itself.
I think Bartlet Lake would have been more interesting on Intel3 with Redwood cove cores instead of raptor cove on 7.
They have the capacity on Intel 3 it's just that ramping that will cause them loss lol if it's not that big of a product.Intel doesn't have the capacity for something like that. And it's intended to be cheap.