Search results

Before the first benchmarks leaked that showed it having lower IPC JF-AMD kept saying "IPC increases!" When the details of the cache design (write-through has NEVER been high performance) it was clear higher IPC was simply not possible. I argued that point with him for quite a while and as more...

True, (IPC * clock rate) * (core count * thread scaling) is what I use :cool: The first part is per-thread performance, the second part is relative core count scaling, together we have the whole CPU performance. No doubt, it could also be 0% in five benchmarks, 100% in a few, and 70% in...

Even in the best case for AMD, Intel need not worry. Intel has being doing nothing on the CPU front beyond FPU improvements for certain large customers and process improvements to save themselves money on each chip produced. The good thing is that they haven't dropped their prices, which gives...

Market share actually doesn't matter as much as you think, it's all about the installed user base. AMD is not as far behind as it may seem, though they are certainly slipping badly. From the steam hardware survey it would appear they have about 25 to 30% of all graphics cards in current use...

That, however, is not how you do proper hardware comparisons. In science, for example, you eliminate every possible variable. If nVidia has a LOD setting of 3, and AMD has a higher quality setting of 0 (I know, totally seems backwards...the best numbers are negative values, LOL!) then you need...

AMD users are in a decidedly better position to take advantage of DX12. 7870XT, 7950/70 have asynchronous shaders ready for use with direct X 12. That's a four year old GPU. On another note, that same four-year old GPU has gained nearly 30% of additional performance due to driver...

Also, you can apparently do downsampling with the following: http://www.neogaf.com/forum/showthread.php?t=807472 Be careful, though, the scaler on the 280X will not work well after about 2600x1500.

What do you mean by 'poor drivers?' I've ever only had one problem with gaming on AMD hardware (well, since the x850 days, LOL... I think I had some problems with MechWarrior back then, IIRC) and that was with Microsoft Flight Sim X in Direct X 10 mode... and the problem is STILL not fixed...

My opinion of the two control panels is the opposite. I go back and forth between them and personally enjoy using AMD's CCC more than nVidia's control panel. Then again, nVidia's is more intuitive and things are a bit easier to find. The biggest complaint I have about nVidia (not sure if...

Turn Post FX down one notch, you won't notice any changes, but the glitches should be gone. No one knows why this happens, no matter very extensive efforts from all parties to try and discover the cause. It may never be repairable by AMD, but require a fix by RockStar - no one knows at this...

Same width, just faster.:D

Okay, since no one else seems to be answering the question you are actually asking, I'll explain. :D There are three main reasons why HBM memory is located adjacent to the GPU: 1. Over 1,000 connections per memory chip. -- Creating even 1,000 connections from a GPU on a PCB is expensive...

The shader code isn't where the problems are. It simply nVidia's excessive use of tessellation. Have you seen the Crysis 2 tessellation trickery? It does nothing... except slow down AMD cards more than nVidia cards.:o

The problem, as I understood it, wasn't a source code level problem at all, it was nVidia's license agreements preventing AMD from receiving the game in a timely manner while nVidia was consulting on the game. AMD would receive infrequent snapshots at best and they wouldn't include GW...

Bandwidth is only up 60% with HBM if you don't consider compression. If throughput has increased with GCN1.3 (likely), then that 45% increased need for bandwidth could easily be quite a bit greater.

Well, and the fact that Fiji uses memory compression, so should act like about 5GB of RAM (or more).

The importance of HBM for AMD's Fiji resides in the fact that Fiji, with 4096SPs, required 45% greater bandwidth to enable proper scaling. Without it, AMD's Fiji would scale very poorly. HBM adds 60% more bandwidth just when the GPU needs 45% more bandwidth to maintain its scaling. What's...

Fury [X] supports HDMI 2.0. Seriously, if you're going 4k, you need a top of the line GPU - NOT a second tier GPU like the 390/x. That would be like buying an R9 280 and trying to run 1600p... Or, trying to run 1080p with a GTX 550.

That's WAY lower than my R9 290 when it's at the same clocks: And the R9 390 also performs about 8.5% better from what I've seen - but my R9 290 performs exactly in line with online comparisons... Is it possible the R9 390 IS a refresh but the 390X is just an up-clocked 290X??? :eek: EDIT...

My R9 290 shows: 1. Device: Hawaiin 1.1 Hardware version: OpenCL 2.0 AMD-APP (1642.5)n 1.2 Softwa re version: 1642.5 (VM)n 1.3 OpenCL C version: OpenCL C 2.0 n 1.4 Parallel compu te units: 40n So I don't think we can go by the OpenCL version...

The evidence is actually quite confusing. The R9 390 outpaces the R9 290 at the same clocks by a surprising amount in FireStrike - between 4 and 8.5%... However, the reports from the 390X is that the performance isn't much more than 5% above the 390 - and are perfectly in line with just...

I'm going to definitely have to disagree about who had the worse drivers in the late 90s early 2000s. I built hundreds of systems using all manner of cards, including the good old Voodoo and no drivers ever gave me more troubles than the nVidia drivers. ESPECIALLY on upgrades or uninstalls for...

No, no it isn't. You don't seem to understand business at all, allow me to enlighten you a little. A business cares not one bit about their market share in terms of sales. They care about profit, profit, profit. If AMD sales 20% less hardware, but earns 1% more net profits as a result...

I think AMD has everything they need to match Haswell, or Sandy Bridge in the very least, in terms of IPC and net throughput... they just need to put it all together into a coherent package. They could take the very same elements in Excavator and reorganize them in the same fashion as the...

Server Bulldozer was shipped on time. Server OEMs got what they wanted when they wanted. H2/11 was the date for anticipated shipping for years.

Has anyone considered just removing the HC1011? Looking at the information and the schematics the end result would be ~45-50 v out on the sat side, more on the sub side, with the extra power merely going to waste in the end. You lose power protection, but you should then have a working...

Prior to upgrading to a new MOBO/CPU I purchased PCI-E Radeon HD 4670 and replaced my AGP Radeon x850 XTP and tried to replicate the issue. That proved very easy - flash in full screen was easy enough :-) I tried everything non-drastic without success. I trashed the enum tree... with...

The problem is rather simple, which doesn't make it any less annoying. Windows was, most likely, still blindly loading the AGP driver considering the AGP bus was still present. The driver loading obviously was a cause for conflict. Rather than formatting, you could have simply completely...