First of all loading PDF is supported only by .NET and Python version of Aspose.Words. This feature is not available in C++ and Java versions.
Aspose.Words is designed to work with MS Word documents. MS Word documents are flow documents and they have structure very similar to Aspose.Words Document Object Model. But on the other hand PDF documents are fixed page format documents. While loading PDF document, Aspose.Words converts Fixed Page Document structure into the Flow Document Object Model. Unfortunately, such conversion does not guaranty 100% fidelity.
On other hand Aspose.PDF is designed to work with fixed page PDF documents.
Conversions from fixed page to flow document model is quite time consuming task. So most likely Aspose.PDF conversion performance will be better while conversion PDF to HTML. But the performance depends on the document size and complexity. So it is better to test both products with your real documents and see that product better fits your requirements.
Evaluation version does not limit the product performance. There are only two limitations in evaluation version - the maximum size of the document is limited to several hundreds of paragraphs and an evaluation watermark is injected into the document.
You can request a free 30-days temporary license to test both Aspose.Words and Aspose.PDF without evaluation version limitations.